diff --git a/assets/grafana/_grafanalib.py b/assets/grafana/_grafanalib.py
index e1e121e538ee8c5c57492b22ca5a5ed1208ffd25..b304809d5735700876fe023f239ed91d3d5e5fd3 100644
--- a/assets/grafana/_grafanalib.py
+++ b/assets/grafana/_grafanalib.py
@@ -13,7 +13,7 @@ def Dashboard(
title=title, refresh=None, schemaVersion=14,
version=version, time=time, timezone='browser', inputs=[
{
- 'name': 'DS_PROMETHEUS',
+ 'name': 'prometheus',
'label': 'prometheus',
'description': '',
'type': 'datasource',
@@ -54,7 +54,7 @@ def SingleStat(
return core.SingleStat(
title=title, id=id, colorValue=colorValue,
- dataSource='${DS_PROMETHEUS}', gauge=gauge,
+ dataSource='prometheus', gauge=gauge,
valueFontSize=valueFontSize, thresholds=thresholds,
valueName=valueName, valueMaps=valueMaps, rangeMaps=rangeMaps,
mappingTypes=mappingTypes, targets=targets,
@@ -81,7 +81,7 @@ def Graph(
return core.Graph(
id=id, title=title, dashLength=dashLength, dashes=dashes,
spaceLength=spaceLength, targets=targets, xAxis=xAxis, yAxes=yAxes,
- dataSource='${DS_PROMETHEUS}', nullPointMode=nullPointMode, editable=False,
+ dataSource='prometheus', nullPointMode=nullPointMode, editable=False,
)
diff --git a/assets/grafana/deployment.dashboard.py b/assets/grafana/deployment.dashboard.py
index 4a2bda4674f9fbdcf6c52db01ca1b156d96a5aa8..6cecd4bf607e0436385f361e54c2f15878cdf38f 100644
--- a/assets/grafana/deployment.dashboard.py
+++ b/assets/grafana/deployment.dashboard.py
@@ -13,7 +13,7 @@ dashboard = Dashboard(
{
'allValue': '.*',
'current': {},
- 'datasource': '${DS_PROMETHEUS}',
+ 'datasource': 'prometheus',
'hide': 0,
'includeAll': False,
'label': 'Namespace',
@@ -34,7 +34,7 @@ dashboard = Dashboard(
{
'allValue': None,
'current': {},
- 'datasource': '${DS_PROMETHEUS}',
+ 'datasource': 'prometheus',
'hide': 0,
'includeAll': False,
'label': 'Deployment',
diff --git a/assets/grafana/kubernetes-capacity-planning.dashboard.py b/assets/grafana/kubernetes-capacity-planning.dashboard.py
index 00e5ada2876404042a573ec5ccda5a94a034166d..9b02010a960df2de13a9a7c0fee5d83b6087e822 100644
--- a/assets/grafana/kubernetes-capacity-planning.dashboard.py
+++ b/assets/grafana/kubernetes-capacity-planning.dashboard.py
@@ -13,7 +13,7 @@ dashboard = Dashboard(
timezone='browser',
inputs=[
{
- 'name': 'DS_PROMETHEUS',
+ 'name': 'prometheus',
'label': 'prometheus',
'description': '',
'type': 'datasource',
@@ -28,7 +28,7 @@ dashboard = Dashboard(
Graph(
title='Idle CPU',
id=3,
- dataSource='${DS_PROMETHEUS}',
+ dataSource='prometheus',
dashLength=10,
dashes=False,
isNew=False,
@@ -55,7 +55,7 @@ dashboard = Dashboard(
Graph(
title='System Load',
id=9,
- dataSource='${DS_PROMETHEUS}',
+ dataSource='prometheus',
dashLength=10,
dashes=False,
isNew=False,
@@ -102,7 +102,7 @@ dashboard = Dashboard(
Graph(
title='Memory Usage',
id=4,
- dataSource='${DS_PROMETHEUS}',
+ dataSource='prometheus',
dashLength=10,
dashes=False,
isNew=False,
@@ -170,7 +170,7 @@ dashboard = Dashboard(
),
SingleStat(
title='Memory Usage',
- dataSource='${DS_PROMETHEUS}',
+ dataSource='prometheus',
id=5,
format='percent',
span=3,
@@ -212,7 +212,7 @@ dashboard = Dashboard(
titleSize='h6', panels=[
Graph(
title='Disk I/O',
- dataSource='${DS_PROMETHEUS}',
+ dataSource='prometheus',
id=6,
dashLength=10,
dashes=False,
@@ -267,7 +267,7 @@ dashboard = Dashboard(
),
SingleStat(
title='Disk Space Usage',
- dataSource='${DS_PROMETHEUS}',
+ dataSource='prometheus',
id=12,
span=3,
editable=False,
@@ -312,7 +312,7 @@ dashboard = Dashboard(
titleSize='h6', panels=[
Graph(
title='Network Received',
- dataSource='${DS_PROMETHEUS}',
+ dataSource='prometheus',
id=8,
dashLength=10,
dashes=False,
@@ -346,7 +346,7 @@ dashboard = Dashboard(
),
Graph(
title='Network Transmitted',
- dataSource='${DS_PROMETHEUS}',
+ dataSource='prometheus',
id=10,
dashLength=10,
dashes=False,
@@ -386,7 +386,7 @@ dashboard = Dashboard(
panels=[
Graph(
title='Cluster Pod Utilization',
- dataSource='${DS_PROMETHEUS}',
+ dataSource='prometheus',
id=11,
span=9,
dashes=False,
@@ -421,7 +421,7 @@ dashboard = Dashboard(
),
SingleStat(
title='Pod Utilization',
- dataSource='${DS_PROMETHEUS}',
+ dataSource='prometheus',
id=7,
editable=False,
span=3,
diff --git a/assets/grafana/kubernetes-cluster-health.dashboard.py b/assets/grafana/kubernetes-cluster-health.dashboard.py
index dbd402f385684703d620ad71b441d7b14550634d..7f1cfe647fdc5376157b4e4d9be8fcfd721884dc 100644
--- a/assets/grafana/kubernetes-cluster-health.dashboard.py
+++ b/assets/grafana/kubernetes-cluster-health.dashboard.py
@@ -11,7 +11,7 @@ dashboard = Dashboard(
timezone='browser',
inputs=[
{
- 'name': 'DS_PROMETHEUS',
+ 'name': 'prometheus',
'label': 'prometheus',
'description': '',
'type': 'datasource',
@@ -26,7 +26,7 @@ dashboard = Dashboard(
SingleStat(
title='Control Plane Components Down',
id=1,
- dataSource='${DS_PROMETHEUS}',
+ dataSource='prometheus',
gauge=Gauge(),
span=3,
thresholds='1, 3',
@@ -76,7 +76,7 @@ dashboard = Dashboard(
SingleStat(
title='Alerts Firing',
id=2,
- dataSource='${DS_PROMETHEUS}',
+ dataSource='prometheus',
gauge=Gauge(),
colorValue=True,
editable=False,
@@ -122,7 +122,7 @@ dashboard = Dashboard(
SingleStat(
title='Alerts Pending',
id=3,
- dataSource='${DS_PROMETHEUS}',
+ dataSource='prometheus',
gauge=Gauge(),
colorValue=True,
editable=False,
@@ -168,7 +168,7 @@ dashboard = Dashboard(
SingleStat(
title='Crashlooping Pods',
id=4,
- dataSource='${DS_PROMETHEUS}',
+ dataSource='prometheus',
gauge=Gauge(),
colorValue=True,
editable=False,
@@ -219,7 +219,7 @@ dashboard = Dashboard(
SingleStat(
title='Node Not Ready',
id=5,
- dataSource='${DS_PROMETHEUS}',
+ dataSource='prometheus',
gauge=Gauge(),
colorValue=True,
editable=False,
@@ -265,7 +265,7 @@ dashboard = Dashboard(
SingleStat(
title='Node Disk Pressure',
id=6,
- dataSource='${DS_PROMETHEUS}',
+ dataSource='prometheus',
gauge=Gauge(),
colorValue=True,
editable=False,
@@ -311,7 +311,7 @@ dashboard = Dashboard(
SingleStat(
title='Node Memory Pressure',
id=7,
- dataSource='${DS_PROMETHEUS}',
+ dataSource='prometheus',
gauge=Gauge(),
colorValue=True,
editable=False,
@@ -357,7 +357,7 @@ dashboard = Dashboard(
SingleStat(
title='Nodes Unschedulable',
id=8,
- dataSource='${DS_PROMETHEUS}',
+ dataSource='prometheus',
gauge=Gauge(),
colorValue=True,
editable=False,
diff --git a/assets/grafana/kubernetes-control-plane-status.dashboard.py b/assets/grafana/kubernetes-control-plane-status.dashboard.py
index fad157f5986dc6c143e89d2df38cc495c53535ff..d2f351296564c41652c0a3d26ef15d45196e9b53 100644
--- a/assets/grafana/kubernetes-control-plane-status.dashboard.py
+++ b/assets/grafana/kubernetes-control-plane-status.dashboard.py
@@ -11,7 +11,7 @@ dashboard = Dashboard(
editable=False,
inputs=[
{
- 'name': 'DS_PROMETHEUS',
+ 'name': 'prometheus',
'label': 'prometheus',
'description': '',
'type': 'datasource',
@@ -25,7 +25,7 @@ dashboard = Dashboard(
panels=[
SingleStat(
title='API Servers UP',
- dataSource='${DS_PROMETHEUS}',
+ dataSource='prometheus',
format='percent',
editable=False,
gauge=Gauge(
@@ -76,7 +76,7 @@ dashboard = Dashboard(
),
SingleStat(
title='Controller Managers UP',
- dataSource='${DS_PROMETHEUS}',
+ dataSource='prometheus',
format='percent',
editable=False,
gauge=Gauge(
@@ -128,7 +128,7 @@ dashboard = Dashboard(
),
SingleStat(
title='Schedulers UP',
- dataSource='${DS_PROMETHEUS}',
+ dataSource='prometheus',
format='percent',
editable=False,
gauge=Gauge(
@@ -179,7 +179,7 @@ dashboard = Dashboard(
),
SingleStat(
title='API Server Request Error Rate',
- dataSource='${DS_PROMETHEUS}',
+ dataSource='prometheus',
format='percent',
editable=False,
gauge=Gauge(
@@ -234,7 +234,7 @@ dashboard = Dashboard(
Graph(
title='API Server Request Latency',
id=7,
- dataSource='${DS_PROMETHEUS}',
+ dataSource='prometheus',
dashLength=10,
dashes=False,
isNew=False,
@@ -269,7 +269,7 @@ dashboard = Dashboard(
Graph(
title='End to End Scheduling Latency',
id=5,
- dataSource='${DS_PROMETHEUS}',
+ dataSource='prometheus',
isNew=False,
editable=False,
dashLength=10,
@@ -300,7 +300,7 @@ dashboard = Dashboard(
Graph(
title='API Server Request Rates',
id=6,
- dataSource='${DS_PROMETHEUS}',
+ dataSource='prometheus',
isNew=False,
editable=False,
dashLength=10,
diff --git a/assets/grafana/kubernetes-resource-requests.dashboard.py b/assets/grafana/kubernetes-resource-requests.dashboard.py
index 6e4c510ce19c0d1b05a0a51218fdb6bf34920ce8..5d5b3bd8c7be1a9d715b9b63aee143f8a202d1f1 100644
--- a/assets/grafana/kubernetes-resource-requests.dashboard.py
+++ b/assets/grafana/kubernetes-resource-requests.dashboard.py
@@ -12,7 +12,7 @@ dashboard = Dashboard(
timezone='browser',
inputs=[
{
- 'name': 'DS_PROMETHEUS',
+ 'name': 'prometheus',
'label': 'prometheus',
'description': '',
'type': 'datasource',
@@ -34,7 +34,7 @@ dashboard = Dashboard(
'community/blob/master/contributors/design-proposals/'
'node-allocatable.md) is also shown.',
id=1,
- dataSource='${DS_PROMETHEUS}',
+ dataSource='prometheus',
dashLength=10,
dashes=False,
isNew=False,
@@ -73,7 +73,7 @@ dashboard = Dashboard(
),
SingleStat(
title='CPU Cores',
- dataSource='${DS_PROMETHEUS}',
+ dataSource='prometheus',
id=2,
format='percent',
editable=False,
@@ -117,7 +117,7 @@ dashboard = Dashboard(
Graph(
title='Memory',
id=3,
- dataSource='${DS_PROMETHEUS}',
+ dataSource='prometheus',
description='This represents the total [memory resource '
'requests](https://kubernetes.io/docs/concepts/'
'configuration/manage-compute-resources-container/'
@@ -163,7 +163,7 @@ dashboard = Dashboard(
),
SingleStat(
title='Memory',
- dataSource='${DS_PROMETHEUS}',
+ dataSource='prometheus',
id=4,
format='percent',
span=3,
diff --git a/assets/grafana/nodes.dashboard.py b/assets/grafana/nodes.dashboard.py
index 89d2b1f9cbeb911f64b8ad3488198e667457072e..da7b7d247834316c06e0218757da3964f70b60c0 100644
--- a/assets/grafana/nodes.dashboard.py
+++ b/assets/grafana/nodes.dashboard.py
@@ -14,7 +14,7 @@ dashboard = Dashboard(
timezone='browser',
inputs=[
{
- 'name': 'DS_PROMETHEUS',
+ 'name': 'prometheus',
'label': 'prometheus',
'description': '',
'type': 'datasource',
@@ -26,7 +26,7 @@ dashboard = Dashboard(
{
'allValue': None,
'current': {},
- 'datasource': '${DS_PROMETHEUS}',
+ 'datasource': 'prometheus',
'hide': 0,
'includeAll': False,
'label': None,
@@ -50,7 +50,7 @@ dashboard = Dashboard(
titleSize='h6', panels=[
Graph(
title='Idle CPU',
- dataSource='${DS_PROMETHEUS}',
+ dataSource='prometheus',
id=3,
isNew=False,
editable=False,
@@ -82,7 +82,7 @@ dashboard = Dashboard(
),
Graph(
title='System Load',
- dataSource='${DS_PROMETHEUS}',
+ dataSource='prometheus',
id=9,
isNew=False,
editable=False,
@@ -129,7 +129,7 @@ dashboard = Dashboard(
titleSize='h6', panels=[
Graph(
title='Memory Usage',
- dataSource='${DS_PROMETHEUS}',
+ dataSource='prometheus',
id=4,
isNew=False,
editable=False,
@@ -198,7 +198,7 @@ dashboard = Dashboard(
),
SingleStat(
title='Memory Usage',
- dataSource='${DS_PROMETHEUS}',
+ dataSource='prometheus',
id=5,
format='percent',
gauge=Gauge(show=True),
@@ -241,7 +241,7 @@ dashboard = Dashboard(
showTitle=False, panels=[
Graph(
title='Disk I/O',
- dataSource='${DS_PROMETHEUS}',
+ dataSource='prometheus',
id=6,
dashLength=10,
dashes=False,
@@ -304,7 +304,7 @@ dashboard = Dashboard(
),
SingleStat(
title='Disk Space Usage',
- dataSource='${DS_PROMETHEUS}',
+ dataSource='prometheus',
id=7,
thresholds='0.75, 0.9',
editable=False,
@@ -351,7 +351,7 @@ dashboard = Dashboard(
panels=[
Graph(
title='Network Received',
- dataSource='${DS_PROMETHEUS}',
+ dataSource='prometheus',
id=8,
dashLength=10,
dashes=False,
@@ -385,7 +385,7 @@ dashboard = Dashboard(
),
Graph(
title='Network Transmitted',
- dataSource='${DS_PROMETHEUS}',
+ dataSource='prometheus',
id=10,
dashLength=10,
dashes=False,
diff --git a/assets/grafana/pods.dashboard.py b/assets/grafana/pods.dashboard.py
index f5258cbce78f3205bed95db30870956a96178fab..84b3fdefcb416ff8b5beb89c2c3e535a5c0425ad 100644
--- a/assets/grafana/pods.dashboard.py
+++ b/assets/grafana/pods.dashboard.py
@@ -12,7 +12,7 @@ dashboard = Dashboard(
timezone='browser',
inputs=[
{
- 'name': 'DS_PROMETHEUS',
+ 'name': 'prometheus',
'label': 'prometheus',
'description': '',
'type': 'datasource',
@@ -24,7 +24,7 @@ dashboard = Dashboard(
{
'allValue': '.*',
'current': {},
- 'datasource': '${DS_PROMETHEUS}',
+ 'datasource': 'prometheus',
'hide': 0,
'includeAll': True,
'label': 'Namespace',
@@ -44,7 +44,7 @@ dashboard = Dashboard(
{
'allValue': None,
'current': {},
- 'datasource': '${DS_PROMETHEUS}',
+ 'datasource': 'prometheus',
'hide': 0,
'includeAll': False,
'label': 'Pod',
@@ -65,7 +65,7 @@ dashboard = Dashboard(
{
'allValue': '.*',
'current': {},
- 'datasource': '${DS_PROMETHEUS}',
+ 'datasource': 'prometheus',
'hide': 0,
'includeAll': True,
'label': 'Container',
@@ -90,7 +90,7 @@ dashboard = Dashboard(
titleSize='h6', panels=[
Graph(
title='Memory Usage',
- dataSource='${DS_PROMETHEUS}',
+ dataSource='prometheus',
id=1,
isNew=False,
editable=False,
@@ -155,7 +155,7 @@ dashboard = Dashboard(
titleSize='h6', panels=[
Graph(
title='CPU Usage',
- dataSource='${DS_PROMETHEUS}',
+ dataSource='prometheus',
id=2,
isNew=False,
editable=False,
@@ -218,7 +218,7 @@ dashboard = Dashboard(
titleSize='h6', panels=[
Graph(
title='Network I/O',
- dataSource='${DS_PROMETHEUS}',
+ dataSource='prometheus',
id=3,
isNew=False,
editable=False,
diff --git a/assets/grafana/raw-json-dashboards/etcd-dashboard.json b/assets/grafana/raw-json-dashboards/etcd-dashboard.json
index 0098ffeac6bb7672f0c8523b65a82f759a59d486..f2a03cec5fc7576649dc37fa0d46866ec1060f5c 100644
--- a/assets/grafana/raw-json-dashboards/etcd-dashboard.json
+++ b/assets/grafana/raw-json-dashboards/etcd-dashboard.json
@@ -1,7 +1,7 @@
{
"__inputs": [
{
- "name": "DS_PROMETHEUS",
+ "name": "prometheus",
"label": "prometheus",
"description": "",
"type": "datasource",
@@ -60,7 +60,7 @@
"rgba(237, 129, 40, 0.89)",
"rgba(50, 172, 45, 0.97)"
],
- "datasource": "${DS_PROMETHEUS}",
+ "datasource": "prometheus",
"editable": false,
"error": false,
"format": "none",
@@ -136,7 +136,7 @@
"bars": false,
"dashLength": 10,
"dashes": false,
- "datasource": "${DS_PROMETHEUS}",
+ "datasource": "prometheus",
"editable": false,
"error": false,
"fill": 0,
@@ -225,7 +225,7 @@
"bars": false,
"dashLength": 10,
"dashes": false,
- "datasource": "${DS_PROMETHEUS}",
+ "datasource": "prometheus",
"editable": false,
"error": false,
"fill": 0,
@@ -326,7 +326,7 @@
"bars": false,
"dashLength": 10,
"dashes": false,
- "datasource": "${DS_PROMETHEUS}",
+ "datasource": "prometheus",
"decimals": null,
"editable": false,
"error": false,
@@ -408,7 +408,7 @@
"bars": false,
"dashLength": 10,
"dashes": false,
- "datasource": "${DS_PROMETHEUS}",
+ "datasource": "prometheus",
"editable": false,
"error": false,
"fill": 0,
@@ -497,7 +497,7 @@
"bars": false,
"dashLength": 10,
"dashes": false,
- "datasource": "${DS_PROMETHEUS}",
+ "datasource": "prometheus",
"editable": false,
"error": false,
"fill": 0,
@@ -589,7 +589,7 @@
"bars": false,
"dashLength": 10,
"dashes": false,
- "datasource": "${DS_PROMETHEUS}",
+ "datasource": "prometheus",
"editable": false,
"error": false,
"fill": 5,
@@ -669,7 +669,7 @@
"bars": false,
"dashLength": 10,
"dashes": false,
- "datasource": "${DS_PROMETHEUS}",
+ "datasource": "prometheus",
"editable": false,
"error": false,
"fill": 5,
@@ -749,7 +749,7 @@
"bars": false,
"dashLength": 10,
"dashes": false,
- "datasource": "${DS_PROMETHEUS}",
+ "datasource": "prometheus",
"editable": false,
"error": false,
"fill": 0,
@@ -829,7 +829,7 @@
"bars": false,
"dashLength": 10,
"dashes": false,
- "datasource": "${DS_PROMETHEUS}",
+ "datasource": "prometheus",
"decimals": null,
"editable": false,
"error": false,
@@ -923,7 +923,7 @@
"bars": false,
"dashLength": 10,
"dashes": false,
- "datasource": "${DS_PROMETHEUS}",
+ "datasource": "prometheus",
"editable": false,
"error": false,
"fill": 0,
@@ -1029,7 +1029,7 @@
"bars": false,
"dashLength": 10,
"dashes": false,
- "datasource": "${DS_PROMETHEUS}",
+ "datasource": "prometheus",
"decimals": 0,
"editable": false,
"error": false,
diff --git a/assets/grafana/statefulset.dashboard.py b/assets/grafana/statefulset.dashboard.py
index a28d70fdf9f5ffb4b6eeb052d34fe19dc62e352f..780630a28741ab752e3907ec81df88c04d931399 100644
--- a/assets/grafana/statefulset.dashboard.py
+++ b/assets/grafana/statefulset.dashboard.py
@@ -13,7 +13,7 @@ dashboard = Dashboard(
{
'allValue': '.*',
'current': {},
- 'datasource': '${DS_PROMETHEUS}',
+ 'datasource': 'prometheus',
'hide': 0,
'includeAll': False,
'label': 'Namespace',
@@ -34,7 +34,7 @@ dashboard = Dashboard(
{
'allValue': None,
'current': {},
- 'datasource': '${DS_PROMETHEUS}',
+ 'datasource': 'prometheus',
'hide': 0,
'includeAll': False,
'label': 'StatefulSet',
diff --git a/grafana-image/Dockerfile b/grafana-image/Dockerfile
index bac01b592e86bec53394f2692a8011d1e4190d54..fdbef2895290b507ed743844e436565875584b6a 100644
--- a/grafana-image/Dockerfile
+++ b/grafana-image/Dockerfile
@@ -1,15 +1,17 @@
FROM debian:9.3-slim
+ARG GRAFANA_VERSION
+
RUN apt-get update && apt-get install -qq -y wget tar sqlite && \
- wget -O /tmp/grafana.tar.gz https://s3-us-west-2.amazonaws.com/grafana-releases/release/grafana-4.6.3.linux-x64.tar.gz && \
- tar -zxvf /tmp/grafana.tar.gz -C /tmp && mv /tmp/grafana-4.6.3 /grafana && \
+ wget -O /tmp/grafana.tar.gz https://s3-us-west-2.amazonaws.com/grafana-releases/release/grafana-$GRAFANA_VERSION.linux-x64.tar.gz && \
+ tar -zxvf /tmp/grafana.tar.gz -C /tmp && mv /tmp/grafana-$GRAFANA_VERSION /grafana && \
rm -rf /tmp/grafana.tar.gz
-ADD config.toml /grafana/conf/config.toml
+ADD config.ini /grafana/conf/config.ini
USER nobody
EXPOSE 3000
VOLUME [ "/data" ]
WORKDIR /grafana
ENTRYPOINT [ "/grafana/bin/grafana-server" ]
-CMD [ "-config=/grafana/conf/config.toml" ]
+CMD [ "-config=/grafana/conf/config.ini" ]
diff --git a/grafana-image/Makefile b/grafana-image/Makefile
index f054a0544daf8c4deeebc81f88b3cb28b1af7744..9b352c170e113dccff3ab6d7442bcd3869d451e2 100644
--- a/grafana-image/Makefile
+++ b/grafana-image/Makefile
@@ -1,2 +1,5 @@
+VERSION=5.0.0-beta1
+IMAGE_TAG=$(VERSION)
+
container:
- docker build . -t quay.io/coreos/monitoring-grafana:4.6.3-non-root.1
+ docker build --build-arg GRAFANA_VERSION=$(VERSION) -t quay.io/coreos/monitoring-grafana:$(IMAGE_TAG) .
diff --git a/grafana-image/config.toml b/grafana-image/config.ini
similarity index 51%
rename from grafana-image/config.toml
rename to grafana-image/config.ini
index 74f5d78b5e5c066c4728c646fa6db9a651b6c977..14e0c7adc25e9f9d26a05252062bbde103accff5 100644
--- a/grafana-image/config.toml
+++ b/grafana-image/config.ini
@@ -5,3 +5,12 @@ path = /data/grafana.db
data = /data
logs = /data/log
plugins = /data/plugins
+
+[session]
+provider = memory
+
+[auth.basic]
+enabled = false
+
+[auth.anonymous]
+enabled = true
diff --git a/hack/grafana-dashboards-configmap-generator/bin/grafana_dashboards_generate.sh b/hack/grafana-dashboards-configmap-generator/bin/grafana_dashboards_generate.sh
index b4273baa707497a6cdfe3eb3fa52435f10247427..d89cfc97f102e7bc621f2a7ecb98b2428f67aeea 100755
--- a/hack/grafana-dashboards-configmap-generator/bin/grafana_dashboards_generate.sh
+++ b/hack/grafana-dashboards-configmap-generator/bin/grafana_dashboards_generate.sh
@@ -70,7 +70,7 @@ DATE_EXEC="$(date "+%Y-%m-%d-%H%M%S")"
BIN_DIR="$( cd "$( dirname "${BASH_SOURCE[0]}" )" && pwd )"
TOOL_HOME="$(dirname $BIN_DIR)"
SCRIPT_BASE=`basename $0 | sed "s/\.[Ss][Hh]//"`
-CONFIGMAP_DASHBOARD_PREFIX="grafana-dashboards"
+CONFIGMAP_DASHBOARD_PREFIX="grafana-dashboard-definitions"
TEMPLATES_DIR="$TOOL_HOME/templates"
DASHBOARD_HEADER_FILE="$TEMPLATES_DIR/dashboard.header"
@@ -327,7 +327,10 @@ initialize-bin-pack
bin-pack-files "$(find $DASHBOARDS_DIR -maxdepth 1 -type f -name "*-dashboard.json" | sort)"
# Continue processing datasources (maintaining the same queue)
-bin-pack-files "$(find $DASHBOARDS_DIR -maxdepth 1 -type f -name "*-datasource.json" | sort )"
+#
+# Commented out, as datasources are provisionable by Grafana by default in Grafana v5, but from a separate directory, meaning a separate ConfigMap for us.
+#
+# bin-pack-files "$(find $DASHBOARDS_DIR -maxdepth 1 -type f -name "*-datasource.json" | sort )"
# Processing remaining data in the queue (or unique)
if [ "$to_process" ]; then
@@ -361,7 +364,7 @@ for (( i=0; i<$total_configmaps_created; i++ )); do
configmap="$CONFIGMAP_DASHBOARD_PREFIX-$i"
echo "# Preparing grafana deployment to support configmap: $configmap"
- test "$VOLUME_MOUNTS" && VOLUME_MOUNTS="$VOLUME_MOUNTS\n- name: $configmap\n mountPath: /var/$configmap" || VOLUME_MOUNTS="- name: $configmap\n mountPath: /var/$configmap"
+ test "$VOLUME_MOUNTS" && VOLUME_MOUNTS="$VOLUME_MOUNTS\n- name: $configmap\n mountPath: /grafana-dashboard-definitions/$i" || VOLUME_MOUNTS="- name: $configmap\n mountPath: /grafana-dashboard-definitions/$i"
test "$VOLUMES" && VOLUMES="$VOLUMES\n- name: $configmap\n configMap:\n name: $configmap" || VOLUMES="- name: $configmap\n configMap:\n name: $configmap"
test "$WATCH_DIR" && WATCH_DIR="$WATCH_DIR\n- '--watch-dir=/var/$configmap'" || WATCH_DIR="- '--watch-dir=/var/$configmap'"
# echo "DEBUG:"
diff --git a/hack/grafana-dashboards-configmap-generator/templates/ConfigMap.header b/hack/grafana-dashboards-configmap-generator/templates/ConfigMap.header
index afc1f42cbfe862f18eaf5d7e0e19750687f59218..73a14b05be083d9e82d849ac0c22177815f791c6 100644
--- a/hack/grafana-dashboards-configmap-generator/templates/ConfigMap.header
+++ b/hack/grafana-dashboards-configmap-generator/templates/ConfigMap.header
@@ -1,5 +1,5 @@
apiVersion: v1
kind: ConfigMap
metadata:
- name: grafana-dashboards
+ name: grafana-dashboard-definitions
data:
diff --git a/hack/grafana-dashboards-configmap-generator/templates/dashboard.foot b/hack/grafana-dashboards-configmap-generator/templates/dashboard.foot
index 81fe9f6fd7de318bb42f0e6884bcf9e6332740ac..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391 100644
--- a/hack/grafana-dashboards-configmap-generator/templates/dashboard.foot
+++ b/hack/grafana-dashboards-configmap-generator/templates/dashboard.foot
@@ -1,11 +0,0 @@
- ,
- "inputs": [
- {
- "name": "DS_PROMETHEUS",
- "pluginId": "prometheus",
- "type": "datasource",
- "value": "prometheus"
- }
- ],
- "overwrite": true
- }
diff --git a/hack/grafana-dashboards-configmap-generator/templates/dashboard.header b/hack/grafana-dashboards-configmap-generator/templates/dashboard.header
index 807e5d38caf69d3642c74e088a527df7192c1df6..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391 100644
--- a/hack/grafana-dashboards-configmap-generator/templates/dashboard.header
+++ b/hack/grafana-dashboards-configmap-generator/templates/dashboard.header
@@ -1,2 +0,0 @@
- {
- "dashboard":
diff --git a/hack/grafana-dashboards-configmap-generator/templates/grafana-deployment-template.yaml b/hack/grafana-dashboards-configmap-generator/templates/grafana-deployment-template.yaml
index a607f353253fdca15f723f03e30a02f5bffd1b84..fae45b25c6628f00d29cf2e87fcf8e551cc262b0 100644
--- a/hack/grafana-dashboards-configmap-generator/templates/grafana-deployment-template.yaml
+++ b/hack/grafana-dashboards-configmap-generator/templates/grafana-deployment-template.yaml
@@ -14,25 +14,15 @@ spec:
runAsUser: 65534
containers:
- name: grafana
- image: quay.io/coreos/monitoring-grafana:4.6.3-non-root.1
- env:
- - name: GF_AUTH_BASIC_ENABLED
- value: "true"
- - name: GF_AUTH_ANONYMOUS_ENABLED
- value: "true"
- - name: GF_SECURITY_ADMIN_USER
- valueFrom:
- secretKeyRef:
- name: grafana-credentials
- key: user
- - name: GF_SECURITY_ADMIN_PASSWORD
- valueFrom:
- secretKeyRef:
- name: grafana-credentials
- key: password
+ image: quay.io/coreos/monitoring-grafana:5.0.0-beta1
volumeMounts:
- name: grafana-storage
mountPath: /data
+ - name: grafana-datasources
+ mountPath: /grafana/conf/provisioning/datasources
+ - name: grafana-dashboards
+ mountPath: /grafana/conf/provisioning/dashboards
+XXX_VOLUME_MOUNTS_XXX
ports:
- name: web
containerPort: 3000
@@ -43,32 +33,13 @@ spec:
limits:
memory: 200Mi
cpu: 200m
- - name: grafana-watcher
- image: quay.io/coreos/grafana-watcher:v0.0.8
- args:
-XXX_WATCH_DIR_XXX
- - '--grafana-url=http://localhost:3000'
- env:
- - name: GRAFANA_USER
- valueFrom:
- secretKeyRef:
- name: grafana-credentials
- key: user
- - name: GRAFANA_PASSWORD
- valueFrom:
- secretKeyRef:
- name: grafana-credentials
- key: password
- resources:
- requests:
- memory: "16Mi"
- cpu: "50m"
- limits:
- memory: "32Mi"
- cpu: "100m"
- volumeMounts:
-XXX_VOLUME_MOUNTS_XXX
volumes:
- name: grafana-storage
emptyDir: {}
+ - name: grafana-datasources
+ configMap:
+ name: grafana-datasources
+ - name: grafana-dashboards
+ configMap:
+ name: grafana-dashboards
XXX_VOLUMES_XXX
diff --git a/hack/scripts/generate-dashboards-configmap.sh b/hack/scripts/generate-dashboards-configmap.sh
index c023874e9d1df03f4ee294656b21434304eea8bf..81c157e7ce47bb4edb1abe84d725b03058d472fa 100755
--- a/hack/scripts/generate-dashboards-configmap.sh
+++ b/hack/scripts/generate-dashboards-configmap.sh
@@ -6,7 +6,7 @@ cat <<-EOF
apiVersion: v1
kind: ConfigMap
metadata:
- name: grafana-dashboards-0
+ name: grafana-dashboard-definitions-0
data:
EOF
@@ -37,10 +37,3 @@ do
hack/scripts/wrap-dashboard.sh $f prometheus-k8s | sed "s/^/ /g"
fi
done
-
-for f in assets/grafana/*-datasource.json
-do
- cp $f assets/grafana/generated/
- echo " $(basename $f): |+"
- cat $f | sed "s/^/ /g"
-done
diff --git a/hack/scripts/generate-manifests.sh b/hack/scripts/generate-manifests.sh
index 4826864c1a2ef8b4c5fb361bf647d5c4aecba5cc..b6f35a2df589b310c70f5dfbd95881230c936771 100755
--- a/hack/scripts/generate-manifests.sh
+++ b/hack/scripts/generate-manifests.sh
@@ -6,16 +6,16 @@ set +x
hack/scripts/generate-rules-configmap.sh > manifests/prometheus/prometheus-k8s-rules.yaml
# Generate Dashboard ConfigMap
-hack/scripts/generate-dashboards-configmap.sh > manifests/grafana/grafana-dashboards.yaml
+hack/scripts/generate-dashboards-configmap.sh > manifests/grafana/grafana-dashboard-definitions.yaml
# Generate Dashboard ConfigMap with configmap-generator tool
# Max Size per ConfigMap: 240000
# Input dir: assets/grafana
# output file: manifests/grafana/grafana-dashboards.yaml
# grafana deployment output file: manifests/grafana/grafana-deployment.yaml
-test -f manifests/grafana/grafana-dashboards.yaml && rm -f manifests/grafana/grafana-dashboards.yaml
+test -f manifests/grafana/grafana-dashboards.yaml && rm -f manifests/grafana/grafana-dashboard-definitions.yaml
test -f manifests/grafana/grafana-deployment.yaml && rm -f manifests/grafana/grafana-deployment.yaml
-hack/grafana-dashboards-configmap-generator/bin/grafana_dashboards_generate.sh -s 240000 -i assets/grafana/generated -o manifests/grafana/grafana-dashboards.yaml -g manifests/grafana/grafana-deployment.yaml
+hack/grafana-dashboards-configmap-generator/bin/grafana_dashboards_generate.sh -s 240000 -i assets/grafana/generated -o manifests/grafana/grafana-dashboard-definitions.yaml -g manifests/grafana/grafana-deployment.yaml
# Generate Grafana Credentials Secret
hack/scripts/generate-grafana-credentials-secret.sh admin admin > manifests/grafana/grafana-credentials.yaml
diff --git a/manifests/grafana/grafana-dashboard-definitions.yaml b/manifests/grafana/grafana-dashboard-definitions.yaml
new file mode 100644
index 0000000000000000000000000000000000000000..dbd2a30d47eeb977dab15bdca78d9b95abe646f4
--- /dev/null
+++ b/manifests/grafana/grafana-dashboard-definitions.yaml
@@ -0,0 +1,7360 @@
+apiVersion: v1
+kind: ConfigMap
+metadata:
+ name: grafana-dashboard-definitions-0
+data:
+ deployment-dashboard.json: |+
+ {
+ "__inputs": [
+ {
+ "description": "",
+ "label": "prometheus",
+ "name": "prometheus",
+ "pluginId": "prometheus",
+ "pluginName": "Prometheus",
+ "type": "datasource"
+ }
+ ],
+ "annotations": {
+ "list": []
+ },
+ "editable": false,
+ "graphTooltip": 1,
+ "hideControls": false,
+ "links": [],
+ "rows": [
+ {
+ "collapse": false,
+ "editable": false,
+ "height": "200px",
+ "panels": [
+ {
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "rgba(245, 54, 54, 0.9)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(50, 172, 45, 0.97)"
+ ],
+ "datasource": "prometheus",
+ "editable": false,
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "id": 8,
+ "links": [],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "postfix": "cores",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 4,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": true
+ },
+ "targets": [
+ {
+ "expr": "sum(rate(container_cpu_usage_seconds_total{namespace=\"$deployment_namespace\",pod_name=~\"$deployment_name.*\"}[3m]))",
+ "intervalFactor": 2,
+ "refId": "A",
+ "step": 600
+ }
+ ],
+ "title": "CPU",
+ "type": "singlestat",
+ "valueFontSize": "110%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "avg"
+ },
+ {
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "rgba(245, 54, 54, 0.9)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(50, 172, 45, 0.97)"
+ ],
+ "datasource": "prometheus",
+ "editable": false,
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "id": 9,
+ "links": [],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "postfix": "GB",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "80%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 4,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": true
+ },
+ "targets": [
+ {
+ "expr": "sum(container_memory_usage_bytes{namespace=\"$deployment_namespace\",pod_name=~\"$deployment_name.*\"}) / 1024^3",
+ "intervalFactor": 2,
+ "refId": "A",
+ "step": 600
+ }
+ ],
+ "title": "Memory",
+ "type": "singlestat",
+ "valueFontSize": "110%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "avg"
+ },
+ {
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "rgba(245, 54, 54, 0.9)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(50, 172, 45, 0.97)"
+ ],
+ "datasource": "prometheus",
+ "editable": false,
+ "format": "Bps",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": false
+ },
+ "id": 7,
+ "links": [],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 4,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": true
+ },
+ "targets": [
+ {
+ "expr": "sum(rate(container_network_transmit_bytes_total{namespace=\"$deployment_namespace\",pod_name=~\"$deployment_name.*\"}[3m])) + sum(rate(container_network_receive_bytes_total{namespace=\"$deployment_namespace\",pod_name=~\"$deployment_name.*\"}[3m]))",
+ "intervalFactor": 2,
+ "refId": "A",
+ "step": 600
+ }
+ ],
+ "title": "Network",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "avg"
+ }
+ ],
+ "showTitle": false,
+ "title": "Dashboard Row",
+ "titleSize": "h6"
+ },
+ {
+ "collapse": false,
+ "editable": false,
+ "height": "100px",
+ "panels": [
+ {
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "rgba(245, 54, 54, 0.9)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(50, 172, 45, 0.97)"
+ ],
+ "datasource": "prometheus",
+ "editable": false,
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": false
+ },
+ "id": 5,
+ "links": [],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 3,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "targets": [
+ {
+ "expr": "max(kube_deployment_spec_replicas{deployment=\"$deployment_name\",namespace=\"$deployment_namespace\"}) without (instance, pod)",
+ "intervalFactor": 2,
+ "metric": "kube_deployment_spec_replicas",
+ "refId": "A",
+ "step": 600
+ }
+ ],
+ "title": "Desired Replicas",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "avg"
+ },
+ {
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "rgba(245, 54, 54, 0.9)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(50, 172, 45, 0.97)"
+ ],
+ "datasource": "prometheus",
+ "editable": false,
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "id": 6,
+ "links": [],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 3,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "targets": [
+ {
+ "expr": "min(kube_deployment_status_replicas_available{deployment=\"$deployment_name\",namespace=\"$deployment_namespace\"}) without (instance, pod)",
+ "intervalFactor": 2,
+ "refId": "A",
+ "step": 600
+ }
+ ],
+ "title": "Available Replicas",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "avg"
+ },
+ {
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "rgba(245, 54, 54, 0.9)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(50, 172, 45, 0.97)"
+ ],
+ "datasource": "prometheus",
+ "editable": false,
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "id": 3,
+ "links": [],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 3,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "targets": [
+ {
+ "expr": "max(kube_deployment_status_observed_generation{deployment=\"$deployment_name\",namespace=\"$deployment_namespace\"}) without (instance, pod)",
+ "intervalFactor": 2,
+ "refId": "A",
+ "step": 600
+ }
+ ],
+ "title": "Observed Generation",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "avg"
+ },
+ {
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "rgba(245, 54, 54, 0.9)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(50, 172, 45, 0.97)"
+ ],
+ "datasource": "prometheus",
+ "editable": false,
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "id": 2,
+ "links": [],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 3,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "targets": [
+ {
+ "expr": "max(kube_deployment_metadata_generation{deployment=\"$deployment_name\",namespace=\"$deployment_namespace\"}) without (instance, pod)",
+ "intervalFactor": 2,
+ "refId": "A",
+ "step": 600
+ }
+ ],
+ "title": "Metadata Generation",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "avg"
+ }
+ ],
+ "showTitle": false,
+ "title": "Dashboard Row",
+ "titleSize": "h6"
+ },
+ {
+ "collapse": false,
+ "editable": false,
+ "height": "350px",
+ "panels": [
+ {
+ "aliasColors": {},
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "prometheus",
+ "editable": false,
+ "error": false,
+ "fill": 1,
+ "grid": {
+ "threshold1Color": "rgba(216, 200, 27, 0.27)",
+ "threshold2Color": "rgba(234, 112, 112, 0.22)"
+ },
+ "id": 1,
+ "isNew": true,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "hideEmpty": false,
+ "hideZero": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "total": false
+ },
+ "lines": true,
+ "linewidth": 2,
+ "links": [],
+ "nullPointMode": "connected",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [],
+ "spaceLength": 10,
+ "span": 12,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "max(kube_deployment_status_replicas{deployment=\"$deployment_name\",namespace=\"$deployment_namespace\"}) without (instance, pod)",
+ "intervalFactor": 2,
+ "legendFormat": "current replicas",
+ "refId": "A",
+ "step": 30
+ },
+ {
+ "expr": "min(kube_deployment_status_replicas_available{deployment=\"$deployment_name\",namespace=\"$deployment_namespace\"}) without (instance, pod)",
+ "intervalFactor": 2,
+ "legendFormat": "available",
+ "refId": "B",
+ "step": 30
+ },
+ {
+ "expr": "max(kube_deployment_status_replicas_unavailable{deployment=\"$deployment_name\",namespace=\"$deployment_namespace\"}) without (instance, pod)",
+ "intervalFactor": 2,
+ "legendFormat": "unavailable",
+ "refId": "C",
+ "step": 30
+ },
+ {
+ "expr": "min(kube_deployment_status_replicas_updated{deployment=\"$deployment_name\",namespace=\"$deployment_namespace\"}) without (instance, pod)",
+ "intervalFactor": 2,
+ "legendFormat": "updated",
+ "refId": "D",
+ "step": 30
+ },
+ {
+ "expr": "max(kube_deployment_spec_replicas{deployment=\"$deployment_name\",namespace=\"$deployment_namespace\"}) without (instance, pod)",
+ "intervalFactor": 2,
+ "legendFormat": "desired",
+ "refId": "E",
+ "step": 30
+ }
+ ],
+ "title": "Replicas",
+ "tooltip": {
+ "msResolution": true,
+ "shared": true,
+ "sort": 0,
+ "value_type": "cumulative"
+ },
+ "type": "graph",
+ "xaxis": {
+ "mode": "time",
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "format": "none",
+ "label": "",
+ "logBase": 1,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": "",
+ "logBase": 1,
+ "show": false
+ }
+ ]
+ }
+ ],
+ "showTitle": false,
+ "title": "Dashboard Row",
+ "titleSize": "h6"
+ }
+ ],
+ "schemaVersion": 14,
+ "sharedCrosshair": false,
+ "style": "dark",
+ "tags": [],
+ "templating": {
+ "list": [
+ {
+ "allValue": ".*",
+ "current": {},
+ "datasource": "prometheus",
+ "hide": 0,
+ "includeAll": false,
+ "label": "Namespace",
+ "multi": false,
+ "name": "deployment_namespace",
+ "options": [],
+ "query": "label_values(kube_deployment_metadata_generation, namespace)",
+ "refresh": 1,
+ "regex": "",
+ "sort": 0,
+ "tagValuesQuery": null,
+ "tags": [],
+ "tagsQuery": "",
+ "type": "query",
+ "useTags": false
+ },
+ {
+ "allValue": null,
+ "current": {},
+ "datasource": "prometheus",
+ "hide": 0,
+ "includeAll": false,
+ "label": "Deployment",
+ "multi": false,
+ "name": "deployment_name",
+ "options": [],
+ "query": "label_values(kube_deployment_metadata_generation{namespace=\"$deployment_namespace\"}, deployment)",
+ "refresh": 1,
+ "regex": "",
+ "sort": 0,
+ "tagValuesQuery": "",
+ "tags": [],
+ "tagsQuery": "deployment",
+ "type": "query",
+ "useTags": false
+ }
+ ]
+ },
+ "time": {
+ "from": "now-6h",
+ "to": "now"
+ },
+ "timepicker": {
+ "refresh_intervals": [
+ "5s",
+ "10s",
+ "30s",
+ "1m",
+ "5m",
+ "15m",
+ "30m",
+ "1h",
+ "2h",
+ "1d"
+ ],
+ "time_options": [
+ "5m",
+ "15m",
+ "1h",
+ "6h",
+ "12h",
+ "24h",
+ "2d",
+ "7d",
+ "30d"
+ ]
+ },
+ "timezone": "browser",
+ "title": "Deployment",
+ "version": 1
+ }
+ etcd-dashboard.json: |+
+ {
+ "__inputs": [
+ {
+ "name": "prometheus",
+ "label": "prometheus",
+ "description": "",
+ "type": "datasource",
+ "pluginId": "prometheus",
+ "pluginName": "Prometheus"
+ }
+ ],
+ "__requires": [
+ {
+ "type": "grafana",
+ "id": "grafana",
+ "name": "Grafana",
+ "version": "4.5.2"
+ },
+ {
+ "type": "panel",
+ "id": "graph",
+ "name": "Graph",
+ "version": ""
+ },
+ {
+ "type": "datasource",
+ "id": "prometheus",
+ "name": "Prometheus",
+ "version": "1.0.0"
+ },
+ {
+ "type": "panel",
+ "id": "singlestat",
+ "name": "Singlestat",
+ "version": ""
+ }
+ ],
+ "annotations": {
+ "list": []
+ },
+ "description": "etcd sample Grafana dashboard with Prometheus",
+ "editable": false,
+ "gnetId": null,
+ "graphTooltip": 0,
+ "hideControls": false,
+ "id": null,
+ "links": [],
+ "refresh": false,
+ "rows": [
+ {
+ "collapse": false,
+ "height": "250px",
+ "panels": [
+ {
+ "cacheTimeout": null,
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "rgba(245, 54, 54, 0.9)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(50, 172, 45, 0.97)"
+ ],
+ "datasource": "prometheus",
+ "editable": false,
+ "error": false,
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "id": 28,
+ "interval": null,
+ "links": [],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "nullText": null,
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 3,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "tableColumn": "",
+ "targets": [
+ {
+ "expr": "sum(etcd_server_has_leader)",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "",
+ "metric": "etcd_server_has_leader",
+ "refId": "A",
+ "step": 20
+ }
+ ],
+ "thresholds": "",
+ "title": "Up",
+ "type": "singlestat",
+ "valueFontSize": "200%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "avg"
+ },
+ {
+ "aliasColors": {},
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "prometheus",
+ "editable": false,
+ "error": false,
+ "fill": 0,
+ "id": 23,
+ "legend": {
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "show": false,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 2,
+ "links": [],
+ "nullPointMode": "connected",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [],
+ "spaceLength": 10,
+ "span": 5,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "sum(rate(grpc_server_started_total{grpc_type=\"unary\"}[5m]))",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "RPC Rate",
+ "metric": "grpc_server_started_total",
+ "refId": "A",
+ "step": 4
+ },
+ {
+ "expr": "sum(rate(grpc_server_handled_total{grpc_type=\"unary\",grpc_code!=\"OK\"}[5m]))",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "RPC Failed Rate",
+ "metric": "grpc_server_handled_total",
+ "refId": "B",
+ "step": 4
+ }
+ ],
+ "thresholds": [],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "RPC Rate",
+ "tooltip": {
+ "msResolution": false,
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "format": "ops",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ }
+ ]
+ },
+ {
+ "aliasColors": {},
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "prometheus",
+ "editable": false,
+ "error": false,
+ "fill": 0,
+ "id": 41,
+ "legend": {
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "show": false,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 2,
+ "links": [],
+ "nullPointMode": "connected",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [],
+ "spaceLength": 10,
+ "span": 4,
+ "stack": true,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "sum(grpc_server_started_total{grpc_service=\"etcdserverpb.Watch\",grpc_type=\"bidi_stream\"}) - sum(grpc_server_handled_total{grpc_service=\"etcdserverpb.Watch\",grpc_type=\"bidi_stream\"})",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "Watch Streams",
+ "metric": "grpc_server_handled_total",
+ "refId": "A",
+ "step": 4
+ },
+ {
+ "expr": "sum(grpc_server_started_total{grpc_service=\"etcdserverpb.Lease\",grpc_type=\"bidi_stream\"}) - sum(grpc_server_handled_total{grpc_service=\"etcdserverpb.Lease\",grpc_type=\"bidi_stream\"})",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "Lease Streams",
+ "metric": "grpc_server_handled_total",
+ "refId": "B",
+ "step": 4
+ }
+ ],
+ "thresholds": [],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Active Streams",
+ "tooltip": {
+ "msResolution": false,
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "format": "short",
+ "label": "",
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ }
+ ]
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": false,
+ "title": "Row",
+ "titleSize": "h6"
+ },
+ {
+ "collapse": false,
+ "height": "250px",
+ "panels": [
+ {
+ "aliasColors": {},
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "prometheus",
+ "decimals": null,
+ "editable": false,
+ "error": false,
+ "fill": 0,
+ "grid": {},
+ "id": 1,
+ "legend": {
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "show": false,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 2,
+ "links": [],
+ "nullPointMode": "connected",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [],
+ "spaceLength": 10,
+ "span": 4,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "etcd_debugging_mvcc_db_total_size_in_bytes",
+ "format": "time_series",
+ "hide": false,
+ "interval": "",
+ "intervalFactor": 2,
+ "legendFormat": "{{instance}} DB Size",
+ "metric": "",
+ "refId": "A",
+ "step": 4
+ }
+ ],
+ "thresholds": [],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "DB Size",
+ "tooltip": {
+ "msResolution": false,
+ "shared": true,
+ "sort": 0,
+ "value_type": "cumulative"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "format": "bytes",
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "short",
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": false
+ }
+ ]
+ },
+ {
+ "aliasColors": {},
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "prometheus",
+ "editable": false,
+ "error": false,
+ "fill": 0,
+ "grid": {},
+ "id": 3,
+ "legend": {
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "show": false,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 2,
+ "links": [],
+ "nullPointMode": "connected",
+ "percentage": false,
+ "pointradius": 1,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [],
+ "spaceLength": 10,
+ "span": 4,
+ "stack": false,
+ "steppedLine": true,
+ "targets": [
+ {
+ "expr": "histogram_quantile(0.99, sum(rate(etcd_disk_wal_fsync_duration_seconds_bucket[5m])) by (instance, le))",
+ "format": "time_series",
+ "hide": false,
+ "intervalFactor": 2,
+ "legendFormat": "{{instance}} WAL fsync",
+ "metric": "etcd_disk_wal_fsync_duration_seconds_bucket",
+ "refId": "A",
+ "step": 4
+ },
+ {
+ "expr": "histogram_quantile(0.99, sum(rate(etcd_disk_backend_commit_duration_seconds_bucket[5m])) by (instance, le))",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "{{instance}} DB fsync",
+ "metric": "etcd_disk_backend_commit_duration_seconds_bucket",
+ "refId": "B",
+ "step": 4
+ }
+ ],
+ "thresholds": [],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Disk Sync Duration",
+ "tooltip": {
+ "msResolution": false,
+ "shared": true,
+ "sort": 0,
+ "value_type": "cumulative"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "format": "s",
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "short",
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": false
+ }
+ ]
+ },
+ {
+ "aliasColors": {},
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "prometheus",
+ "editable": false,
+ "error": false,
+ "fill": 0,
+ "id": 29,
+ "legend": {
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "show": false,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 2,
+ "links": [],
+ "nullPointMode": "connected",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [],
+ "spaceLength": 10,
+ "span": 4,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "process_resident_memory_bytes",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "{{instance}} Resident Memory",
+ "metric": "process_resident_memory_bytes",
+ "refId": "A",
+ "step": 4
+ }
+ ],
+ "thresholds": [],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Memory",
+ "tooltip": {
+ "msResolution": false,
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "format": "bytes",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ }
+ ]
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": false,
+ "title": "New row",
+ "titleSize": "h6"
+ },
+ {
+ "collapse": false,
+ "height": "250px",
+ "panels": [
+ {
+ "aliasColors": {},
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "prometheus",
+ "editable": false,
+ "error": false,
+ "fill": 5,
+ "id": 22,
+ "legend": {
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "show": false,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 2,
+ "links": [],
+ "nullPointMode": "connected",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [],
+ "spaceLength": 10,
+ "span": 3,
+ "stack": true,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "rate(etcd_network_client_grpc_received_bytes_total[5m])",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "{{instance}} Client Traffic In",
+ "metric": "etcd_network_client_grpc_received_bytes_total",
+ "refId": "A",
+ "step": 4
+ }
+ ],
+ "thresholds": [],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Client Traffic In",
+ "tooltip": {
+ "msResolution": false,
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "format": "Bps",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ }
+ ]
+ },
+ {
+ "aliasColors": {},
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "prometheus",
+ "editable": false,
+ "error": false,
+ "fill": 5,
+ "id": 21,
+ "legend": {
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "show": false,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 2,
+ "links": [],
+ "nullPointMode": "connected",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [],
+ "spaceLength": 10,
+ "span": 3,
+ "stack": true,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "rate(etcd_network_client_grpc_sent_bytes_total[5m])",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "{{instance}} Client Traffic Out",
+ "metric": "etcd_network_client_grpc_sent_bytes_total",
+ "refId": "A",
+ "step": 4
+ }
+ ],
+ "thresholds": [],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Client Traffic Out",
+ "tooltip": {
+ "msResolution": false,
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "format": "Bps",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ }
+ ]
+ },
+ {
+ "aliasColors": {},
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "prometheus",
+ "editable": false,
+ "error": false,
+ "fill": 0,
+ "id": 20,
+ "legend": {
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "show": false,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 2,
+ "links": [],
+ "nullPointMode": "connected",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [],
+ "spaceLength": 10,
+ "span": 3,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "sum(rate(etcd_network_peer_received_bytes_total[5m])) by (instance)",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "{{instance}} Peer Traffic In",
+ "metric": "etcd_network_peer_received_bytes_total",
+ "refId": "A",
+ "step": 4
+ }
+ ],
+ "thresholds": [],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Peer Traffic In",
+ "tooltip": {
+ "msResolution": false,
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "format": "Bps",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ }
+ ]
+ },
+ {
+ "aliasColors": {},
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "prometheus",
+ "decimals": null,
+ "editable": false,
+ "error": false,
+ "fill": 0,
+ "grid": {},
+ "id": 16,
+ "legend": {
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "show": false,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 2,
+ "links": [],
+ "nullPointMode": "connected",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [],
+ "spaceLength": 10,
+ "span": 3,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "sum(rate(etcd_network_peer_sent_bytes_total[5m])) by (instance)",
+ "format": "time_series",
+ "hide": false,
+ "interval": "",
+ "intervalFactor": 2,
+ "legendFormat": "{{instance}} Peer Traffic Out",
+ "metric": "etcd_network_peer_sent_bytes_total",
+ "refId": "A",
+ "step": 4
+ }
+ ],
+ "thresholds": [],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Peer Traffic Out",
+ "tooltip": {
+ "msResolution": false,
+ "shared": true,
+ "sort": 0,
+ "value_type": "cumulative"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "format": "Bps",
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "short",
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ }
+ ]
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": false,
+ "title": "New row",
+ "titleSize": "h6"
+ },
+ {
+ "collapse": false,
+ "height": "250px",
+ "panels": [
+ {
+ "aliasColors": {},
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "prometheus",
+ "editable": false,
+ "error": false,
+ "fill": 0,
+ "id": 40,
+ "legend": {
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "show": false,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 2,
+ "links": [],
+ "nullPointMode": "connected",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [],
+ "spaceLength": 10,
+ "span": 6,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "sum(rate(etcd_server_proposals_failed_total[5m]))",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "Proposal Failure Rate",
+ "metric": "etcd_server_proposals_failed_total",
+ "refId": "A",
+ "step": 2
+ },
+ {
+ "expr": "sum(etcd_server_proposals_pending)",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "Proposal Pending Total",
+ "metric": "etcd_server_proposals_pending",
+ "refId": "B",
+ "step": 2
+ },
+ {
+ "expr": "sum(rate(etcd_server_proposals_committed_total[5m]))",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "Proposal Commit Rate",
+ "metric": "etcd_server_proposals_committed_total",
+ "refId": "C",
+ "step": 2
+ },
+ {
+ "expr": "sum(rate(etcd_server_proposals_applied_total[5m]))",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "Proposal Apply Rate",
+ "refId": "D",
+ "step": 2
+ }
+ ],
+ "thresholds": [],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Raft Proposals",
+ "tooltip": {
+ "msResolution": false,
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "format": "short",
+ "label": "",
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ }
+ ]
+ },
+ {
+ "aliasColors": {},
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "prometheus",
+ "decimals": 0,
+ "editable": false,
+ "error": false,
+ "fill": 0,
+ "id": 19,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": false,
+ "total": false,
+ "values": false
+ },
+ "lines": true,
+ "linewidth": 2,
+ "links": [],
+ "nullPointMode": "connected",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [],
+ "spaceLength": 10,
+ "span": 6,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "changes(etcd_server_leader_changes_seen_total[1d])",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "{{instance}} Total Leader Elections Per Day",
+ "metric": "etcd_server_leader_changes_seen_total",
+ "refId": "A",
+ "step": 2
+ }
+ ],
+ "thresholds": [],
+ "timeFrom": null,
+ "timeShift": null,
+ "title": "Total Leader Elections Per Day",
+ "tooltip": {
+ "msResolution": false,
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "buckets": null,
+ "mode": "time",
+ "name": null,
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": null,
+ "logBase": 1,
+ "max": null,
+ "min": null,
+ "show": true
+ }
+ ]
+ }
+ ],
+ "repeat": null,
+ "repeatIteration": null,
+ "repeatRowId": null,
+ "showTitle": false,
+ "title": "New row",
+ "titleSize": "h6"
+ }
+ ],
+ "schemaVersion": 14,
+ "style": "dark",
+ "tags": [],
+ "templating": {
+ "list": []
+ },
+ "time": {
+ "from": "now-15m",
+ "to": "now"
+ },
+ "timepicker": {
+ "now": true,
+ "refresh_intervals": [
+ "5s",
+ "10s",
+ "30s",
+ "1m",
+ "5m",
+ "15m",
+ "30m",
+ "1h",
+ "2h",
+ "1d"
+ ],
+ "time_options": [
+ "5m",
+ "15m",
+ "1h",
+ "6h",
+ "12h",
+ "24h",
+ "2d",
+ "7d",
+ "30d"
+ ]
+ },
+ "timezone": "browser",
+ "title": "etcd",
+ "version": 4
+ }
+ kubernetes-capacity-planning-dashboard.json: |+
+ {
+ "__inputs": [
+ {
+ "description": "",
+ "label": "prometheus",
+ "name": "prometheus",
+ "pluginId": "prometheus",
+ "pluginName": "Prometheus",
+ "type": "datasource"
+ }
+ ],
+ "annotations": {
+ "list": []
+ },
+ "editable": false,
+ "gnetId": 22,
+ "graphTooltip": 0,
+ "hideControls": false,
+ "links": [],
+ "refresh": false,
+ "rows": [
+ {
+ "collapse": false,
+ "editable": false,
+ "height": "250px",
+ "panels": [
+ {
+ "aliasColors": {},
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "prometheus",
+ "editable": false,
+ "error": false,
+ "fill": 1,
+ "grid": {
+ "threshold1Color": "rgba(216, 200, 27, 0.27)",
+ "threshold2Color": "rgba(234, 112, 112, 0.22)"
+ },
+ "id": 3,
+ "isNew": false,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "hideEmpty": false,
+ "hideZero": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "total": false
+ },
+ "lines": true,
+ "linewidth": 2,
+ "links": [],
+ "nullPointMode": "connected",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [],
+ "spaceLength": 10,
+ "span": 6,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "sum(rate(node_cpu{mode=\"idle\"}[2m])) * 100",
+ "hide": false,
+ "intervalFactor": 10,
+ "legendFormat": "",
+ "refId": "A",
+ "step": 50
+ }
+ ],
+ "title": "Idle CPU",
+ "tooltip": {
+ "msResolution": false,
+ "shared": true,
+ "sort": 0,
+ "value_type": "cumulative"
+ },
+ "type": "graph",
+ "xaxis": {
+ "mode": "time",
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "format": "percent",
+ "label": "cpu usage",
+ "logBase": 1,
+ "min": 0,
+ "show": true
+ },
+ {
+ "format": "short",
+ "logBase": 1,
+ "show": true
+ }
+ ]
+ },
+ {
+ "aliasColors": {},
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "prometheus",
+ "editable": false,
+ "error": false,
+ "fill": 1,
+ "grid": {
+ "threshold1Color": "rgba(216, 200, 27, 0.27)",
+ "threshold2Color": "rgba(234, 112, 112, 0.22)"
+ },
+ "id": 9,
+ "isNew": false,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "hideEmpty": false,
+ "hideZero": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "total": false
+ },
+ "lines": true,
+ "linewidth": 2,
+ "links": [],
+ "nullPointMode": "connected",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [],
+ "spaceLength": 10,
+ "span": 6,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "sum(node_load1)",
+ "intervalFactor": 4,
+ "legendFormat": "load 1m",
+ "refId": "A",
+ "step": 20,
+ "target": ""
+ },
+ {
+ "expr": "sum(node_load5)",
+ "intervalFactor": 4,
+ "legendFormat": "load 5m",
+ "refId": "B",
+ "step": 20,
+ "target": ""
+ },
+ {
+ "expr": "sum(node_load15)",
+ "intervalFactor": 4,
+ "legendFormat": "load 15m",
+ "refId": "C",
+ "step": 20,
+ "target": ""
+ }
+ ],
+ "title": "System Load",
+ "tooltip": {
+ "msResolution": false,
+ "shared": true,
+ "sort": 0,
+ "value_type": "cumulative"
+ },
+ "type": "graph",
+ "xaxis": {
+ "mode": "time",
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "format": "percentunit",
+ "logBase": 1,
+ "show": true
+ },
+ {
+ "format": "short",
+ "logBase": 1,
+ "show": true
+ }
+ ]
+ }
+ ],
+ "showTitle": false,
+ "title": "New Row",
+ "titleSize": "h6"
+ },
+ {
+ "collapse": false,
+ "editable": false,
+ "height": "250px",
+ "panels": [
+ {
+ "aliasColors": {},
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "prometheus",
+ "editable": false,
+ "error": false,
+ "fill": 1,
+ "grid": {
+ "threshold1Color": "rgba(216, 200, 27, 0.27)",
+ "threshold2Color": "rgba(234, 112, 112, 0.22)"
+ },
+ "id": 4,
+ "isNew": false,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "hideEmpty": false,
+ "hideZero": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "total": false
+ },
+ "lines": true,
+ "linewidth": 2,
+ "links": [],
+ "nullPointMode": "connected",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [
+ {
+ "alias": "node_memory_SwapFree{instance=\"172.17.0.1:9100\",job=\"prometheus\"}",
+ "yaxis": 2
+ }
+ ],
+ "spaceLength": 10,
+ "span": 9,
+ "stack": true,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "sum(node_memory_MemTotal) - sum(node_memory_MemFree) - sum(node_memory_Buffers) - sum(node_memory_Cached)",
+ "intervalFactor": 2,
+ "legendFormat": "memory usage",
+ "metric": "memo",
+ "refId": "A",
+ "step": 10,
+ "target": ""
+ },
+ {
+ "expr": "sum(node_memory_Buffers)",
+ "interval": "",
+ "intervalFactor": 2,
+ "legendFormat": "memory buffers",
+ "metric": "memo",
+ "refId": "B",
+ "step": 10,
+ "target": ""
+ },
+ {
+ "expr": "sum(node_memory_Cached)",
+ "interval": "",
+ "intervalFactor": 2,
+ "legendFormat": "memory cached",
+ "metric": "memo",
+ "refId": "C",
+ "step": 10,
+ "target": ""
+ },
+ {
+ "expr": "sum(node_memory_MemFree)",
+ "interval": "",
+ "intervalFactor": 2,
+ "legendFormat": "memory free",
+ "metric": "memo",
+ "refId": "D",
+ "step": 10,
+ "target": ""
+ }
+ ],
+ "title": "Memory Usage",
+ "tooltip": {
+ "msResolution": false,
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "mode": "time",
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "format": "bytes",
+ "logBase": 1,
+ "min": "0",
+ "show": true
+ },
+ {
+ "format": "short",
+ "logBase": 1,
+ "show": true
+ }
+ ]
+ },
+ {
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "rgba(50, 172, 45, 0.97)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(245, 54, 54, 0.9)"
+ ],
+ "datasource": "prometheus",
+ "editable": false,
+ "format": "percent",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": true,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "hideTimeOverride": false,
+ "id": 5,
+ "links": [],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 3,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "targets": [
+ {
+ "expr": "((sum(node_memory_MemTotal) - sum(node_memory_MemFree) - sum(node_memory_Buffers) - sum(node_memory_Cached)) / sum(node_memory_MemTotal)) * 100",
+ "intervalFactor": 2,
+ "metric": "",
+ "refId": "A",
+ "step": 60,
+ "target": ""
+ }
+ ],
+ "thresholds": "80, 90",
+ "title": "Memory Usage",
+ "transparent": false,
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "avg"
+ }
+ ],
+ "showTitle": false,
+ "title": "New Row",
+ "titleSize": "h6"
+ },
+ {
+ "collapse": false,
+ "editable": false,
+ "height": "246px",
+ "panels": [
+ {
+ "aliasColors": {},
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "prometheus",
+ "editable": false,
+ "error": false,
+ "fill": 1,
+ "grid": {
+ "threshold1Color": "rgba(216, 200, 27, 0.27)",
+ "threshold2Color": "rgba(234, 112, 112, 0.22)"
+ },
+ "id": 6,
+ "isNew": false,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "hideEmpty": false,
+ "hideZero": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "total": false
+ },
+ "lines": true,
+ "linewidth": 2,
+ "links": [],
+ "nullPointMode": "connected",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [
+ {
+ "alias": "read",
+ "yaxis": 1
+ },
+ {
+ "alias": "{instance=\"172.17.0.1:9100\"}",
+ "yaxis": 2
+ },
+ {
+ "alias": "io time",
+ "yaxis": 2
+ }
+ ],
+ "spaceLength": 10,
+ "span": 9,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "sum(rate(node_disk_bytes_read[5m]))",
+ "hide": false,
+ "intervalFactor": 4,
+ "legendFormat": "read",
+ "refId": "A",
+ "step": 20,
+ "target": ""
+ },
+ {
+ "expr": "sum(rate(node_disk_bytes_written[5m]))",
+ "intervalFactor": 4,
+ "legendFormat": "written",
+ "refId": "B",
+ "step": 20
+ },
+ {
+ "expr": "sum(rate(node_disk_io_time_ms[5m]))",
+ "intervalFactor": 4,
+ "legendFormat": "io time",
+ "refId": "C",
+ "step": 20
+ }
+ ],
+ "title": "Disk I/O",
+ "tooltip": {
+ "msResolution": false,
+ "shared": true,
+ "sort": 0,
+ "value_type": "cumulative"
+ },
+ "type": "graph",
+ "xaxis": {
+ "mode": "time",
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "format": "bytes",
+ "logBase": 1,
+ "show": true
+ },
+ {
+ "format": "ms",
+ "logBase": 1,
+ "show": true
+ }
+ ]
+ },
+ {
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "rgba(50, 172, 45, 0.97)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(245, 54, 54, 0.9)"
+ ],
+ "datasource": "prometheus",
+ "editable": false,
+ "format": "percentunit",
+ "gauge": {
+ "maxValue": 1,
+ "minValue": 0,
+ "show": true,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "hideTimeOverride": false,
+ "id": 12,
+ "links": [],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 3,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "targets": [
+ {
+ "expr": "(sum(node_filesystem_size{device!=\"rootfs\"}) - sum(node_filesystem_free{device!=\"rootfs\"})) / sum(node_filesystem_size{device!=\"rootfs\"})",
+ "intervalFactor": 2,
+ "refId": "A",
+ "step": 60,
+ "target": ""
+ }
+ ],
+ "thresholds": "0.75, 0.9",
+ "title": "Disk Space Usage",
+ "transparent": false,
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ }
+ ],
+ "showTitle": false,
+ "title": "New Row",
+ "titleSize": "h6"
+ },
+ {
+ "collapse": false,
+ "editable": false,
+ "height": "250px",
+ "panels": [
+ {
+ "aliasColors": {},
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "prometheus",
+ "editable": false,
+ "error": false,
+ "fill": 1,
+ "grid": {
+ "threshold1Color": "rgba(216, 200, 27, 0.27)",
+ "threshold2Color": "rgba(234, 112, 112, 0.22)"
+ },
+ "id": 8,
+ "isNew": false,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "hideEmpty": false,
+ "hideZero": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "total": false
+ },
+ "lines": true,
+ "linewidth": 2,
+ "links": [],
+ "nullPointMode": "connected",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [
+ {
+ "alias": "transmitted",
+ "yaxis": 2
+ }
+ ],
+ "spaceLength": 10,
+ "span": 6,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "sum(rate(node_network_receive_bytes{device!~\"lo\"}[5m]))",
+ "hide": false,
+ "intervalFactor": 2,
+ "legendFormat": "",
+ "refId": "A",
+ "step": 10,
+ "target": ""
+ }
+ ],
+ "title": "Network Received",
+ "tooltip": {
+ "msResolution": false,
+ "shared": true,
+ "sort": 0,
+ "value_type": "cumulative"
+ },
+ "type": "graph",
+ "xaxis": {
+ "mode": "time",
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "format": "bytes",
+ "logBase": 1,
+ "show": true
+ },
+ {
+ "format": "bytes",
+ "logBase": 1,
+ "show": true
+ }
+ ]
+ },
+ {
+ "aliasColors": {},
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "prometheus",
+ "editable": false,
+ "error": false,
+ "fill": 1,
+ "grid": {
+ "threshold1Color": "rgba(216, 200, 27, 0.27)",
+ "threshold2Color": "rgba(234, 112, 112, 0.22)"
+ },
+ "id": 10,
+ "isNew": false,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "hideEmpty": false,
+ "hideZero": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "total": false
+ },
+ "lines": true,
+ "linewidth": 2,
+ "links": [],
+ "nullPointMode": "connected",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [
+ {
+ "alias": "transmitted",
+ "yaxis": 2
+ }
+ ],
+ "spaceLength": 10,
+ "span": 6,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "sum(rate(node_network_transmit_bytes{device!~\"lo\"}[5m]))",
+ "hide": false,
+ "intervalFactor": 2,
+ "legendFormat": "",
+ "refId": "B",
+ "step": 10,
+ "target": ""
+ }
+ ],
+ "title": "Network Transmitted",
+ "tooltip": {
+ "msResolution": false,
+ "shared": true,
+ "sort": 0,
+ "value_type": "cumulative"
+ },
+ "type": "graph",
+ "xaxis": {
+ "mode": "time",
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "format": "bytes",
+ "logBase": 1,
+ "show": true
+ },
+ {
+ "format": "bytes",
+ "logBase": 1,
+ "show": true
+ }
+ ]
+ }
+ ],
+ "showTitle": false,
+ "title": "New Row",
+ "titleSize": "h6"
+ },
+ {
+ "collapse": false,
+ "editable": false,
+ "height": "276px",
+ "panels": [
+ {
+ "aliasColors": {},
+ "bars": false,
+ "dashes": false,
+ "datasource": "prometheus",
+ "editable": false,
+ "error": false,
+ "fill": 1,
+ "grid": {
+ "threshold1Color": "rgba(216, 200, 27, 0.27)",
+ "threshold2Color": "rgba(234, 112, 112, 0.22)"
+ },
+ "id": 11,
+ "isNew": true,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "hideEmpty": false,
+ "hideZero": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "total": false
+ },
+ "lines": true,
+ "linewidth": 2,
+ "links": [],
+ "nullPointMode": "connected",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [],
+ "spaceLength": 11,
+ "span": 9,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "sum(kube_pod_info)",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "Current number of Pods",
+ "refId": "A",
+ "step": 10
+ },
+ {
+ "expr": "sum(kube_node_status_capacity_pods)",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "Maximum capacity of pods",
+ "refId": "B",
+ "step": 10
+ }
+ ],
+ "title": "Cluster Pod Utilization",
+ "tooltip": {
+ "msResolution": false,
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "mode": "time",
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "format": "short",
+ "logBase": 1,
+ "show": true
+ },
+ {
+ "format": "short",
+ "logBase": 1,
+ "show": true
+ }
+ ]
+ },
+ {
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "rgba(50, 172, 45, 0.97)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(245, 54, 54, 0.9)"
+ ],
+ "datasource": "prometheus",
+ "editable": false,
+ "format": "percent",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": true,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "hideTimeOverride": false,
+ "id": 7,
+ "links": [],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 3,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "targets": [
+ {
+ "expr": "100 - (sum(kube_node_status_capacity_pods) - sum(kube_pod_info)) / sum(kube_node_status_capacity_pods) * 100",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "",
+ "refId": "A",
+ "step": 60,
+ "target": ""
+ }
+ ],
+ "thresholds": "80, 90",
+ "title": "Pod Utilization",
+ "transparent": false,
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ }
+ ],
+ "showTitle": false,
+ "title": "New Row",
+ "titleSize": "h6"
+ }
+ ],
+ "schemaVersion": 14,
+ "sharedCrosshair": false,
+ "style": "dark",
+ "tags": [],
+ "templating": {
+ "list": []
+ },
+ "time": {
+ "from": "now-1h",
+ "to": "now"
+ },
+ "timepicker": {
+ "refresh_intervals": [
+ "5s",
+ "10s",
+ "30s",
+ "1m",
+ "5m",
+ "15m",
+ "30m",
+ "1h",
+ "2h",
+ "1d"
+ ],
+ "time_options": [
+ "5m",
+ "15m",
+ "1h",
+ "6h",
+ "12h",
+ "24h",
+ "2d",
+ "7d",
+ "30d"
+ ]
+ },
+ "timezone": "browser",
+ "title": "Kubernetes Capacity Planning",
+ "version": 4
+ }
+ kubernetes-cluster-health-dashboard.json: |+
+ {
+ "__inputs": [
+ {
+ "description": "",
+ "label": "prometheus",
+ "name": "prometheus",
+ "pluginId": "prometheus",
+ "pluginName": "Prometheus",
+ "type": "datasource"
+ }
+ ],
+ "annotations": {
+ "list": []
+ },
+ "editable": false,
+ "graphTooltip": 0,
+ "hideControls": false,
+ "links": [],
+ "refresh": "10s",
+ "rows": [
+ {
+ "collapse": false,
+ "editable": false,
+ "height": "254px",
+ "panels": [
+ {
+ "colorBackground": false,
+ "colorValue": true,
+ "colors": [
+ "rgba(50, 172, 45, 0.97)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(245, 54, 54, 0.9)"
+ ],
+ "datasource": "prometheus",
+ "editable": false,
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "hideTimeOverride": false,
+ "id": 1,
+ "links": [],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 3,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "targets": [
+ {
+ "expr": "sum(up{job=~\"apiserver|kube-scheduler|kube-controller-manager\"} == 0)",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "",
+ "refId": "A",
+ "step": 600
+ }
+ ],
+ "thresholds": "1, 3",
+ "title": "Control Plane Components Down",
+ "transparent": false,
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "Everything UP and healthy",
+ "value": "null"
+ },
+ {
+ "op": "=",
+ "text": "",
+ "value": ""
+ }
+ ],
+ "valueName": "avg"
+ },
+ {
+ "colorBackground": false,
+ "colorValue": true,
+ "colors": [
+ "rgba(50, 172, 45, 0.97)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(245, 54, 54, 0.9)"
+ ],
+ "datasource": "prometheus",
+ "editable": false,
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "hideTimeOverride": false,
+ "id": 2,
+ "links": [],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 3,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "targets": [
+ {
+ "expr": "sum(ALERTS{alertstate=\"firing\",alertname!=\"DeadMansSwitch\"})",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "",
+ "refId": "A",
+ "step": 600
+ }
+ ],
+ "thresholds": "1, 3",
+ "title": "Alerts Firing",
+ "transparent": false,
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "0",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ },
+ {
+ "colorBackground": false,
+ "colorValue": true,
+ "colors": [
+ "rgba(50, 172, 45, 0.97)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(245, 54, 54, 0.9)"
+ ],
+ "datasource": "prometheus",
+ "editable": false,
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "hideTimeOverride": false,
+ "id": 3,
+ "links": [],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 3,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "targets": [
+ {
+ "expr": "sum(ALERTS{alertstate=\"pending\",alertname!=\"DeadMansSwitch\"})",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "",
+ "refId": "A",
+ "step": 600
+ }
+ ],
+ "thresholds": "3, 5",
+ "title": "Alerts Pending",
+ "transparent": false,
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "0",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ },
+ {
+ "colorBackground": false,
+ "colorValue": true,
+ "colors": [
+ "rgba(50, 172, 45, 0.97)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(245, 54, 54, 0.9)"
+ ],
+ "datasource": "prometheus",
+ "editable": false,
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "hideTimeOverride": false,
+ "id": 4,
+ "links": [],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 3,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "targets": [
+ {
+ "expr": "count(increase(kube_pod_container_status_restarts[1h]) > 5)",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "",
+ "refId": "A",
+ "step": 600
+ }
+ ],
+ "thresholds": "1, 3",
+ "title": "Crashlooping Pods",
+ "transparent": false,
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "0",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ }
+ ],
+ "showTitle": false,
+ "title": "Row",
+ "titleSize": "h6"
+ },
+ {
+ "collapse": false,
+ "editable": false,
+ "height": "250px",
+ "panels": [
+ {
+ "colorBackground": false,
+ "colorValue": true,
+ "colors": [
+ "rgba(50, 172, 45, 0.97)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(245, 54, 54, 0.9)"
+ ],
+ "datasource": "prometheus",
+ "editable": false,
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "hideTimeOverride": false,
+ "id": 5,
+ "links": [],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 3,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "targets": [
+ {
+ "expr": "sum(kube_node_status_condition{condition=\"Ready\",status!=\"true\"})",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "",
+ "refId": "A",
+ "step": 600
+ }
+ ],
+ "thresholds": "1, 3",
+ "title": "Node Not Ready",
+ "transparent": false,
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ },
+ {
+ "colorBackground": false,
+ "colorValue": true,
+ "colors": [
+ "rgba(50, 172, 45, 0.97)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(245, 54, 54, 0.9)"
+ ],
+ "datasource": "prometheus",
+ "editable": false,
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "hideTimeOverride": false,
+ "id": 6,
+ "links": [],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 3,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "targets": [
+ {
+ "expr": "sum(kube_node_status_condition{condition=\"DiskPressure\",status=\"true\"})",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "",
+ "refId": "A",
+ "step": 600
+ }
+ ],
+ "thresholds": "1, 3",
+ "title": "Node Disk Pressure",
+ "transparent": false,
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ },
+ {
+ "colorBackground": false,
+ "colorValue": true,
+ "colors": [
+ "rgba(50, 172, 45, 0.97)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(245, 54, 54, 0.9)"
+ ],
+ "datasource": "prometheus",
+ "editable": false,
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "hideTimeOverride": false,
+ "id": 7,
+ "links": [],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 3,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "targets": [
+ {
+ "expr": "sum(kube_node_status_condition{condition=\"MemoryPressure\",status=\"true\"})",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "",
+ "refId": "A",
+ "step": 600
+ }
+ ],
+ "thresholds": "1, 3",
+ "title": "Node Memory Pressure",
+ "transparent": false,
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ },
+ {
+ "colorBackground": false,
+ "colorValue": true,
+ "colors": [
+ "rgba(50, 172, 45, 0.97)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(245, 54, 54, 0.9)"
+ ],
+ "datasource": "prometheus",
+ "editable": false,
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "hideTimeOverride": false,
+ "id": 8,
+ "links": [],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 3,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "targets": [
+ {
+ "expr": "sum(kube_node_spec_unschedulable)",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "",
+ "refId": "A",
+ "step": 600
+ }
+ ],
+ "thresholds": "1, 3",
+ "title": "Nodes Unschedulable",
+ "transparent": false,
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ }
+ ],
+ "showTitle": false,
+ "title": "Row",
+ "titleSize": "h6"
+ }
+ ],
+ "schemaVersion": 14,
+ "sharedCrosshair": false,
+ "style": "dark",
+ "tags": [],
+ "templating": {
+ "list": []
+ },
+ "time": {
+ "from": "now-6h",
+ "to": "now"
+ },
+ "timepicker": {
+ "refresh_intervals": [
+ "5s",
+ "10s",
+ "30s",
+ "1m",
+ "5m",
+ "15m",
+ "30m",
+ "1h",
+ "2h",
+ "1d"
+ ],
+ "time_options": [
+ "5m",
+ "15m",
+ "1h",
+ "6h",
+ "12h",
+ "24h",
+ "2d",
+ "7d",
+ "30d"
+ ]
+ },
+ "timezone": "browser",
+ "title": "Kubernetes Cluster Health",
+ "version": 9
+ }
+ kubernetes-cluster-status-dashboard.json: |+
+ {
+ "__inputs": [
+ {
+ "description": "",
+ "label": "prometheus",
+ "name": "prometheus",
+ "pluginId": "prometheus",
+ "pluginName": "Prometheus",
+ "type": "datasource"
+ }
+ ],
+ "annotations": {
+ "list": []
+ },
+ "editable": false,
+ "graphTooltip": 0,
+ "hideControls": false,
+ "links": [],
+ "rows": [
+ {
+ "collapse": false,
+ "editable": false,
+ "height": "129px",
+ "panels": [
+ {
+ "colorBackground": false,
+ "colorValue": true,
+ "colors": [
+ "rgba(50, 172, 45, 0.97)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(245, 54, 54, 0.9)"
+ ],
+ "datasource": "prometheus",
+ "editable": false,
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "id": 5,
+ "links": [],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 6,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "targets": [
+ {
+ "expr": "sum(up{job=~\"apiserver|kube-scheduler|kube-controller-manager\"} == 0)",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "refId": "A",
+ "step": 600
+ }
+ ],
+ "thresholds": "1, 3",
+ "title": "Control Plane UP",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "UP",
+ "value": "null"
+ }
+ ],
+ "valueName": "total"
+ },
+ {
+ "colorBackground": false,
+ "colorValue": true,
+ "colors": [
+ "rgba(50, 172, 45, 0.97)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(245, 54, 54, 0.9)"
+ ],
+ "datasource": "prometheus",
+ "editable": false,
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "id": 6,
+ "links": [],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 6,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "targets": [
+ {
+ "expr": "sum(ALERTS{alertstate=\"firing\",alertname!=\"DeadMansSwitch\"})",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "refId": "A",
+ "step": 600
+ }
+ ],
+ "thresholds": "3, 5",
+ "title": "Alerts Firing",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "0",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ }
+ ],
+ "showTitle": true,
+ "title": "Cluster Health",
+ "titleSize": "h6"
+ },
+ {
+ "collapse": false,
+ "editable": false,
+ "height": "168px",
+ "panels": [
+ {
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "rgba(245, 54, 54, 0.9)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(50, 172, 45, 0.97)"
+ ],
+ "datasource": "prometheus",
+ "editable": false,
+ "format": "percent",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": true,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "id": 1,
+ "links": [],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 3,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "targets": [
+ {
+ "expr": "(sum(up{job=\"apiserver\"} == 1) / count(up{job=\"apiserver\"})) * 100",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "refId": "A",
+ "step": 600
+ }
+ ],
+ "thresholds": "50, 80",
+ "title": "API Servers UP",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ },
+ {
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "rgba(245, 54, 54, 0.9)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(50, 172, 45, 0.97)"
+ ],
+ "datasource": "prometheus",
+ "editable": false,
+ "format": "percent",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": true,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "id": 2,
+ "links": [],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 3,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "targets": [
+ {
+ "expr": "(sum(up{job=\"kube-controller-manager\"} == 1) / count(up{job=\"kube-controller-manager\"})) * 100",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "refId": "A",
+ "step": 600
+ }
+ ],
+ "thresholds": "50, 80",
+ "title": "Controller Managers UP",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ },
+ {
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "rgba(245, 54, 54, 0.9)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(50, 172, 45, 0.97)"
+ ],
+ "datasource": "prometheus",
+ "editable": false,
+ "format": "percent",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": true,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "id": 3,
+ "links": [],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 3,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "targets": [
+ {
+ "expr": "(sum(up{job=\"kube-scheduler\"} == 1) / count(up{job=\"kube-scheduler\"})) * 100",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "refId": "A",
+ "step": 600
+ }
+ ],
+ "thresholds": "50, 80",
+ "title": "Schedulers UP",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ },
+ {
+ "colorBackground": false,
+ "colorValue": true,
+ "colors": [
+ "rgba(50, 172, 45, 0.97)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(245, 54, 54, 0.9)"
+ ],
+ "datasource": "prometheus",
+ "editable": false,
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "id": 4,
+ "links": [],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 3,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "targets": [
+ {
+ "expr": "count(increase(kube_pod_container_status_restarts{namespace=~\"kube-system|tectonic-system\"}[1h]) > 5)",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "refId": "A",
+ "step": 600
+ }
+ ],
+ "thresholds": "1, 3",
+ "title": "Crashlooping Control Plane Pods",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "0",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ }
+ ],
+ "showTitle": true,
+ "title": "Control Plane Status",
+ "titleSize": "h6"
+ },
+ {
+ "collapse": false,
+ "editable": false,
+ "height": "158px",
+ "panels": [
+ {
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "rgba(50, 172, 45, 0.97)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(245, 54, 54, 0.9)"
+ ],
+ "datasource": "prometheus",
+ "editable": false,
+ "format": "percent",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": true,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "id": 8,
+ "links": [],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 3,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "targets": [
+ {
+ "expr": "sum(100 - (avg by (instance) (rate(node_cpu{job=\"node-exporter\",mode=\"idle\"}[5m])) * 100)) / count(node_cpu{job=\"node-exporter\",mode=\"idle\"})",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "refId": "A",
+ "step": 600
+ }
+ ],
+ "thresholds": "80, 90",
+ "title": "CPU Utilization",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "avg"
+ },
+ {
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "rgba(50, 172, 45, 0.97)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(245, 54, 54, 0.9)"
+ ],
+ "datasource": "prometheus",
+ "editable": false,
+ "format": "percent",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": true,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "id": 7,
+ "links": [],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 3,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "targets": [
+ {
+ "expr": "((sum(node_memory_MemTotal) - sum(node_memory_MemFree) - sum(node_memory_Buffers) - sum(node_memory_Cached)) / sum(node_memory_MemTotal)) * 100",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "refId": "A",
+ "step": 600
+ }
+ ],
+ "thresholds": "80, 90",
+ "title": "Memory Utilization",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "avg"
+ },
+ {
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "rgba(50, 172, 45, 0.97)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(245, 54, 54, 0.9)"
+ ],
+ "datasource": "prometheus",
+ "editable": false,
+ "format": "percent",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": true,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "id": 9,
+ "links": [],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 3,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "targets": [
+ {
+ "expr": "(sum(node_filesystem_size{device!=\"rootfs\"}) - sum(node_filesystem_free{device!=\"rootfs\"})) / sum(node_filesystem_size{device!=\"rootfs\"})",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "refId": "A",
+ "step": 600
+ }
+ ],
+ "thresholds": "80, 90",
+ "title": "Filesystem Utilization",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "avg"
+ },
+ {
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "rgba(50, 172, 45, 0.97)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(245, 54, 54, 0.9)"
+ ],
+ "datasource": "prometheus",
+ "editable": false,
+ "format": "percent",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": true,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "id": 10,
+ "links": [],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 3,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "targets": [
+ {
+ "expr": "100 - (sum(kube_node_status_capacity_pods) - sum(kube_pod_info)) / sum(kube_node_status_capacity_pods) * 100",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "refId": "A",
+ "step": 600
+ }
+ ],
+ "thresholds": "80, 90",
+ "title": "Pod Utilization",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "avg"
+ }
+ ],
+ "showTitle": true,
+ "title": "Capacity Planning",
+ "titleSize": "h6"
+ }
+ ],
+ "schemaVersion": 14,
+ "sharedCrosshair": false,
+ "style": "dark",
+ "tags": [],
+ "templating": {
+ "list": []
+ },
+ "time": {
+ "from": "now-6h",
+ "to": "now"
+ },
+ "timepicker": {
+ "refresh_intervals": [
+ "5s",
+ "10s",
+ "30s",
+ "1m",
+ "5m",
+ "15m",
+ "30m",
+ "1h",
+ "2h",
+ "1d"
+ ],
+ "time_options": [
+ "5m",
+ "15m",
+ "1h",
+ "6h",
+ "12h",
+ "24h",
+ "2d",
+ "7d",
+ "30d"
+ ]
+ },
+ "timezone": "browser",
+ "title": "Kubernetes Cluster Status",
+ "version": 3
+ }
+ kubernetes-control-plane-status-dashboard.json: |+
+ {
+ "__inputs": [
+ {
+ "description": "",
+ "label": "prometheus",
+ "name": "prometheus",
+ "pluginId": "prometheus",
+ "pluginName": "Prometheus",
+ "type": "datasource"
+ }
+ ],
+ "annotations": {
+ "list": []
+ },
+ "editable": false,
+ "graphTooltip": 0,
+ "hideControls": false,
+ "links": [],
+ "rows": [
+ {
+ "collapse": false,
+ "editable": false,
+ "height": "250px",
+ "panels": [
+ {
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "rgba(245, 54, 54, 0.9)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(50, 172, 45, 0.97)"
+ ],
+ "datasource": "prometheus",
+ "editable": false,
+ "format": "percent",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": true,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "hideTimeOverride": false,
+ "id": 1,
+ "links": [],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 3,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "targets": [
+ {
+ "expr": "(sum(up{job=\"apiserver\"} == 1) / sum(up{job=\"apiserver\"})) * 100",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "refId": "A",
+ "step": 600
+ }
+ ],
+ "thresholds": "50, 80",
+ "title": "API Servers UP",
+ "transparent": false,
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "avg"
+ },
+ {
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "rgba(245, 54, 54, 0.9)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(50, 172, 45, 0.97)"
+ ],
+ "datasource": "prometheus",
+ "editable": false,
+ "format": "percent",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": true,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "hideTimeOverride": false,
+ "id": 2,
+ "links": [],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 3,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "targets": [
+ {
+ "expr": "(sum(up{job=\"kube-controller-manager\"} == 1) / sum(up{job=\"kube-controller-manager\"})) * 100",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "refId": "A",
+ "step": 600
+ }
+ ],
+ "thresholds": "50, 80",
+ "title": "Controller Managers UP",
+ "transparent": false,
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "avg"
+ },
+ {
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "rgba(245, 54, 54, 0.9)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(50, 172, 45, 0.97)"
+ ],
+ "datasource": "prometheus",
+ "editable": false,
+ "format": "percent",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": true,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "hideTimeOverride": false,
+ "id": 3,
+ "links": [],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 3,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "targets": [
+ {
+ "expr": "(sum(up{job=\"kube-scheduler\"} == 1) / sum(up{job=\"kube-scheduler\"})) * 100",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "refId": "A",
+ "step": 600
+ }
+ ],
+ "thresholds": "50, 80",
+ "title": "Schedulers UP",
+ "transparent": false,
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "avg"
+ },
+ {
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "rgba(50, 172, 45, 0.97)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(245, 54, 54, 0.9)"
+ ],
+ "datasource": "prometheus",
+ "editable": false,
+ "format": "percent",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": true,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "hideTimeOverride": false,
+ "id": 4,
+ "links": [],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 3,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "targets": [
+ {
+ "expr": "max(sum by(instance) (rate(apiserver_request_count{code=~\"5..\"}[5m])) / sum by(instance) (rate(apiserver_request_count[5m]))) * 100",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "",
+ "refId": "A",
+ "step": 600
+ }
+ ],
+ "thresholds": "5, 10",
+ "title": "API Server Request Error Rate",
+ "transparent": false,
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "0",
+ "value": "null"
+ }
+ ],
+ "valueName": "avg"
+ }
+ ],
+ "showTitle": false,
+ "title": "Dashboard Row",
+ "titleSize": "h6"
+ },
+ {
+ "collapse": false,
+ "editable": false,
+ "height": "250px",
+ "panels": [
+ {
+ "aliasColors": {},
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "prometheus",
+ "editable": false,
+ "error": false,
+ "fill": 1,
+ "grid": {
+ "threshold1Color": "rgba(216, 200, 27, 0.27)",
+ "threshold2Color": "rgba(234, 112, 112, 0.22)"
+ },
+ "id": 7,
+ "isNew": false,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "hideEmpty": false,
+ "hideZero": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "total": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [],
+ "spaceLength": 10,
+ "span": 12,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "sum by(verb) (rate(apiserver_latency_seconds:quantile[5m]) >= 0)",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "",
+ "refId": "A",
+ "step": 30
+ }
+ ],
+ "title": "API Server Request Latency",
+ "tooltip": {
+ "msResolution": false,
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "mode": "time",
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "format": "short",
+ "logBase": 1,
+ "show": true
+ },
+ {
+ "format": "short",
+ "logBase": 1,
+ "show": true
+ }
+ ]
+ }
+ ],
+ "showTitle": false,
+ "title": "Dashboard Row",
+ "titleSize": "h6"
+ },
+ {
+ "collapse": false,
+ "editable": false,
+ "height": "250px",
+ "panels": [
+ {
+ "aliasColors": {},
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "prometheus",
+ "editable": false,
+ "error": false,
+ "fill": 1,
+ "grid": {
+ "threshold1Color": "rgba(216, 200, 27, 0.27)",
+ "threshold2Color": "rgba(234, 112, 112, 0.22)"
+ },
+ "id": 5,
+ "isNew": false,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "hideEmpty": false,
+ "hideZero": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "total": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [],
+ "spaceLength": 10,
+ "span": 6,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "cluster:scheduler_e2e_scheduling_latency_seconds:quantile",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "refId": "A",
+ "step": 60
+ }
+ ],
+ "title": "End to End Scheduling Latency",
+ "tooltip": {
+ "msResolution": false,
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "mode": "time",
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "format": "short",
+ "logBase": 1,
+ "show": true
+ },
+ {
+ "format": "dtdurations",
+ "logBase": 1,
+ "show": true
+ }
+ ]
+ },
+ {
+ "aliasColors": {},
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "prometheus",
+ "editable": false,
+ "error": false,
+ "fill": 1,
+ "grid": {
+ "threshold1Color": "rgba(216, 200, 27, 0.27)",
+ "threshold2Color": "rgba(234, 112, 112, 0.22)"
+ },
+ "id": 6,
+ "isNew": false,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "hideEmpty": false,
+ "hideZero": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "total": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [],
+ "spaceLength": 10,
+ "span": 6,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "sum by(instance) (rate(apiserver_request_count{code!~\"2..\"}[5m]))",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "Error Rate",
+ "refId": "A",
+ "step": 60
+ },
+ {
+ "expr": "sum by(instance) (rate(apiserver_request_count[5m]))",
+ "format": "time_series",
+ "intervalFactor": 2,
+ "legendFormat": "Request Rate",
+ "refId": "B",
+ "step": 60
+ }
+ ],
+ "title": "API Server Request Rates",
+ "tooltip": {
+ "msResolution": false,
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "mode": "time",
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "format": "short",
+ "logBase": 1,
+ "show": true
+ },
+ {
+ "format": "short",
+ "logBase": 1,
+ "show": true
+ }
+ ]
+ }
+ ],
+ "showTitle": false,
+ "title": "Dashboard Row",
+ "titleSize": "h6"
+ }
+ ],
+ "schemaVersion": 14,
+ "sharedCrosshair": false,
+ "style": "dark",
+ "tags": [],
+ "templating": {
+ "list": []
+ },
+ "time": {
+ "from": "now-6h",
+ "to": "now"
+ },
+ "timepicker": {
+ "refresh_intervals": [
+ "5s",
+ "10s",
+ "30s",
+ "1m",
+ "5m",
+ "15m",
+ "30m",
+ "1h",
+ "2h",
+ "1d"
+ ],
+ "time_options": [
+ "5m",
+ "15m",
+ "1h",
+ "6h",
+ "12h",
+ "24h",
+ "2d",
+ "7d",
+ "30d"
+ ]
+ },
+ "timezone": "browser",
+ "title": "Kubernetes Control Plane Status",
+ "version": 3
+ }
+ kubernetes-resource-requests-dashboard.json: |+
+ {
+ "__inputs": [
+ {
+ "description": "",
+ "label": "prometheus",
+ "name": "prometheus",
+ "pluginId": "prometheus",
+ "pluginName": "Prometheus",
+ "type": "datasource"
+ }
+ ],
+ "annotations": {
+ "list": []
+ },
+ "editable": false,
+ "graphTooltip": 0,
+ "hideControls": false,
+ "links": [],
+ "refresh": false,
+ "rows": [
+ {
+ "collapse": false,
+ "editable": false,
+ "height": "300px",
+ "panels": [
+ {
+ "aliasColors": {},
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "prometheus",
+ "description": "This represents the total [CPU resource requests](https://kubernetes.io/docs/concepts/configuration/manage-compute-resources-container/#meaning-of-cpu) in the cluster.\nFor comparison the total [allocatable CPU cores](https://github.com/kubernetes/community/blob/master/contributors/design-proposals/node-allocatable.md) is also shown.",
+ "editable": false,
+ "error": false,
+ "fill": 1,
+ "grid": {
+ "threshold1Color": "rgba(216, 200, 27, 0.27)",
+ "threshold2Color": "rgba(234, 112, 112, 0.22)"
+ },
+ "id": 1,
+ "isNew": false,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "hideEmpty": false,
+ "hideZero": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "total": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [],
+ "spaceLength": 10,
+ "span": 9,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "min(sum(kube_node_status_allocatable_cpu_cores) by (instance))",
+ "hide": false,
+ "intervalFactor": 2,
+ "legendFormat": "Allocatable CPU Cores",
+ "refId": "A",
+ "step": 20
+ },
+ {
+ "expr": "max(sum(kube_pod_container_resource_requests_cpu_cores) by (instance))",
+ "hide": false,
+ "intervalFactor": 2,
+ "legendFormat": "Requested CPU Cores",
+ "refId": "B",
+ "step": 20
+ }
+ ],
+ "title": "CPU Cores",
+ "tooltip": {
+ "msResolution": false,
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "mode": "time",
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "format": "short",
+ "label": "CPU Cores",
+ "logBase": 1,
+ "show": true
+ },
+ {
+ "format": "short",
+ "logBase": 1,
+ "show": true
+ }
+ ]
+ },
+ {
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "rgba(50, 172, 45, 0.97)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(245, 54, 54, 0.9)"
+ ],
+ "datasource": "prometheus",
+ "editable": false,
+ "format": "percent",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": true,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "hideTimeOverride": false,
+ "id": 2,
+ "links": [],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 3,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": true
+ },
+ "targets": [
+ {
+ "expr": "max(sum(kube_pod_container_resource_requests_cpu_cores) by (instance)) / min(sum(kube_node_status_allocatable_cpu_cores) by (instance)) * 100",
+ "intervalFactor": 2,
+ "legendFormat": "",
+ "refId": "A",
+ "step": 240
+ }
+ ],
+ "thresholds": "80, 90",
+ "title": "CPU Cores",
+ "transparent": false,
+ "type": "singlestat",
+ "valueFontSize": "110%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "avg"
+ }
+ ],
+ "showTitle": false,
+ "title": "CPU Cores",
+ "titleSize": "h6"
+ },
+ {
+ "collapse": false,
+ "editable": false,
+ "height": "300px",
+ "panels": [
+ {
+ "aliasColors": {},
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "prometheus",
+ "description": "This represents the total [memory resource requests](https://kubernetes.io/docs/concepts/configuration/manage-compute-resources-container/#meaning-of-memory) in the cluster.\nFor comparison the total [allocatable memory](https://github.com/kubernetes/community/blob/master/contributors/design-proposals/node-allocatable.md) is also shown.",
+ "editable": false,
+ "error": false,
+ "fill": 1,
+ "grid": {
+ "threshold1Color": "rgba(216, 200, 27, 0.27)",
+ "threshold2Color": "rgba(234, 112, 112, 0.22)"
+ },
+ "id": 3,
+ "isNew": false,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "hideEmpty": false,
+ "hideZero": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "total": false
+ },
+ "lines": true,
+ "linewidth": 1,
+ "links": [],
+ "nullPointMode": "null",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [],
+ "spaceLength": 10,
+ "span": 9,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "min(sum(kube_node_status_allocatable_memory_bytes) by (instance))",
+ "hide": false,
+ "intervalFactor": 2,
+ "legendFormat": "Allocatable Memory",
+ "refId": "A",
+ "step": 20
+ },
+ {
+ "expr": "max(sum(kube_pod_container_resource_requests_memory_bytes) by (instance))",
+ "hide": false,
+ "intervalFactor": 2,
+ "legendFormat": "Requested Memory",
+ "refId": "B",
+ "step": 20
+ }
+ ],
+ "title": "Memory",
+ "tooltip": {
+ "msResolution": false,
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "mode": "time",
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "format": "bytes",
+ "label": "Memory",
+ "logBase": 1,
+ "show": true
+ },
+ {
+ "format": "short",
+ "logBase": 1,
+ "show": true
+ }
+ ]
+ },
+ {
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "rgba(50, 172, 45, 0.97)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(245, 54, 54, 0.9)"
+ ],
+ "datasource": "prometheus",
+ "editable": false,
+ "format": "percent",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": true,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "hideTimeOverride": false,
+ "id": 4,
+ "links": [],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 3,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": true
+ },
+ "targets": [
+ {
+ "expr": "max(sum(kube_pod_container_resource_requests_memory_bytes) by (instance)) / min(sum(kube_node_status_allocatable_memory_bytes) by (instance)) * 100",
+ "intervalFactor": 2,
+ "legendFormat": "",
+ "refId": "A",
+ "step": 240
+ }
+ ],
+ "thresholds": "80, 90",
+ "title": "Memory",
+ "transparent": false,
+ "type": "singlestat",
+ "valueFontSize": "110%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "avg"
+ }
+ ],
+ "showTitle": false,
+ "title": "Memory",
+ "titleSize": "h6"
+ }
+ ],
+ "schemaVersion": 14,
+ "sharedCrosshair": false,
+ "style": "dark",
+ "tags": [],
+ "templating": {
+ "list": []
+ },
+ "time": {
+ "from": "now-3h",
+ "to": "now"
+ },
+ "timepicker": {
+ "refresh_intervals": [
+ "5s",
+ "10s",
+ "30s",
+ "1m",
+ "5m",
+ "15m",
+ "30m",
+ "1h",
+ "2h",
+ "1d"
+ ],
+ "time_options": [
+ "5m",
+ "15m",
+ "1h",
+ "6h",
+ "12h",
+ "24h",
+ "2d",
+ "7d",
+ "30d"
+ ]
+ },
+ "timezone": "browser",
+ "title": "Kubernetes Resource Requests",
+ "version": 2
+ }
+ nodes-dashboard.json: |+
+ {
+ "__inputs": [
+ {
+ "description": "",
+ "label": "prometheus",
+ "name": "prometheus",
+ "pluginId": "prometheus",
+ "pluginName": "Prometheus",
+ "type": "datasource"
+ }
+ ],
+ "annotations": {
+ "list": []
+ },
+ "description": "Dashboard to get an overview of one server",
+ "editable": false,
+ "gnetId": 22,
+ "graphTooltip": 0,
+ "hideControls": false,
+ "links": [],
+ "refresh": false,
+ "rows": [
+ {
+ "collapse": false,
+ "editable": false,
+ "height": "250px",
+ "panels": [
+ {
+ "aliasColors": {},
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "prometheus",
+ "editable": false,
+ "error": false,
+ "fill": 1,
+ "grid": {
+ "threshold1Color": "rgba(216, 200, 27, 0.27)",
+ "threshold2Color": "rgba(234, 112, 112, 0.22)"
+ },
+ "id": 3,
+ "isNew": false,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "hideEmpty": false,
+ "hideZero": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "total": false
+ },
+ "lines": true,
+ "linewidth": 2,
+ "links": [],
+ "nullPointMode": "connected",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [],
+ "spaceLength": 10,
+ "span": 6,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "100 - (avg by (cpu) (irate(node_cpu{mode=\"idle\", instance=\"$server\"}[5m])) * 100)",
+ "hide": false,
+ "intervalFactor": 10,
+ "legendFormat": "{{cpu}}",
+ "refId": "A",
+ "step": 50
+ }
+ ],
+ "title": "Idle CPU",
+ "tooltip": {
+ "msResolution": false,
+ "shared": true,
+ "sort": 0,
+ "value_type": "cumulative"
+ },
+ "type": "graph",
+ "xaxis": {
+ "mode": "time",
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "format": "percent",
+ "label": "cpu usage",
+ "logBase": 1,
+ "max": 100,
+ "min": 0,
+ "show": true
+ },
+ {
+ "format": "short",
+ "logBase": 1,
+ "show": true
+ }
+ ]
+ },
+ {
+ "aliasColors": {},
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "prometheus",
+ "editable": false,
+ "error": false,
+ "fill": 1,
+ "grid": {
+ "threshold1Color": "rgba(216, 200, 27, 0.27)",
+ "threshold2Color": "rgba(234, 112, 112, 0.22)"
+ },
+ "id": 9,
+ "isNew": false,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "hideEmpty": false,
+ "hideZero": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "total": false
+ },
+ "lines": true,
+ "linewidth": 2,
+ "links": [],
+ "nullPointMode": "connected",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [],
+ "spaceLength": 10,
+ "span": 6,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "node_load1{instance=\"$server\"}",
+ "intervalFactor": 4,
+ "legendFormat": "load 1m",
+ "refId": "A",
+ "step": 20,
+ "target": ""
+ },
+ {
+ "expr": "node_load5{instance=\"$server\"}",
+ "intervalFactor": 4,
+ "legendFormat": "load 5m",
+ "refId": "B",
+ "step": 20,
+ "target": ""
+ },
+ {
+ "expr": "node_load15{instance=\"$server\"}",
+ "intervalFactor": 4,
+ "legendFormat": "load 15m",
+ "refId": "C",
+ "step": 20,
+ "target": ""
+ }
+ ],
+ "title": "System Load",
+ "tooltip": {
+ "msResolution": false,
+ "shared": true,
+ "sort": 0,
+ "value_type": "cumulative"
+ },
+ "type": "graph",
+ "xaxis": {
+ "mode": "time",
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "format": "percentunit",
+ "logBase": 1,
+ "show": true
+ },
+ {
+ "format": "short",
+ "logBase": 1,
+ "show": true
+ }
+ ]
+ }
+ ],
+ "showTitle": false,
+ "title": "New Row",
+ "titleSize": "h6"
+ },
+ {
+ "collapse": false,
+ "editable": false,
+ "height": "250px",
+ "panels": [
+ {
+ "aliasColors": {},
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "prometheus",
+ "editable": false,
+ "error": false,
+ "fill": 1,
+ "grid": {
+ "threshold1Color": "rgba(216, 200, 27, 0.27)",
+ "threshold2Color": "rgba(234, 112, 112, 0.22)"
+ },
+ "id": 4,
+ "isNew": false,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "hideEmpty": false,
+ "hideZero": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "total": false
+ },
+ "lines": true,
+ "linewidth": 2,
+ "links": [],
+ "nullPointMode": "connected",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [
+ {
+ "alias": "node_memory_SwapFree{instance=\"172.17.0.1:9100\",job=\"prometheus\"}",
+ "yaxis": 2
+ }
+ ],
+ "spaceLength": 10,
+ "span": 9,
+ "stack": true,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "node_memory_MemTotal{instance=\"$server\"} - node_memory_MemFree{instance=\"$server\"} - node_memory_Buffers{instance=\"$server\"} - node_memory_Cached{instance=\"$server\"}",
+ "hide": false,
+ "interval": "",
+ "intervalFactor": 2,
+ "legendFormat": "memory used",
+ "metric": "",
+ "refId": "C",
+ "step": 10
+ },
+ {
+ "expr": "node_memory_Buffers{instance=\"$server\"}",
+ "interval": "",
+ "intervalFactor": 2,
+ "legendFormat": "memory buffers",
+ "metric": "",
+ "refId": "E",
+ "step": 10
+ },
+ {
+ "expr": "node_memory_Cached{instance=\"$server\"}",
+ "intervalFactor": 2,
+ "legendFormat": "memory cached",
+ "metric": "",
+ "refId": "F",
+ "step": 10
+ },
+ {
+ "expr": "node_memory_MemFree{instance=\"$server\"}",
+ "intervalFactor": 2,
+ "legendFormat": "memory free",
+ "metric": "",
+ "refId": "D",
+ "step": 10
+ }
+ ],
+ "title": "Memory Usage",
+ "tooltip": {
+ "msResolution": false,
+ "shared": true,
+ "sort": 0,
+ "value_type": "individual"
+ },
+ "type": "graph",
+ "xaxis": {
+ "mode": "time",
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "format": "bytes",
+ "logBase": 1,
+ "min": "0",
+ "show": true
+ },
+ {
+ "format": "short",
+ "logBase": 1,
+ "show": true
+ }
+ ]
+ },
+ {
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "rgba(50, 172, 45, 0.97)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(245, 54, 54, 0.9)"
+ ],
+ "datasource": "prometheus",
+ "editable": false,
+ "format": "percent",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": true,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "hideTimeOverride": false,
+ "id": 5,
+ "links": [],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 3,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "targets": [
+ {
+ "expr": "((node_memory_MemTotal{instance=\"$server\"} - node_memory_MemFree{instance=\"$server\"} - node_memory_Buffers{instance=\"$server\"} - node_memory_Cached{instance=\"$server\"}) / node_memory_MemTotal{instance=\"$server\"}) * 100",
+ "intervalFactor": 2,
+ "refId": "A",
+ "step": 60,
+ "target": ""
+ }
+ ],
+ "thresholds": "80, 90",
+ "title": "Memory Usage",
+ "transparent": false,
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "avg"
+ }
+ ],
+ "showTitle": false,
+ "title": "New Row",
+ "titleSize": "h6"
+ },
+ {
+ "collapse": false,
+ "editable": false,
+ "height": "250px",
+ "panels": [
+ {
+ "aliasColors": {},
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "prometheus",
+ "editable": false,
+ "error": false,
+ "fill": 1,
+ "grid": {
+ "threshold1Color": "rgba(216, 200, 27, 0.27)",
+ "threshold2Color": "rgba(234, 112, 112, 0.22)"
+ },
+ "id": 6,
+ "isNew": true,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "hideEmpty": false,
+ "hideZero": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "total": false
+ },
+ "lines": true,
+ "linewidth": 2,
+ "links": [],
+ "nullPointMode": "connected",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [
+ {
+ "alias": "read",
+ "yaxis": 1
+ },
+ {
+ "alias": "{instance=\"172.17.0.1:9100\"}",
+ "yaxis": 2
+ },
+ {
+ "alias": "io time",
+ "yaxis": 2
+ }
+ ],
+ "spaceLength": 10,
+ "span": 9,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "sum by (instance) (rate(node_disk_bytes_read{instance=\"$server\"}[2m]))",
+ "hide": false,
+ "intervalFactor": 4,
+ "legendFormat": "read",
+ "refId": "A",
+ "step": 20,
+ "target": ""
+ },
+ {
+ "expr": "sum by (instance) (rate(node_disk_bytes_written{instance=\"$server\"}[2m]))",
+ "intervalFactor": 4,
+ "legendFormat": "written",
+ "refId": "B",
+ "step": 20
+ },
+ {
+ "expr": "sum by (instance) (rate(node_disk_io_time_ms{instance=\"$server\"}[2m]))",
+ "intervalFactor": 4,
+ "legendFormat": "io time",
+ "refId": "C",
+ "step": 20
+ }
+ ],
+ "title": "Disk I/O",
+ "tooltip": {
+ "msResolution": false,
+ "shared": true,
+ "sort": 0,
+ "value_type": "cumulative"
+ },
+ "type": "graph",
+ "xaxis": {
+ "mode": "time",
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "format": "bytes",
+ "logBase": 1,
+ "show": true
+ },
+ {
+ "format": "ms",
+ "logBase": 1,
+ "show": true
+ }
+ ]
+ },
+ {
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "rgba(50, 172, 45, 0.97)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(245, 54, 54, 0.9)"
+ ],
+ "datasource": "prometheus",
+ "editable": false,
+ "format": "percentunit",
+ "gauge": {
+ "maxValue": 1,
+ "minValue": 0,
+ "show": true,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "hideTimeOverride": false,
+ "id": 7,
+ "links": [],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 3,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "targets": [
+ {
+ "expr": "(sum(node_filesystem_size{device!=\"rootfs\",instance=\"$server\"}) - sum(node_filesystem_free{device!=\"rootfs\",instance=\"$server\"})) / sum(node_filesystem_size{device!=\"rootfs\",instance=\"$server\"})",
+ "intervalFactor": 2,
+ "refId": "A",
+ "step": 60,
+ "target": ""
+ }
+ ],
+ "thresholds": "0.75, 0.9",
+ "title": "Disk Space Usage",
+ "transparent": false,
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "current"
+ }
+ ],
+ "showTitle": false,
+ "title": "New Row",
+ "titleSize": "h6"
+ },
+ {
+ "collapse": false,
+ "editable": false,
+ "height": "250px",
+ "panels": [
+ {
+ "aliasColors": {},
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "prometheus",
+ "editable": false,
+ "error": false,
+ "fill": 1,
+ "grid": {
+ "threshold1Color": "rgba(216, 200, 27, 0.27)",
+ "threshold2Color": "rgba(234, 112, 112, 0.22)"
+ },
+ "id": 8,
+ "isNew": false,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "hideEmpty": false,
+ "hideZero": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "total": false
+ },
+ "lines": true,
+ "linewidth": 2,
+ "links": [],
+ "nullPointMode": "connected",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [
+ {
+ "alias": "transmitted",
+ "yaxis": 2
+ }
+ ],
+ "spaceLength": 10,
+ "span": 6,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "rate(node_network_receive_bytes{instance=\"$server\",device!~\"lo\"}[5m])",
+ "hide": false,
+ "intervalFactor": 2,
+ "legendFormat": "{{device}}",
+ "refId": "A",
+ "step": 10,
+ "target": ""
+ }
+ ],
+ "title": "Network Received",
+ "tooltip": {
+ "msResolution": false,
+ "shared": true,
+ "sort": 0,
+ "value_type": "cumulative"
+ },
+ "type": "graph",
+ "xaxis": {
+ "mode": "time",
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "format": "bytes",
+ "logBase": 1,
+ "show": true
+ },
+ {
+ "format": "bytes",
+ "logBase": 1,
+ "show": true
+ }
+ ]
+ },
+ {
+ "aliasColors": {},
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "prometheus",
+ "editable": false,
+ "error": false,
+ "fill": 1,
+ "grid": {
+ "threshold1Color": "rgba(216, 200, 27, 0.27)",
+ "threshold2Color": "rgba(234, 112, 112, 0.22)"
+ },
+ "id": 10,
+ "isNew": false,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "hideEmpty": false,
+ "hideZero": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "total": false
+ },
+ "lines": true,
+ "linewidth": 2,
+ "links": [],
+ "nullPointMode": "connected",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [
+ {
+ "alias": "transmitted",
+ "yaxis": 2
+ }
+ ],
+ "spaceLength": 10,
+ "span": 6,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "rate(node_network_transmit_bytes{instance=\"$server\",device!~\"lo\"}[5m])",
+ "hide": false,
+ "intervalFactor": 2,
+ "legendFormat": "{{device}}",
+ "refId": "B",
+ "step": 10,
+ "target": ""
+ }
+ ],
+ "title": "Network Transmitted",
+ "tooltip": {
+ "msResolution": false,
+ "shared": true,
+ "sort": 0,
+ "value_type": "cumulative"
+ },
+ "type": "graph",
+ "xaxis": {
+ "mode": "time",
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "format": "bytes",
+ "logBase": 1,
+ "show": true
+ },
+ {
+ "format": "bytes",
+ "logBase": 1,
+ "show": true
+ }
+ ]
+ }
+ ],
+ "showTitle": false,
+ "title": "New Row",
+ "titleSize": "h6"
+ }
+ ],
+ "schemaVersion": 14,
+ "sharedCrosshair": false,
+ "style": "dark",
+ "tags": [],
+ "templating": {
+ "list": [
+ {
+ "allValue": null,
+ "current": {},
+ "datasource": "prometheus",
+ "hide": 0,
+ "includeAll": false,
+ "label": null,
+ "multi": false,
+ "name": "server",
+ "options": [],
+ "query": "label_values(node_boot_time, instance)",
+ "refresh": 1,
+ "regex": "",
+ "sort": 0,
+ "tagValuesQuery": "",
+ "tags": [],
+ "tagsQuery": "",
+ "type": "query",
+ "useTags": false
+ }
+ ]
+ },
+ "time": {
+ "from": "now-1h",
+ "to": "now"
+ },
+ "timepicker": {
+ "refresh_intervals": [
+ "5s",
+ "10s",
+ "30s",
+ "1m",
+ "5m",
+ "15m",
+ "30m",
+ "1h",
+ "2h",
+ "1d"
+ ],
+ "time_options": [
+ "5m",
+ "15m",
+ "1h",
+ "6h",
+ "12h",
+ "24h",
+ "2d",
+ "7d",
+ "30d"
+ ]
+ },
+ "timezone": "browser",
+ "title": "Nodes",
+ "version": 2
+ }
+ pods-dashboard.json: |+
+ {
+ "__inputs": [
+ {
+ "description": "",
+ "label": "prometheus",
+ "name": "prometheus",
+ "pluginId": "prometheus",
+ "pluginName": "Prometheus",
+ "type": "datasource"
+ }
+ ],
+ "annotations": {
+ "list": []
+ },
+ "editable": false,
+ "graphTooltip": 1,
+ "hideControls": false,
+ "links": [],
+ "refresh": false,
+ "rows": [
+ {
+ "collapse": false,
+ "editable": false,
+ "height": "250px",
+ "panels": [
+ {
+ "aliasColors": {},
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "prometheus",
+ "editable": false,
+ "error": false,
+ "fill": 1,
+ "grid": {
+ "threshold1Color": "rgba(216, 200, 27, 0.27)",
+ "threshold2Color": "rgba(234, 112, 112, 0.22)"
+ },
+ "id": 1,
+ "isNew": false,
+ "legend": {
+ "alignAsTable": true,
+ "avg": true,
+ "current": true,
+ "hideEmpty": false,
+ "hideZero": false,
+ "max": false,
+ "min": false,
+ "rightSide": true,
+ "show": true,
+ "total": false,
+ "values": true
+ },
+ "lines": true,
+ "linewidth": 2,
+ "links": [],
+ "nullPointMode": "connected",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [],
+ "spaceLength": 10,
+ "span": 12,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "sum by(container_name) (container_memory_usage_bytes{pod_name=\"$pod\", container_name=~\"$container\", container_name!=\"POD\"})",
+ "interval": "10s",
+ "intervalFactor": 1,
+ "legendFormat": "Current: {{ container_name }}",
+ "metric": "container_memory_usage_bytes",
+ "refId": "A",
+ "step": 15
+ },
+ {
+ "expr": "kube_pod_container_resource_requests_memory_bytes{pod=\"$pod\", container=~\"$container\"}",
+ "interval": "10s",
+ "intervalFactor": 2,
+ "legendFormat": "Requested: {{ container }}",
+ "metric": "kube_pod_container_resource_requests_memory_bytes",
+ "refId": "B",
+ "step": 20
+ },
+ {
+ "expr": "kube_pod_container_resource_limits_memory_bytes{pod=\"$pod\", container=~\"$container\"}",
+ "interval": "10s",
+ "intervalFactor": 2,
+ "legendFormat": "Limit: {{ container }}",
+ "metric": "kube_pod_container_resource_limits_memory_bytes",
+ "refId": "C",
+ "step": 20
+ }
+ ],
+ "title": "Memory Usage",
+ "tooltip": {
+ "msResolution": true,
+ "shared": true,
+ "sort": 0,
+ "value_type": "cumulative"
+ },
+ "type": "graph",
+ "xaxis": {
+ "mode": "time",
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "format": "bytes",
+ "logBase": 1,
+ "show": true
+ },
+ {
+ "format": "short",
+ "logBase": 1,
+ "show": true
+ }
+ ]
+ }
+ ],
+ "showTitle": false,
+ "title": "Row",
+ "titleSize": "h6"
+ },
+ {
+ "collapse": false,
+ "editable": false,
+ "height": "250px",
+ "panels": [
+ {
+ "aliasColors": {},
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "prometheus",
+ "editable": false,
+ "error": false,
+ "fill": 1,
+ "grid": {
+ "threshold1Color": "rgba(216, 200, 27, 0.27)",
+ "threshold2Color": "rgba(234, 112, 112, 0.22)"
+ },
+ "id": 2,
+ "isNew": false,
+ "legend": {
+ "alignAsTable": true,
+ "avg": true,
+ "current": true,
+ "hideEmpty": false,
+ "hideZero": false,
+ "max": false,
+ "min": false,
+ "rightSide": true,
+ "show": true,
+ "total": false,
+ "values": true
+ },
+ "lines": true,
+ "linewidth": 2,
+ "links": [],
+ "nullPointMode": "connected",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [],
+ "spaceLength": 10,
+ "span": 12,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "sum by (container_name)(rate(container_cpu_usage_seconds_total{image!=\"\",container_name!=\"POD\",pod_name=\"$pod\"}[1m]))",
+ "intervalFactor": 2,
+ "legendFormat": "{{ container_name }}",
+ "refId": "A",
+ "step": 30
+ },
+ {
+ "expr": "kube_pod_container_resource_requests_cpu_cores{pod=\"$pod\", container=~\"$container\"}",
+ "interval": "10s",
+ "intervalFactor": 2,
+ "legendFormat": "Requested: {{ container }}",
+ "metric": "kube_pod_container_resource_requests_cpu_cores",
+ "refId": "B",
+ "step": 20
+ },
+ {
+ "expr": "kube_pod_container_resource_limits_cpu_cores{pod=\"$pod\", container=~\"$container\"}",
+ "interval": "10s",
+ "intervalFactor": 2,
+ "legendFormat": "Limit: {{ container }}",
+ "metric": "kube_pod_container_resource_limits_memory_bytes",
+ "refId": "C",
+ "step": 20
+ }
+ ],
+ "title": "CPU Usage",
+ "tooltip": {
+ "msResolution": true,
+ "shared": true,
+ "sort": 0,
+ "value_type": "cumulative"
+ },
+ "type": "graph",
+ "xaxis": {
+ "mode": "time",
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "format": "short",
+ "logBase": 1,
+ "show": true
+ },
+ {
+ "format": "short",
+ "logBase": 1,
+ "show": true
+ }
+ ]
+ }
+ ],
+ "showTitle": false,
+ "title": "Row",
+ "titleSize": "h6"
+ },
+ {
+ "collapse": false,
+ "editable": false,
+ "height": "250px",
+ "panels": [
+ {
+ "aliasColors": {},
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "prometheus",
+ "editable": false,
+ "error": false,
+ "fill": 1,
+ "grid": {
+ "threshold1Color": "rgba(216, 200, 27, 0.27)",
+ "threshold2Color": "rgba(234, 112, 112, 0.22)"
+ },
+ "id": 3,
+ "isNew": false,
+ "legend": {
+ "alignAsTable": true,
+ "avg": true,
+ "current": true,
+ "hideEmpty": false,
+ "hideZero": false,
+ "max": false,
+ "min": false,
+ "rightSide": true,
+ "show": true,
+ "total": false,
+ "values": true
+ },
+ "lines": true,
+ "linewidth": 2,
+ "links": [],
+ "nullPointMode": "connected",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [],
+ "spaceLength": 10,
+ "span": 12,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "sort_desc(sum by (pod_name) (rate(container_network_receive_bytes_total{pod_name=\"$pod\"}[1m])))",
+ "intervalFactor": 2,
+ "legendFormat": "{{ pod_name }}",
+ "refId": "A",
+ "step": 30
+ }
+ ],
+ "title": "Network I/O",
+ "tooltip": {
+ "msResolution": true,
+ "shared": true,
+ "sort": 0,
+ "value_type": "cumulative"
+ },
+ "type": "graph",
+ "xaxis": {
+ "mode": "time",
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "format": "bytes",
+ "logBase": 1,
+ "show": true
+ },
+ {
+ "format": "short",
+ "logBase": 1,
+ "show": true
+ }
+ ]
+ }
+ ],
+ "showTitle": false,
+ "title": "New Row",
+ "titleSize": "h6"
+ }
+ ],
+ "schemaVersion": 14,
+ "sharedCrosshair": false,
+ "style": "dark",
+ "tags": [],
+ "templating": {
+ "list": [
+ {
+ "allValue": ".*",
+ "current": {},
+ "datasource": "prometheus",
+ "hide": 0,
+ "includeAll": true,
+ "label": "Namespace",
+ "multi": false,
+ "name": "namespace",
+ "options": [],
+ "query": "label_values(kube_pod_info, namespace)",
+ "refresh": 1,
+ "regex": "",
+ "sort": 0,
+ "tagValuesQuery": "",
+ "tags": [],
+ "tagsQuery": "",
+ "type": "query",
+ "useTags": false
+ },
+ {
+ "allValue": null,
+ "current": {},
+ "datasource": "prometheus",
+ "hide": 0,
+ "includeAll": false,
+ "label": "Pod",
+ "multi": false,
+ "name": "pod",
+ "options": [],
+ "query": "label_values(kube_pod_info{namespace=~\"$namespace\"}, pod)",
+ "refresh": 1,
+ "regex": "",
+ "sort": 0,
+ "tagValuesQuery": "",
+ "tags": [],
+ "tagsQuery": "",
+ "type": "query",
+ "useTags": false
+ },
+ {
+ "allValue": ".*",
+ "current": {},
+ "datasource": "prometheus",
+ "hide": 0,
+ "includeAll": true,
+ "label": "Container",
+ "multi": false,
+ "name": "container",
+ "options": [],
+ "query": "label_values(kube_pod_container_info{namespace=\"$namespace\", pod=\"$pod\"}, container)",
+ "refresh": 1,
+ "regex": "",
+ "sort": 0,
+ "tagValuesQuery": "",
+ "tags": [],
+ "tagsQuery": "",
+ "type": "query",
+ "useTags": false
+ }
+ ]
+ },
+ "time": {
+ "from": "now-6h",
+ "to": "now"
+ },
+ "timepicker": {
+ "refresh_intervals": [
+ "5s",
+ "10s",
+ "30s",
+ "1m",
+ "5m",
+ "15m",
+ "30m",
+ "1h",
+ "2h",
+ "1d"
+ ],
+ "time_options": [
+ "5m",
+ "15m",
+ "1h",
+ "6h",
+ "12h",
+ "24h",
+ "2d",
+ "7d",
+ "30d"
+ ]
+ },
+ "timezone": "browser",
+ "title": "Pods",
+ "version": 1
+ }
+ statefulset-dashboard.json: |+
+ {
+ "__inputs": [
+ {
+ "description": "",
+ "label": "prometheus",
+ "name": "prometheus",
+ "pluginId": "prometheus",
+ "pluginName": "Prometheus",
+ "type": "datasource"
+ }
+ ],
+ "annotations": {
+ "list": []
+ },
+ "editable": false,
+ "graphTooltip": 1,
+ "hideControls": false,
+ "links": [],
+ "rows": [
+ {
+ "collapse": false,
+ "editable": false,
+ "height": "200px",
+ "panels": [
+ {
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "rgba(245, 54, 54, 0.9)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(50, 172, 45, 0.97)"
+ ],
+ "datasource": "prometheus",
+ "editable": false,
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "id": 8,
+ "links": [],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "postfix": "cores",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 4,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": true
+ },
+ "targets": [
+ {
+ "expr": "sum(rate(container_cpu_usage_seconds_total{namespace=\"$statefulset_namespace\",pod_name=~\"$statefulset_name.*\"}[3m]))",
+ "intervalFactor": 2,
+ "refId": "A",
+ "step": 600
+ }
+ ],
+ "title": "CPU",
+ "type": "singlestat",
+ "valueFontSize": "110%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "avg"
+ },
+ {
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "rgba(245, 54, 54, 0.9)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(50, 172, 45, 0.97)"
+ ],
+ "datasource": "prometheus",
+ "editable": false,
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "id": 9,
+ "links": [],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "postfix": "GB",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "80%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 4,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": true
+ },
+ "targets": [
+ {
+ "expr": "sum(container_memory_usage_bytes{namespace=\"$statefulset_namespace\",pod_name=~\"$statefulset_name.*\"}) / 1024^3",
+ "intervalFactor": 2,
+ "refId": "A",
+ "step": 600
+ }
+ ],
+ "title": "Memory",
+ "type": "singlestat",
+ "valueFontSize": "110%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "avg"
+ },
+ {
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "rgba(245, 54, 54, 0.9)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(50, 172, 45, 0.97)"
+ ],
+ "datasource": "prometheus",
+ "editable": false,
+ "format": "Bps",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": false
+ },
+ "id": 7,
+ "links": [],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "postfix": "",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 4,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": true
+ },
+ "targets": [
+ {
+ "expr": "sum(rate(container_network_transmit_bytes_total{namespace=\"$statefulset_namespace\",pod_name=~\"$statefulset_name.*\"}[3m])) + sum(rate(container_network_receive_bytes_total{namespace=\"$statefulset_namespace\",pod_name=~\"$statefulset_name.*\"}[3m]))",
+ "intervalFactor": 2,
+ "refId": "A",
+ "step": 600
+ }
+ ],
+ "title": "Network",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "avg"
+ }
+ ],
+ "showTitle": false,
+ "title": "Dashboard Row",
+ "titleSize": "h6"
+ },
+ {
+ "collapse": false,
+ "editable": false,
+ "height": "100px",
+ "panels": [
+ {
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "rgba(245, 54, 54, 0.9)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(50, 172, 45, 0.97)"
+ ],
+ "datasource": "prometheus",
+ "editable": false,
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": false
+ },
+ "id": 5,
+ "links": [],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 3,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "targets": [
+ {
+ "expr": "max(kube_statefulset_replicas{statefulset=\"$statefulset_name\",namespace=\"$statefulset_namespace\"}) without (instance, pod)",
+ "intervalFactor": 2,
+ "metric": "kube_statefulset_replicas",
+ "refId": "A",
+ "step": 600
+ }
+ ],
+ "title": "Desired Replicas",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "avg"
+ },
+ {
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "rgba(245, 54, 54, 0.9)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(50, 172, 45, 0.97)"
+ ],
+ "datasource": "prometheus",
+ "editable": false,
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "id": 6,
+ "links": [],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 3,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "targets": [
+ {
+ "expr": "min(kube_statefulset_status_replicas{statefulset=\"$statefulset_name\",namespace=\"$statefulset_namespace\"}) without (instance, pod)",
+ "intervalFactor": 2,
+ "refId": "A",
+ "step": 600
+ }
+ ],
+ "title": "Available Replicas",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "avg"
+ },
+ {
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "rgba(245, 54, 54, 0.9)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(50, 172, 45, 0.97)"
+ ],
+ "datasource": "prometheus",
+ "editable": false,
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "id": 3,
+ "links": [],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 3,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "targets": [
+ {
+ "expr": "max(kube_statefulset_status_observed_generation{statefulset=\"$statefulset_name\",namespace=\"$statefulset_namespace\"}) without (instance, pod)",
+ "intervalFactor": 2,
+ "refId": "A",
+ "step": 600
+ }
+ ],
+ "title": "Observed Generation",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "avg"
+ },
+ {
+ "colorBackground": false,
+ "colorValue": false,
+ "colors": [
+ "rgba(245, 54, 54, 0.9)",
+ "rgba(237, 129, 40, 0.89)",
+ "rgba(50, 172, 45, 0.97)"
+ ],
+ "datasource": "prometheus",
+ "editable": false,
+ "format": "none",
+ "gauge": {
+ "maxValue": 100,
+ "minValue": 0,
+ "show": false,
+ "thresholdLabels": false,
+ "thresholdMarkers": true
+ },
+ "id": 2,
+ "links": [],
+ "mappingType": 1,
+ "mappingTypes": [
+ {
+ "name": "value to text",
+ "value": 1
+ },
+ {
+ "name": "range to text",
+ "value": 2
+ }
+ ],
+ "maxDataPoints": 100,
+ "nullPointMode": "connected",
+ "postfixFontSize": "50%",
+ "prefix": "",
+ "prefixFontSize": "50%",
+ "rangeMaps": [
+ {
+ "from": "null",
+ "text": "N/A",
+ "to": "null"
+ }
+ ],
+ "span": 3,
+ "sparkline": {
+ "fillColor": "rgba(31, 118, 189, 0.18)",
+ "full": false,
+ "lineColor": "rgb(31, 120, 193)",
+ "show": false
+ },
+ "targets": [
+ {
+ "expr": "max(kube_statefulset_metadata_generation{statefulset=\"$statefulset_name\",namespace=\"$statefulset_namespace\"}) without (instance, pod)",
+ "intervalFactor": 2,
+ "refId": "A",
+ "step": 600
+ }
+ ],
+ "title": "Metadata Generation",
+ "type": "singlestat",
+ "valueFontSize": "80%",
+ "valueMaps": [
+ {
+ "op": "=",
+ "text": "N/A",
+ "value": "null"
+ }
+ ],
+ "valueName": "avg"
+ }
+ ],
+ "showTitle": false,
+ "title": "Dashboard Row",
+ "titleSize": "h6"
+ },
+ {
+ "collapse": false,
+ "editable": false,
+ "height": "350px",
+ "panels": [
+ {
+ "aliasColors": {},
+ "bars": false,
+ "dashLength": 10,
+ "dashes": false,
+ "datasource": "prometheus",
+ "editable": false,
+ "error": false,
+ "fill": 1,
+ "grid": {
+ "threshold1Color": "rgba(216, 200, 27, 0.27)",
+ "threshold2Color": "rgba(234, 112, 112, 0.22)"
+ },
+ "id": 1,
+ "isNew": true,
+ "legend": {
+ "alignAsTable": false,
+ "avg": false,
+ "current": false,
+ "hideEmpty": false,
+ "hideZero": false,
+ "max": false,
+ "min": false,
+ "rightSide": false,
+ "show": true,
+ "total": false
+ },
+ "lines": true,
+ "linewidth": 2,
+ "links": [],
+ "nullPointMode": "connected",
+ "percentage": false,
+ "pointradius": 5,
+ "points": false,
+ "renderer": "flot",
+ "seriesOverrides": [],
+ "spaceLength": 10,
+ "span": 12,
+ "stack": false,
+ "steppedLine": false,
+ "targets": [
+ {
+ "expr": "min(kube_statefulset_status_replicas{statefulset=\"$statefulset_name\",namespace=\"$statefulset_namespace\"}) without (instance, pod)",
+ "intervalFactor": 2,
+ "legendFormat": "available",
+ "refId": "B",
+ "step": 30
+ },
+ {
+ "expr": "max(kube_statefulset_replicas{statefulset=\"$statefulset_name\",namespace=\"$statefulset_namespace\"}) without (instance, pod)",
+ "intervalFactor": 2,
+ "legendFormat": "desired",
+ "refId": "E",
+ "step": 30
+ }
+ ],
+ "title": "Replicas",
+ "tooltip": {
+ "msResolution": true,
+ "shared": true,
+ "sort": 0,
+ "value_type": "cumulative"
+ },
+ "type": "graph",
+ "xaxis": {
+ "mode": "time",
+ "show": true,
+ "values": []
+ },
+ "yaxes": [
+ {
+ "format": "none",
+ "label": "",
+ "logBase": 1,
+ "show": true
+ },
+ {
+ "format": "short",
+ "label": "",
+ "logBase": 1,
+ "show": false
+ }
+ ]
+ }
+ ],
+ "showTitle": false,
+ "title": "Dashboard Row",
+ "titleSize": "h6"
+ }
+ ],
+ "schemaVersion": 14,
+ "sharedCrosshair": false,
+ "style": "dark",
+ "tags": [],
+ "templating": {
+ "list": [
+ {
+ "allValue": ".*",
+ "current": {},
+ "datasource": "prometheus",
+ "hide": 0,
+ "includeAll": false,
+ "label": "Namespace",
+ "multi": false,
+ "name": "statefulset_namespace",
+ "options": [],
+ "query": "label_values(kube_statefulset_metadata_generation, namespace)",
+ "refresh": 1,
+ "regex": "",
+ "sort": 0,
+ "tagValuesQuery": null,
+ "tags": [],
+ "tagsQuery": "",
+ "type": "query",
+ "useTags": false
+ },
+ {
+ "allValue": null,
+ "current": {},
+ "datasource": "prometheus",
+ "hide": 0,
+ "includeAll": false,
+ "label": "StatefulSet",
+ "multi": false,
+ "name": "statefulset_name",
+ "options": [],
+ "query": "label_values(kube_statefulset_metadata_generation{namespace=\"$statefulset_namespace\"}, statefulset)",
+ "refresh": 1,
+ "regex": "",
+ "sort": 0,
+ "tagValuesQuery": "",
+ "tags": [],
+ "tagsQuery": "statefulset",
+ "type": "query",
+ "useTags": false
+ }
+ ]
+ },
+ "time": {
+ "from": "now-6h",
+ "to": "now"
+ },
+ "timepicker": {
+ "refresh_intervals": [
+ "5s",
+ "10s",
+ "30s",
+ "1m",
+ "5m",
+ "15m",
+ "30m",
+ "1h",
+ "2h",
+ "1d"
+ ],
+ "time_options": [
+ "5m",
+ "15m",
+ "1h",
+ "6h",
+ "12h",
+ "24h",
+ "2d",
+ "7d",
+ "30d"
+ ]
+ },
+ "timezone": "browser",
+ "title": "StatefulSet",
+ "version": 1
+ }
+---
diff --git a/manifests/grafana/grafana-dashboards.yaml b/manifests/grafana/grafana-dashboards.yaml
index 8242856452e014e38e1b1cb504438e8fd6e3f0dc..ed3fe5fe0e7d178e062aef56cfee4afae3c590be 100644
--- a/manifests/grafana/grafana-dashboards.yaml
+++ b/manifests/grafana/grafana-dashboards.yaml
@@ -1,7498 +1,12 @@
apiVersion: v1
kind: ConfigMap
metadata:
- name: grafana-dashboards-0
+ name: grafana-dashboards
data:
- deployment-dashboard.json: |+
- {
- "dashboard":
- {
- "__inputs": [
- {
- "description": "",
- "label": "prometheus",
- "name": "DS_PROMETHEUS",
- "pluginId": "prometheus",
- "pluginName": "Prometheus",
- "type": "datasource"
- }
- ],
- "annotations": {
- "list": []
- },
- "editable": false,
- "graphTooltip": 1,
- "hideControls": false,
- "links": [],
- "rows": [
- {
- "collapse": false,
- "editable": false,
- "height": "200px",
- "panels": [
- {
- "colorBackground": false,
- "colorValue": false,
- "colors": [
- "rgba(245, 54, 54, 0.9)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(50, 172, 45, 0.97)"
- ],
- "datasource": "${DS_PROMETHEUS}",
- "editable": false,
- "format": "none",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": false,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "id": 8,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "postfix": "cores",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 4,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": true
- },
- "targets": [
- {
- "expr": "sum(rate(container_cpu_usage_seconds_total{namespace=\"$deployment_namespace\",pod_name=~\"$deployment_name.*\"}[3m]))",
- "intervalFactor": 2,
- "refId": "A",
- "step": 600
- }
- ],
- "title": "CPU",
- "type": "singlestat",
- "valueFontSize": "110%",
- "valueMaps": [
- {
- "op": "=",
- "text": "N/A",
- "value": "null"
- }
- ],
- "valueName": "avg"
- },
- {
- "colorBackground": false,
- "colorValue": false,
- "colors": [
- "rgba(245, 54, 54, 0.9)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(50, 172, 45, 0.97)"
- ],
- "datasource": "${DS_PROMETHEUS}",
- "editable": false,
- "format": "none",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": false,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "id": 9,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "postfix": "GB",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "80%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 4,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": true
- },
- "targets": [
- {
- "expr": "sum(container_memory_usage_bytes{namespace=\"$deployment_namespace\",pod_name=~\"$deployment_name.*\"}) / 1024^3",
- "intervalFactor": 2,
- "refId": "A",
- "step": 600
- }
- ],
- "title": "Memory",
- "type": "singlestat",
- "valueFontSize": "110%",
- "valueMaps": [
- {
- "op": "=",
- "text": "N/A",
- "value": "null"
- }
- ],
- "valueName": "avg"
- },
- {
- "colorBackground": false,
- "colorValue": false,
- "colors": [
- "rgba(245, 54, 54, 0.9)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(50, 172, 45, 0.97)"
- ],
- "datasource": "${DS_PROMETHEUS}",
- "editable": false,
- "format": "Bps",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": false,
- "thresholdLabels": false,
- "thresholdMarkers": false
- },
- "id": 7,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "postfix": "",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 4,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": true
- },
- "targets": [
- {
- "expr": "sum(rate(container_network_transmit_bytes_total{namespace=\"$deployment_namespace\",pod_name=~\"$deployment_name.*\"}[3m])) + sum(rate(container_network_receive_bytes_total{namespace=\"$deployment_namespace\",pod_name=~\"$deployment_name.*\"}[3m]))",
- "intervalFactor": 2,
- "refId": "A",
- "step": 600
- }
- ],
- "title": "Network",
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "N/A",
- "value": "null"
- }
- ],
- "valueName": "avg"
- }
- ],
- "showTitle": false,
- "title": "Dashboard Row",
- "titleSize": "h6"
- },
- {
- "collapse": false,
- "editable": false,
- "height": "100px",
- "panels": [
- {
- "colorBackground": false,
- "colorValue": false,
- "colors": [
- "rgba(245, 54, 54, 0.9)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(50, 172, 45, 0.97)"
- ],
- "datasource": "${DS_PROMETHEUS}",
- "editable": false,
- "format": "none",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": false,
- "thresholdLabels": false,
- "thresholdMarkers": false
- },
- "id": 5,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 3,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false
- },
- "targets": [
- {
- "expr": "max(kube_deployment_spec_replicas{deployment=\"$deployment_name\",namespace=\"$deployment_namespace\"}) without (instance, pod)",
- "intervalFactor": 2,
- "metric": "kube_deployment_spec_replicas",
- "refId": "A",
- "step": 600
- }
- ],
- "title": "Desired Replicas",
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "N/A",
- "value": "null"
- }
- ],
- "valueName": "avg"
- },
- {
- "colorBackground": false,
- "colorValue": false,
- "colors": [
- "rgba(245, 54, 54, 0.9)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(50, 172, 45, 0.97)"
- ],
- "datasource": "${DS_PROMETHEUS}",
- "editable": false,
- "format": "none",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": false,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "id": 6,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 3,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false
- },
- "targets": [
- {
- "expr": "min(kube_deployment_status_replicas_available{deployment=\"$deployment_name\",namespace=\"$deployment_namespace\"}) without (instance, pod)",
- "intervalFactor": 2,
- "refId": "A",
- "step": 600
- }
- ],
- "title": "Available Replicas",
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "N/A",
- "value": "null"
- }
- ],
- "valueName": "avg"
- },
- {
- "colorBackground": false,
- "colorValue": false,
- "colors": [
- "rgba(245, 54, 54, 0.9)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(50, 172, 45, 0.97)"
- ],
- "datasource": "${DS_PROMETHEUS}",
- "editable": false,
- "format": "none",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": false,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "id": 3,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 3,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false
- },
- "targets": [
- {
- "expr": "max(kube_deployment_status_observed_generation{deployment=\"$deployment_name\",namespace=\"$deployment_namespace\"}) without (instance, pod)",
- "intervalFactor": 2,
- "refId": "A",
- "step": 600
- }
- ],
- "title": "Observed Generation",
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "N/A",
- "value": "null"
- }
- ],
- "valueName": "avg"
- },
- {
- "colorBackground": false,
- "colorValue": false,
- "colors": [
- "rgba(245, 54, 54, 0.9)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(50, 172, 45, 0.97)"
- ],
- "datasource": "${DS_PROMETHEUS}",
- "editable": false,
- "format": "none",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": false,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "id": 2,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 3,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false
- },
- "targets": [
- {
- "expr": "max(kube_deployment_metadata_generation{deployment=\"$deployment_name\",namespace=\"$deployment_namespace\"}) without (instance, pod)",
- "intervalFactor": 2,
- "refId": "A",
- "step": 600
- }
- ],
- "title": "Metadata Generation",
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "N/A",
- "value": "null"
- }
- ],
- "valueName": "avg"
- }
- ],
- "showTitle": false,
- "title": "Dashboard Row",
- "titleSize": "h6"
- },
- {
- "collapse": false,
- "editable": false,
- "height": "350px",
- "panels": [
- {
- "aliasColors": {},
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": "${DS_PROMETHEUS}",
- "editable": false,
- "error": false,
- "fill": 1,
- "grid": {
- "threshold1Color": "rgba(216, 200, 27, 0.27)",
- "threshold2Color": "rgba(234, 112, 112, 0.22)"
- },
- "id": 1,
- "isNew": true,
- "legend": {
- "alignAsTable": false,
- "avg": false,
- "current": false,
- "hideEmpty": false,
- "hideZero": false,
- "max": false,
- "min": false,
- "rightSide": false,
- "show": true,
- "total": false
- },
- "lines": true,
- "linewidth": 2,
- "links": [],
- "nullPointMode": "connected",
- "percentage": false,
- "pointradius": 5,
- "points": false,
- "renderer": "flot",
- "seriesOverrides": [],
- "spaceLength": 10,
- "span": 12,
- "stack": false,
- "steppedLine": false,
- "targets": [
- {
- "expr": "max(kube_deployment_status_replicas{deployment=\"$deployment_name\",namespace=\"$deployment_namespace\"}) without (instance, pod)",
- "intervalFactor": 2,
- "legendFormat": "current replicas",
- "refId": "A",
- "step": 30
- },
- {
- "expr": "min(kube_deployment_status_replicas_available{deployment=\"$deployment_name\",namespace=\"$deployment_namespace\"}) without (instance, pod)",
- "intervalFactor": 2,
- "legendFormat": "available",
- "refId": "B",
- "step": 30
- },
- {
- "expr": "max(kube_deployment_status_replicas_unavailable{deployment=\"$deployment_name\",namespace=\"$deployment_namespace\"}) without (instance, pod)",
- "intervalFactor": 2,
- "legendFormat": "unavailable",
- "refId": "C",
- "step": 30
- },
- {
- "expr": "min(kube_deployment_status_replicas_updated{deployment=\"$deployment_name\",namespace=\"$deployment_namespace\"}) without (instance, pod)",
- "intervalFactor": 2,
- "legendFormat": "updated",
- "refId": "D",
- "step": 30
- },
- {
- "expr": "max(kube_deployment_spec_replicas{deployment=\"$deployment_name\",namespace=\"$deployment_namespace\"}) without (instance, pod)",
- "intervalFactor": 2,
- "legendFormat": "desired",
- "refId": "E",
- "step": 30
- }
- ],
- "title": "Replicas",
- "tooltip": {
- "msResolution": true,
- "shared": true,
- "sort": 0,
- "value_type": "cumulative"
- },
- "type": "graph",
- "xaxis": {
- "mode": "time",
- "show": true,
- "values": []
- },
- "yaxes": [
- {
- "format": "none",
- "label": "",
- "logBase": 1,
- "show": true
- },
- {
- "format": "short",
- "label": "",
- "logBase": 1,
- "show": false
- }
- ]
- }
- ],
- "showTitle": false,
- "title": "Dashboard Row",
- "titleSize": "h6"
- }
- ],
- "schemaVersion": 14,
- "sharedCrosshair": false,
- "style": "dark",
- "tags": [],
- "templating": {
- "list": [
- {
- "allValue": ".*",
- "current": {},
- "datasource": "${DS_PROMETHEUS}",
- "hide": 0,
- "includeAll": false,
- "label": "Namespace",
- "multi": false,
- "name": "deployment_namespace",
- "options": [],
- "query": "label_values(kube_deployment_metadata_generation, namespace)",
- "refresh": 1,
- "regex": "",
- "sort": 0,
- "tagValuesQuery": null,
- "tags": [],
- "tagsQuery": "",
- "type": "query",
- "useTags": false
- },
- {
- "allValue": null,
- "current": {},
- "datasource": "${DS_PROMETHEUS}",
- "hide": 0,
- "includeAll": false,
- "label": "Deployment",
- "multi": false,
- "name": "deployment_name",
- "options": [],
- "query": "label_values(kube_deployment_metadata_generation{namespace=\"$deployment_namespace\"}, deployment)",
- "refresh": 1,
- "regex": "",
- "sort": 0,
- "tagValuesQuery": "",
- "tags": [],
- "tagsQuery": "deployment",
- "type": "query",
- "useTags": false
- }
- ]
- },
- "time": {
- "from": "now-6h",
- "to": "now"
- },
- "timepicker": {
- "refresh_intervals": [
- "5s",
- "10s",
- "30s",
- "1m",
- "5m",
- "15m",
- "30m",
- "1h",
- "2h",
- "1d"
- ],
- "time_options": [
- "5m",
- "15m",
- "1h",
- "6h",
- "12h",
- "24h",
- "2d",
- "7d",
- "30d"
- ]
- },
- "timezone": "browser",
- "title": "Deployment",
- "version": 1
- }
- ,
- "inputs": [
- {
- "name": "DS_PROMETHEUS",
- "pluginId": "prometheus",
- "type": "datasource",
- "value": "prometheus"
- }
- ],
- "overwrite": true
- }
- etcd-dashboard.json: |+
- {
- "dashboard":
- {
- "__inputs": [
- {
- "name": "DS_PROMETHEUS",
- "label": "prometheus",
- "description": "",
- "type": "datasource",
- "pluginId": "prometheus",
- "pluginName": "Prometheus"
- }
- ],
- "__requires": [
- {
- "type": "grafana",
- "id": "grafana",
- "name": "Grafana",
- "version": "4.5.2"
- },
- {
- "type": "panel",
- "id": "graph",
- "name": "Graph",
- "version": ""
- },
- {
- "type": "datasource",
- "id": "prometheus",
- "name": "Prometheus",
- "version": "1.0.0"
- },
- {
- "type": "panel",
- "id": "singlestat",
- "name": "Singlestat",
- "version": ""
- }
- ],
- "annotations": {
- "list": []
- },
- "description": "etcd sample Grafana dashboard with Prometheus",
- "editable": false,
- "gnetId": null,
- "graphTooltip": 0,
- "hideControls": false,
- "id": null,
- "links": [],
- "refresh": false,
- "rows": [
- {
- "collapse": false,
- "height": "250px",
- "panels": [
- {
- "cacheTimeout": null,
- "colorBackground": false,
- "colorValue": false,
- "colors": [
- "rgba(245, 54, 54, 0.9)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(50, 172, 45, 0.97)"
- ],
- "datasource": "${DS_PROMETHEUS}",
- "editable": false,
- "error": false,
- "format": "none",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": false,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "id": 28,
- "interval": null,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "nullText": null,
- "postfix": "",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 3,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false
- },
- "tableColumn": "",
- "targets": [
- {
- "expr": "sum(etcd_server_has_leader)",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "",
- "metric": "etcd_server_has_leader",
- "refId": "A",
- "step": 20
- }
- ],
- "thresholds": "",
- "title": "Up",
- "type": "singlestat",
- "valueFontSize": "200%",
- "valueMaps": [
- {
- "op": "=",
- "text": "N/A",
- "value": "null"
- }
- ],
- "valueName": "avg"
- },
- {
- "aliasColors": {},
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": "${DS_PROMETHEUS}",
- "editable": false,
- "error": false,
- "fill": 0,
- "id": 23,
- "legend": {
- "avg": false,
- "current": false,
- "max": false,
- "min": false,
- "show": false,
- "total": false,
- "values": false
- },
- "lines": true,
- "linewidth": 2,
- "links": [],
- "nullPointMode": "connected",
- "percentage": false,
- "pointradius": 5,
- "points": false,
- "renderer": "flot",
- "seriesOverrides": [],
- "spaceLength": 10,
- "span": 5,
- "stack": false,
- "steppedLine": false,
- "targets": [
- {
- "expr": "sum(rate(grpc_server_started_total{grpc_type=\"unary\"}[5m]))",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "RPC Rate",
- "metric": "grpc_server_started_total",
- "refId": "A",
- "step": 4
- },
- {
- "expr": "sum(rate(grpc_server_handled_total{grpc_type=\"unary\",grpc_code!=\"OK\"}[5m]))",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "RPC Failed Rate",
- "metric": "grpc_server_handled_total",
- "refId": "B",
- "step": 4
- }
- ],
- "thresholds": [],
- "timeFrom": null,
- "timeShift": null,
- "title": "RPC Rate",
- "tooltip": {
- "msResolution": false,
- "shared": true,
- "sort": 0,
- "value_type": "individual"
- },
- "type": "graph",
- "xaxis": {
- "buckets": null,
- "mode": "time",
- "name": null,
- "show": true,
- "values": []
- },
- "yaxes": [
- {
- "format": "ops",
- "label": null,
- "logBase": 1,
- "max": null,
- "min": null,
- "show": true
- },
- {
- "format": "short",
- "label": null,
- "logBase": 1,
- "max": null,
- "min": null,
- "show": true
- }
- ]
- },
- {
- "aliasColors": {},
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": "${DS_PROMETHEUS}",
- "editable": false,
- "error": false,
- "fill": 0,
- "id": 41,
- "legend": {
- "avg": false,
- "current": false,
- "max": false,
- "min": false,
- "show": false,
- "total": false,
- "values": false
- },
- "lines": true,
- "linewidth": 2,
- "links": [],
- "nullPointMode": "connected",
- "percentage": false,
- "pointradius": 5,
- "points": false,
- "renderer": "flot",
- "seriesOverrides": [],
- "spaceLength": 10,
- "span": 4,
- "stack": true,
- "steppedLine": false,
- "targets": [
- {
- "expr": "sum(grpc_server_started_total{grpc_service=\"etcdserverpb.Watch\",grpc_type=\"bidi_stream\"}) - sum(grpc_server_handled_total{grpc_service=\"etcdserverpb.Watch\",grpc_type=\"bidi_stream\"})",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "Watch Streams",
- "metric": "grpc_server_handled_total",
- "refId": "A",
- "step": 4
- },
- {
- "expr": "sum(grpc_server_started_total{grpc_service=\"etcdserverpb.Lease\",grpc_type=\"bidi_stream\"}) - sum(grpc_server_handled_total{grpc_service=\"etcdserverpb.Lease\",grpc_type=\"bidi_stream\"})",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "Lease Streams",
- "metric": "grpc_server_handled_total",
- "refId": "B",
- "step": 4
- }
- ],
- "thresholds": [],
- "timeFrom": null,
- "timeShift": null,
- "title": "Active Streams",
- "tooltip": {
- "msResolution": false,
- "shared": true,
- "sort": 0,
- "value_type": "individual"
- },
- "type": "graph",
- "xaxis": {
- "buckets": null,
- "mode": "time",
- "name": null,
- "show": true,
- "values": []
- },
- "yaxes": [
- {
- "format": "short",
- "label": "",
- "logBase": 1,
- "max": null,
- "min": null,
- "show": true
- },
- {
- "format": "short",
- "label": null,
- "logBase": 1,
- "max": null,
- "min": null,
- "show": true
- }
- ]
- }
- ],
- "repeat": null,
- "repeatIteration": null,
- "repeatRowId": null,
- "showTitle": false,
- "title": "Row",
- "titleSize": "h6"
- },
- {
- "collapse": false,
- "height": "250px",
- "panels": [
- {
- "aliasColors": {},
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": "${DS_PROMETHEUS}",
- "decimals": null,
- "editable": false,
- "error": false,
- "fill": 0,
- "grid": {},
- "id": 1,
- "legend": {
- "avg": false,
- "current": false,
- "max": false,
- "min": false,
- "show": false,
- "total": false,
- "values": false
- },
- "lines": true,
- "linewidth": 2,
- "links": [],
- "nullPointMode": "connected",
- "percentage": false,
- "pointradius": 5,
- "points": false,
- "renderer": "flot",
- "seriesOverrides": [],
- "spaceLength": 10,
- "span": 4,
- "stack": false,
- "steppedLine": false,
- "targets": [
- {
- "expr": "etcd_debugging_mvcc_db_total_size_in_bytes",
- "format": "time_series",
- "hide": false,
- "interval": "",
- "intervalFactor": 2,
- "legendFormat": "{{instance}} DB Size",
- "metric": "",
- "refId": "A",
- "step": 4
- }
- ],
- "thresholds": [],
- "timeFrom": null,
- "timeShift": null,
- "title": "DB Size",
- "tooltip": {
- "msResolution": false,
- "shared": true,
- "sort": 0,
- "value_type": "cumulative"
- },
- "type": "graph",
- "xaxis": {
- "buckets": null,
- "mode": "time",
- "name": null,
- "show": true,
- "values": []
- },
- "yaxes": [
- {
- "format": "bytes",
- "logBase": 1,
- "max": null,
- "min": null,
- "show": true
- },
- {
- "format": "short",
- "logBase": 1,
- "max": null,
- "min": null,
- "show": false
- }
- ]
- },
- {
- "aliasColors": {},
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": "${DS_PROMETHEUS}",
- "editable": false,
- "error": false,
- "fill": 0,
- "grid": {},
- "id": 3,
- "legend": {
- "avg": false,
- "current": false,
- "max": false,
- "min": false,
- "show": false,
- "total": false,
- "values": false
- },
- "lines": true,
- "linewidth": 2,
- "links": [],
- "nullPointMode": "connected",
- "percentage": false,
- "pointradius": 1,
- "points": false,
- "renderer": "flot",
- "seriesOverrides": [],
- "spaceLength": 10,
- "span": 4,
- "stack": false,
- "steppedLine": true,
- "targets": [
- {
- "expr": "histogram_quantile(0.99, sum(rate(etcd_disk_wal_fsync_duration_seconds_bucket[5m])) by (instance, le))",
- "format": "time_series",
- "hide": false,
- "intervalFactor": 2,
- "legendFormat": "{{instance}} WAL fsync",
- "metric": "etcd_disk_wal_fsync_duration_seconds_bucket",
- "refId": "A",
- "step": 4
- },
- {
- "expr": "histogram_quantile(0.99, sum(rate(etcd_disk_backend_commit_duration_seconds_bucket[5m])) by (instance, le))",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "{{instance}} DB fsync",
- "metric": "etcd_disk_backend_commit_duration_seconds_bucket",
- "refId": "B",
- "step": 4
- }
- ],
- "thresholds": [],
- "timeFrom": null,
- "timeShift": null,
- "title": "Disk Sync Duration",
- "tooltip": {
- "msResolution": false,
- "shared": true,
- "sort": 0,
- "value_type": "cumulative"
- },
- "type": "graph",
- "xaxis": {
- "buckets": null,
- "mode": "time",
- "name": null,
- "show": true,
- "values": []
- },
- "yaxes": [
- {
- "format": "s",
- "logBase": 1,
- "max": null,
- "min": null,
- "show": true
- },
- {
- "format": "short",
- "logBase": 1,
- "max": null,
- "min": null,
- "show": false
- }
- ]
- },
- {
- "aliasColors": {},
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": "${DS_PROMETHEUS}",
- "editable": false,
- "error": false,
- "fill": 0,
- "id": 29,
- "legend": {
- "avg": false,
- "current": false,
- "max": false,
- "min": false,
- "show": false,
- "total": false,
- "values": false
- },
- "lines": true,
- "linewidth": 2,
- "links": [],
- "nullPointMode": "connected",
- "percentage": false,
- "pointradius": 5,
- "points": false,
- "renderer": "flot",
- "seriesOverrides": [],
- "spaceLength": 10,
- "span": 4,
- "stack": false,
- "steppedLine": false,
- "targets": [
- {
- "expr": "process_resident_memory_bytes",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "{{instance}} Resident Memory",
- "metric": "process_resident_memory_bytes",
- "refId": "A",
- "step": 4
- }
- ],
- "thresholds": [],
- "timeFrom": null,
- "timeShift": null,
- "title": "Memory",
- "tooltip": {
- "msResolution": false,
- "shared": true,
- "sort": 0,
- "value_type": "individual"
- },
- "type": "graph",
- "xaxis": {
- "buckets": null,
- "mode": "time",
- "name": null,
- "show": true,
- "values": []
- },
- "yaxes": [
- {
- "format": "bytes",
- "label": null,
- "logBase": 1,
- "max": null,
- "min": null,
- "show": true
- },
- {
- "format": "short",
- "label": null,
- "logBase": 1,
- "max": null,
- "min": null,
- "show": true
- }
- ]
- }
- ],
- "repeat": null,
- "repeatIteration": null,
- "repeatRowId": null,
- "showTitle": false,
- "title": "New row",
- "titleSize": "h6"
- },
- {
- "collapse": false,
- "height": "250px",
- "panels": [
- {
- "aliasColors": {},
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": "${DS_PROMETHEUS}",
- "editable": false,
- "error": false,
- "fill": 5,
- "id": 22,
- "legend": {
- "avg": false,
- "current": false,
- "max": false,
- "min": false,
- "show": false,
- "total": false,
- "values": false
- },
- "lines": true,
- "linewidth": 2,
- "links": [],
- "nullPointMode": "connected",
- "percentage": false,
- "pointradius": 5,
- "points": false,
- "renderer": "flot",
- "seriesOverrides": [],
- "spaceLength": 10,
- "span": 3,
- "stack": true,
- "steppedLine": false,
- "targets": [
- {
- "expr": "rate(etcd_network_client_grpc_received_bytes_total[5m])",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "{{instance}} Client Traffic In",
- "metric": "etcd_network_client_grpc_received_bytes_total",
- "refId": "A",
- "step": 4
- }
- ],
- "thresholds": [],
- "timeFrom": null,
- "timeShift": null,
- "title": "Client Traffic In",
- "tooltip": {
- "msResolution": false,
- "shared": true,
- "sort": 0,
- "value_type": "individual"
- },
- "type": "graph",
- "xaxis": {
- "buckets": null,
- "mode": "time",
- "name": null,
- "show": true,
- "values": []
- },
- "yaxes": [
- {
- "format": "Bps",
- "label": null,
- "logBase": 1,
- "max": null,
- "min": null,
- "show": true
- },
- {
- "format": "short",
- "label": null,
- "logBase": 1,
- "max": null,
- "min": null,
- "show": true
- }
- ]
- },
- {
- "aliasColors": {},
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": "${DS_PROMETHEUS}",
- "editable": false,
- "error": false,
- "fill": 5,
- "id": 21,
- "legend": {
- "avg": false,
- "current": false,
- "max": false,
- "min": false,
- "show": false,
- "total": false,
- "values": false
- },
- "lines": true,
- "linewidth": 2,
- "links": [],
- "nullPointMode": "connected",
- "percentage": false,
- "pointradius": 5,
- "points": false,
- "renderer": "flot",
- "seriesOverrides": [],
- "spaceLength": 10,
- "span": 3,
- "stack": true,
- "steppedLine": false,
- "targets": [
- {
- "expr": "rate(etcd_network_client_grpc_sent_bytes_total[5m])",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "{{instance}} Client Traffic Out",
- "metric": "etcd_network_client_grpc_sent_bytes_total",
- "refId": "A",
- "step": 4
- }
- ],
- "thresholds": [],
- "timeFrom": null,
- "timeShift": null,
- "title": "Client Traffic Out",
- "tooltip": {
- "msResolution": false,
- "shared": true,
- "sort": 0,
- "value_type": "individual"
- },
- "type": "graph",
- "xaxis": {
- "buckets": null,
- "mode": "time",
- "name": null,
- "show": true,
- "values": []
- },
- "yaxes": [
- {
- "format": "Bps",
- "label": null,
- "logBase": 1,
- "max": null,
- "min": null,
- "show": true
- },
- {
- "format": "short",
- "label": null,
- "logBase": 1,
- "max": null,
- "min": null,
- "show": true
- }
- ]
- },
- {
- "aliasColors": {},
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": "${DS_PROMETHEUS}",
- "editable": false,
- "error": false,
- "fill": 0,
- "id": 20,
- "legend": {
- "avg": false,
- "current": false,
- "max": false,
- "min": false,
- "show": false,
- "total": false,
- "values": false
- },
- "lines": true,
- "linewidth": 2,
- "links": [],
- "nullPointMode": "connected",
- "percentage": false,
- "pointradius": 5,
- "points": false,
- "renderer": "flot",
- "seriesOverrides": [],
- "spaceLength": 10,
- "span": 3,
- "stack": false,
- "steppedLine": false,
- "targets": [
- {
- "expr": "sum(rate(etcd_network_peer_received_bytes_total[5m])) by (instance)",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "{{instance}} Peer Traffic In",
- "metric": "etcd_network_peer_received_bytes_total",
- "refId": "A",
- "step": 4
- }
- ],
- "thresholds": [],
- "timeFrom": null,
- "timeShift": null,
- "title": "Peer Traffic In",
- "tooltip": {
- "msResolution": false,
- "shared": true,
- "sort": 0,
- "value_type": "individual"
- },
- "type": "graph",
- "xaxis": {
- "buckets": null,
- "mode": "time",
- "name": null,
- "show": true,
- "values": []
- },
- "yaxes": [
- {
- "format": "Bps",
- "label": null,
- "logBase": 1,
- "max": null,
- "min": null,
- "show": true
- },
- {
- "format": "short",
- "label": null,
- "logBase": 1,
- "max": null,
- "min": null,
- "show": true
- }
- ]
- },
- {
- "aliasColors": {},
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": "${DS_PROMETHEUS}",
- "decimals": null,
- "editable": false,
- "error": false,
- "fill": 0,
- "grid": {},
- "id": 16,
- "legend": {
- "avg": false,
- "current": false,
- "max": false,
- "min": false,
- "show": false,
- "total": false,
- "values": false
- },
- "lines": true,
- "linewidth": 2,
- "links": [],
- "nullPointMode": "connected",
- "percentage": false,
- "pointradius": 5,
- "points": false,
- "renderer": "flot",
- "seriesOverrides": [],
- "spaceLength": 10,
- "span": 3,
- "stack": false,
- "steppedLine": false,
- "targets": [
- {
- "expr": "sum(rate(etcd_network_peer_sent_bytes_total[5m])) by (instance)",
- "format": "time_series",
- "hide": false,
- "interval": "",
- "intervalFactor": 2,
- "legendFormat": "{{instance}} Peer Traffic Out",
- "metric": "etcd_network_peer_sent_bytes_total",
- "refId": "A",
- "step": 4
- }
- ],
- "thresholds": [],
- "timeFrom": null,
- "timeShift": null,
- "title": "Peer Traffic Out",
- "tooltip": {
- "msResolution": false,
- "shared": true,
- "sort": 0,
- "value_type": "cumulative"
- },
- "type": "graph",
- "xaxis": {
- "buckets": null,
- "mode": "time",
- "name": null,
- "show": true,
- "values": []
- },
- "yaxes": [
- {
- "format": "Bps",
- "logBase": 1,
- "max": null,
- "min": null,
- "show": true
- },
- {
- "format": "short",
- "logBase": 1,
- "max": null,
- "min": null,
- "show": true
- }
- ]
- }
- ],
- "repeat": null,
- "repeatIteration": null,
- "repeatRowId": null,
- "showTitle": false,
- "title": "New row",
- "titleSize": "h6"
- },
- {
- "collapse": false,
- "height": "250px",
- "panels": [
- {
- "aliasColors": {},
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": "${DS_PROMETHEUS}",
- "editable": false,
- "error": false,
- "fill": 0,
- "id": 40,
- "legend": {
- "avg": false,
- "current": false,
- "max": false,
- "min": false,
- "show": false,
- "total": false,
- "values": false
- },
- "lines": true,
- "linewidth": 2,
- "links": [],
- "nullPointMode": "connected",
- "percentage": false,
- "pointradius": 5,
- "points": false,
- "renderer": "flot",
- "seriesOverrides": [],
- "spaceLength": 10,
- "span": 6,
- "stack": false,
- "steppedLine": false,
- "targets": [
- {
- "expr": "sum(rate(etcd_server_proposals_failed_total[5m]))",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "Proposal Failure Rate",
- "metric": "etcd_server_proposals_failed_total",
- "refId": "A",
- "step": 2
- },
- {
- "expr": "sum(etcd_server_proposals_pending)",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "Proposal Pending Total",
- "metric": "etcd_server_proposals_pending",
- "refId": "B",
- "step": 2
- },
- {
- "expr": "sum(rate(etcd_server_proposals_committed_total[5m]))",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "Proposal Commit Rate",
- "metric": "etcd_server_proposals_committed_total",
- "refId": "C",
- "step": 2
- },
- {
- "expr": "sum(rate(etcd_server_proposals_applied_total[5m]))",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "Proposal Apply Rate",
- "refId": "D",
- "step": 2
- }
- ],
- "thresholds": [],
- "timeFrom": null,
- "timeShift": null,
- "title": "Raft Proposals",
- "tooltip": {
- "msResolution": false,
- "shared": true,
- "sort": 0,
- "value_type": "individual"
- },
- "type": "graph",
- "xaxis": {
- "buckets": null,
- "mode": "time",
- "name": null,
- "show": true,
- "values": []
- },
- "yaxes": [
- {
- "format": "short",
- "label": "",
- "logBase": 1,
- "max": null,
- "min": null,
- "show": true
- },
- {
- "format": "short",
- "label": null,
- "logBase": 1,
- "max": null,
- "min": null,
- "show": true
- }
- ]
- },
- {
- "aliasColors": {},
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": "${DS_PROMETHEUS}",
- "decimals": 0,
- "editable": false,
- "error": false,
- "fill": 0,
- "id": 19,
- "legend": {
- "alignAsTable": false,
- "avg": false,
- "current": false,
- "max": false,
- "min": false,
- "rightSide": false,
- "show": false,
- "total": false,
- "values": false
- },
- "lines": true,
- "linewidth": 2,
- "links": [],
- "nullPointMode": "connected",
- "percentage": false,
- "pointradius": 5,
- "points": false,
- "renderer": "flot",
- "seriesOverrides": [],
- "spaceLength": 10,
- "span": 6,
- "stack": false,
- "steppedLine": false,
- "targets": [
- {
- "expr": "changes(etcd_server_leader_changes_seen_total[1d])",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "{{instance}} Total Leader Elections Per Day",
- "metric": "etcd_server_leader_changes_seen_total",
- "refId": "A",
- "step": 2
- }
- ],
- "thresholds": [],
- "timeFrom": null,
- "timeShift": null,
- "title": "Total Leader Elections Per Day",
- "tooltip": {
- "msResolution": false,
- "shared": true,
- "sort": 0,
- "value_type": "individual"
- },
- "type": "graph",
- "xaxis": {
- "buckets": null,
- "mode": "time",
- "name": null,
- "show": true,
- "values": []
- },
- "yaxes": [
- {
- "format": "short",
- "label": null,
- "logBase": 1,
- "max": null,
- "min": null,
- "show": true
- },
- {
- "format": "short",
- "label": null,
- "logBase": 1,
- "max": null,
- "min": null,
- "show": true
- }
- ]
- }
- ],
- "repeat": null,
- "repeatIteration": null,
- "repeatRowId": null,
- "showTitle": false,
- "title": "New row",
- "titleSize": "h6"
- }
- ],
- "schemaVersion": 14,
- "style": "dark",
- "tags": [],
- "templating": {
- "list": []
- },
- "time": {
- "from": "now-15m",
- "to": "now"
- },
- "timepicker": {
- "now": true,
- "refresh_intervals": [
- "5s",
- "10s",
- "30s",
- "1m",
- "5m",
- "15m",
- "30m",
- "1h",
- "2h",
- "1d"
- ],
- "time_options": [
- "5m",
- "15m",
- "1h",
- "6h",
- "12h",
- "24h",
- "2d",
- "7d",
- "30d"
- ]
- },
- "timezone": "browser",
- "title": "etcd",
- "version": 4
- }
- ,
- "inputs": [
- {
- "name": "DS_PROMETHEUS",
- "pluginId": "prometheus",
- "type": "datasource",
- "value": "prometheus"
- }
- ],
- "overwrite": true
- }
- kubernetes-capacity-planning-dashboard.json: |+
- {
- "dashboard":
- {
- "__inputs": [
- {
- "description": "",
- "label": "prometheus",
- "name": "DS_PROMETHEUS",
- "pluginId": "prometheus",
- "pluginName": "Prometheus",
- "type": "datasource"
- }
- ],
- "annotations": {
- "list": []
- },
- "editable": false,
- "gnetId": 22,
- "graphTooltip": 0,
- "hideControls": false,
- "links": [],
- "refresh": false,
- "rows": [
- {
- "collapse": false,
- "editable": false,
- "height": "250px",
- "panels": [
- {
- "aliasColors": {},
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": "${DS_PROMETHEUS}",
- "editable": false,
- "error": false,
- "fill": 1,
- "grid": {
- "threshold1Color": "rgba(216, 200, 27, 0.27)",
- "threshold2Color": "rgba(234, 112, 112, 0.22)"
- },
- "id": 3,
- "isNew": false,
- "legend": {
- "alignAsTable": false,
- "avg": false,
- "current": false,
- "hideEmpty": false,
- "hideZero": false,
- "max": false,
- "min": false,
- "rightSide": false,
- "show": true,
- "total": false
- },
- "lines": true,
- "linewidth": 2,
- "links": [],
- "nullPointMode": "connected",
- "percentage": false,
- "pointradius": 5,
- "points": false,
- "renderer": "flot",
- "seriesOverrides": [],
- "spaceLength": 10,
- "span": 6,
- "stack": false,
- "steppedLine": false,
- "targets": [
- {
- "expr": "sum(rate(node_cpu{mode=\"idle\"}[2m])) * 100",
- "hide": false,
- "intervalFactor": 10,
- "legendFormat": "",
- "refId": "A",
- "step": 50
- }
- ],
- "title": "Idle CPU",
- "tooltip": {
- "msResolution": false,
- "shared": true,
- "sort": 0,
- "value_type": "cumulative"
- },
- "type": "graph",
- "xaxis": {
- "mode": "time",
- "show": true,
- "values": []
- },
- "yaxes": [
- {
- "format": "percent",
- "label": "cpu usage",
- "logBase": 1,
- "min": 0,
- "show": true
- },
- {
- "format": "short",
- "logBase": 1,
- "show": true
- }
- ]
- },
- {
- "aliasColors": {},
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": "${DS_PROMETHEUS}",
- "editable": false,
- "error": false,
- "fill": 1,
- "grid": {
- "threshold1Color": "rgba(216, 200, 27, 0.27)",
- "threshold2Color": "rgba(234, 112, 112, 0.22)"
- },
- "id": 9,
- "isNew": false,
- "legend": {
- "alignAsTable": false,
- "avg": false,
- "current": false,
- "hideEmpty": false,
- "hideZero": false,
- "max": false,
- "min": false,
- "rightSide": false,
- "show": true,
- "total": false
- },
- "lines": true,
- "linewidth": 2,
- "links": [],
- "nullPointMode": "connected",
- "percentage": false,
- "pointradius": 5,
- "points": false,
- "renderer": "flot",
- "seriesOverrides": [],
- "spaceLength": 10,
- "span": 6,
- "stack": false,
- "steppedLine": false,
- "targets": [
- {
- "expr": "sum(node_load1)",
- "intervalFactor": 4,
- "legendFormat": "load 1m",
- "refId": "A",
- "step": 20,
- "target": ""
- },
- {
- "expr": "sum(node_load5)",
- "intervalFactor": 4,
- "legendFormat": "load 5m",
- "refId": "B",
- "step": 20,
- "target": ""
- },
- {
- "expr": "sum(node_load15)",
- "intervalFactor": 4,
- "legendFormat": "load 15m",
- "refId": "C",
- "step": 20,
- "target": ""
- }
- ],
- "title": "System Load",
- "tooltip": {
- "msResolution": false,
- "shared": true,
- "sort": 0,
- "value_type": "cumulative"
- },
- "type": "graph",
- "xaxis": {
- "mode": "time",
- "show": true,
- "values": []
- },
- "yaxes": [
- {
- "format": "percentunit",
- "logBase": 1,
- "show": true
- },
- {
- "format": "short",
- "logBase": 1,
- "show": true
- }
- ]
- }
- ],
- "showTitle": false,
- "title": "New Row",
- "titleSize": "h6"
- },
- {
- "collapse": false,
- "editable": false,
- "height": "250px",
- "panels": [
- {
- "aliasColors": {},
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": "${DS_PROMETHEUS}",
- "editable": false,
- "error": false,
- "fill": 1,
- "grid": {
- "threshold1Color": "rgba(216, 200, 27, 0.27)",
- "threshold2Color": "rgba(234, 112, 112, 0.22)"
- },
- "id": 4,
- "isNew": false,
- "legend": {
- "alignAsTable": false,
- "avg": false,
- "current": false,
- "hideEmpty": false,
- "hideZero": false,
- "max": false,
- "min": false,
- "rightSide": false,
- "show": true,
- "total": false
- },
- "lines": true,
- "linewidth": 2,
- "links": [],
- "nullPointMode": "connected",
- "percentage": false,
- "pointradius": 5,
- "points": false,
- "renderer": "flot",
- "seriesOverrides": [
- {
- "alias": "node_memory_SwapFree{instance=\"172.17.0.1:9100\",job=\"prometheus\"}",
- "yaxis": 2
- }
- ],
- "spaceLength": 10,
- "span": 9,
- "stack": true,
- "steppedLine": false,
- "targets": [
- {
- "expr": "sum(node_memory_MemTotal) - sum(node_memory_MemFree) - sum(node_memory_Buffers) - sum(node_memory_Cached)",
- "intervalFactor": 2,
- "legendFormat": "memory usage",
- "metric": "memo",
- "refId": "A",
- "step": 10,
- "target": ""
- },
- {
- "expr": "sum(node_memory_Buffers)",
- "interval": "",
- "intervalFactor": 2,
- "legendFormat": "memory buffers",
- "metric": "memo",
- "refId": "B",
- "step": 10,
- "target": ""
- },
- {
- "expr": "sum(node_memory_Cached)",
- "interval": "",
- "intervalFactor": 2,
- "legendFormat": "memory cached",
- "metric": "memo",
- "refId": "C",
- "step": 10,
- "target": ""
- },
- {
- "expr": "sum(node_memory_MemFree)",
- "interval": "",
- "intervalFactor": 2,
- "legendFormat": "memory free",
- "metric": "memo",
- "refId": "D",
- "step": 10,
- "target": ""
- }
- ],
- "title": "Memory Usage",
- "tooltip": {
- "msResolution": false,
- "shared": true,
- "sort": 0,
- "value_type": "individual"
- },
- "type": "graph",
- "xaxis": {
- "mode": "time",
- "show": true,
- "values": []
- },
- "yaxes": [
- {
- "format": "bytes",
- "logBase": 1,
- "min": "0",
- "show": true
- },
- {
- "format": "short",
- "logBase": 1,
- "show": true
- }
- ]
- },
- {
- "colorBackground": false,
- "colorValue": false,
- "colors": [
- "rgba(50, 172, 45, 0.97)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(245, 54, 54, 0.9)"
- ],
- "datasource": "${DS_PROMETHEUS}",
- "editable": false,
- "format": "percent",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": true,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "hideTimeOverride": false,
- "id": 5,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "postfix": "",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 3,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false
- },
- "targets": [
- {
- "expr": "((sum(node_memory_MemTotal) - sum(node_memory_MemFree) - sum(node_memory_Buffers) - sum(node_memory_Cached)) / sum(node_memory_MemTotal)) * 100",
- "intervalFactor": 2,
- "metric": "",
- "refId": "A",
- "step": 60,
- "target": ""
- }
- ],
- "thresholds": "80, 90",
- "title": "Memory Usage",
- "transparent": false,
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "N/A",
- "value": "null"
- }
- ],
- "valueName": "avg"
- }
- ],
- "showTitle": false,
- "title": "New Row",
- "titleSize": "h6"
- },
- {
- "collapse": false,
- "editable": false,
- "height": "246px",
- "panels": [
- {
- "aliasColors": {},
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": "${DS_PROMETHEUS}",
- "editable": false,
- "error": false,
- "fill": 1,
- "grid": {
- "threshold1Color": "rgba(216, 200, 27, 0.27)",
- "threshold2Color": "rgba(234, 112, 112, 0.22)"
- },
- "id": 6,
- "isNew": false,
- "legend": {
- "alignAsTable": false,
- "avg": false,
- "current": false,
- "hideEmpty": false,
- "hideZero": false,
- "max": false,
- "min": false,
- "rightSide": false,
- "show": true,
- "total": false
- },
- "lines": true,
- "linewidth": 2,
- "links": [],
- "nullPointMode": "connected",
- "percentage": false,
- "pointradius": 5,
- "points": false,
- "renderer": "flot",
- "seriesOverrides": [
- {
- "alias": "read",
- "yaxis": 1
- },
- {
- "alias": "{instance=\"172.17.0.1:9100\"}",
- "yaxis": 2
- },
- {
- "alias": "io time",
- "yaxis": 2
- }
- ],
- "spaceLength": 10,
- "span": 9,
- "stack": false,
- "steppedLine": false,
- "targets": [
- {
- "expr": "sum(rate(node_disk_bytes_read[5m]))",
- "hide": false,
- "intervalFactor": 4,
- "legendFormat": "read",
- "refId": "A",
- "step": 20,
- "target": ""
- },
- {
- "expr": "sum(rate(node_disk_bytes_written[5m]))",
- "intervalFactor": 4,
- "legendFormat": "written",
- "refId": "B",
- "step": 20
- },
- {
- "expr": "sum(rate(node_disk_io_time_ms[5m]))",
- "intervalFactor": 4,
- "legendFormat": "io time",
- "refId": "C",
- "step": 20
- }
- ],
- "title": "Disk I/O",
- "tooltip": {
- "msResolution": false,
- "shared": true,
- "sort": 0,
- "value_type": "cumulative"
- },
- "type": "graph",
- "xaxis": {
- "mode": "time",
- "show": true,
- "values": []
- },
- "yaxes": [
- {
- "format": "bytes",
- "logBase": 1,
- "show": true
- },
- {
- "format": "ms",
- "logBase": 1,
- "show": true
- }
- ]
- },
- {
- "colorBackground": false,
- "colorValue": false,
- "colors": [
- "rgba(50, 172, 45, 0.97)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(245, 54, 54, 0.9)"
- ],
- "datasource": "${DS_PROMETHEUS}",
- "editable": false,
- "format": "percentunit",
- "gauge": {
- "maxValue": 1,
- "minValue": 0,
- "show": true,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "hideTimeOverride": false,
- "id": 12,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "postfix": "",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 3,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false
- },
- "targets": [
- {
- "expr": "(sum(node_filesystem_size{device!=\"rootfs\"}) - sum(node_filesystem_free{device!=\"rootfs\"})) / sum(node_filesystem_size{device!=\"rootfs\"})",
- "intervalFactor": 2,
- "refId": "A",
- "step": 60,
- "target": ""
- }
- ],
- "thresholds": "0.75, 0.9",
- "title": "Disk Space Usage",
- "transparent": false,
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "N/A",
- "value": "null"
- }
- ],
- "valueName": "current"
- }
- ],
- "showTitle": false,
- "title": "New Row",
- "titleSize": "h6"
- },
- {
- "collapse": false,
- "editable": false,
- "height": "250px",
- "panels": [
- {
- "aliasColors": {},
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": "${DS_PROMETHEUS}",
- "editable": false,
- "error": false,
- "fill": 1,
- "grid": {
- "threshold1Color": "rgba(216, 200, 27, 0.27)",
- "threshold2Color": "rgba(234, 112, 112, 0.22)"
- },
- "id": 8,
- "isNew": false,
- "legend": {
- "alignAsTable": false,
- "avg": false,
- "current": false,
- "hideEmpty": false,
- "hideZero": false,
- "max": false,
- "min": false,
- "rightSide": false,
- "show": true,
- "total": false
- },
- "lines": true,
- "linewidth": 2,
- "links": [],
- "nullPointMode": "connected",
- "percentage": false,
- "pointradius": 5,
- "points": false,
- "renderer": "flot",
- "seriesOverrides": [
- {
- "alias": "transmitted",
- "yaxis": 2
- }
- ],
- "spaceLength": 10,
- "span": 6,
- "stack": false,
- "steppedLine": false,
- "targets": [
- {
- "expr": "sum(rate(node_network_receive_bytes{device!~\"lo\"}[5m]))",
- "hide": false,
- "intervalFactor": 2,
- "legendFormat": "",
- "refId": "A",
- "step": 10,
- "target": ""
- }
- ],
- "title": "Network Received",
- "tooltip": {
- "msResolution": false,
- "shared": true,
- "sort": 0,
- "value_type": "cumulative"
- },
- "type": "graph",
- "xaxis": {
- "mode": "time",
- "show": true,
- "values": []
- },
- "yaxes": [
- {
- "format": "bytes",
- "logBase": 1,
- "show": true
- },
- {
- "format": "bytes",
- "logBase": 1,
- "show": true
- }
- ]
- },
- {
- "aliasColors": {},
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": "${DS_PROMETHEUS}",
- "editable": false,
- "error": false,
- "fill": 1,
- "grid": {
- "threshold1Color": "rgba(216, 200, 27, 0.27)",
- "threshold2Color": "rgba(234, 112, 112, 0.22)"
- },
- "id": 10,
- "isNew": false,
- "legend": {
- "alignAsTable": false,
- "avg": false,
- "current": false,
- "hideEmpty": false,
- "hideZero": false,
- "max": false,
- "min": false,
- "rightSide": false,
- "show": true,
- "total": false
- },
- "lines": true,
- "linewidth": 2,
- "links": [],
- "nullPointMode": "connected",
- "percentage": false,
- "pointradius": 5,
- "points": false,
- "renderer": "flot",
- "seriesOverrides": [
- {
- "alias": "transmitted",
- "yaxis": 2
- }
- ],
- "spaceLength": 10,
- "span": 6,
- "stack": false,
- "steppedLine": false,
- "targets": [
- {
- "expr": "sum(rate(node_network_transmit_bytes{device!~\"lo\"}[5m]))",
- "hide": false,
- "intervalFactor": 2,
- "legendFormat": "",
- "refId": "B",
- "step": 10,
- "target": ""
- }
- ],
- "title": "Network Transmitted",
- "tooltip": {
- "msResolution": false,
- "shared": true,
- "sort": 0,
- "value_type": "cumulative"
- },
- "type": "graph",
- "xaxis": {
- "mode": "time",
- "show": true,
- "values": []
- },
- "yaxes": [
- {
- "format": "bytes",
- "logBase": 1,
- "show": true
- },
- {
- "format": "bytes",
- "logBase": 1,
- "show": true
- }
- ]
- }
- ],
- "showTitle": false,
- "title": "New Row",
- "titleSize": "h6"
- },
- {
- "collapse": false,
- "editable": false,
- "height": "276px",
- "panels": [
- {
- "aliasColors": {},
- "bars": false,
- "dashes": false,
- "datasource": "${DS_PROMETHEUS}",
- "editable": false,
- "error": false,
- "fill": 1,
- "grid": {
- "threshold1Color": "rgba(216, 200, 27, 0.27)",
- "threshold2Color": "rgba(234, 112, 112, 0.22)"
- },
- "id": 11,
- "isNew": true,
- "legend": {
- "alignAsTable": false,
- "avg": false,
- "current": false,
- "hideEmpty": false,
- "hideZero": false,
- "max": false,
- "min": false,
- "rightSide": false,
- "show": true,
- "total": false
- },
- "lines": true,
- "linewidth": 2,
- "links": [],
- "nullPointMode": "connected",
- "percentage": false,
- "pointradius": 5,
- "points": false,
- "renderer": "flot",
- "seriesOverrides": [],
- "spaceLength": 11,
- "span": 9,
- "stack": false,
- "steppedLine": false,
- "targets": [
- {
- "expr": "sum(kube_pod_info)",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "Current number of Pods",
- "refId": "A",
- "step": 10
- },
- {
- "expr": "sum(kube_node_status_capacity_pods)",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "Maximum capacity of pods",
- "refId": "B",
- "step": 10
- }
- ],
- "title": "Cluster Pod Utilization",
- "tooltip": {
- "msResolution": false,
- "shared": true,
- "sort": 0,
- "value_type": "individual"
- },
- "type": "graph",
- "xaxis": {
- "mode": "time",
- "show": true,
- "values": []
- },
- "yaxes": [
- {
- "format": "short",
- "logBase": 1,
- "show": true
- },
- {
- "format": "short",
- "logBase": 1,
- "show": true
- }
- ]
- },
- {
- "colorBackground": false,
- "colorValue": false,
- "colors": [
- "rgba(50, 172, 45, 0.97)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(245, 54, 54, 0.9)"
- ],
- "datasource": "${DS_PROMETHEUS}",
- "editable": false,
- "format": "percent",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": true,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "hideTimeOverride": false,
- "id": 7,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "postfix": "",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 3,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false
- },
- "targets": [
- {
- "expr": "100 - (sum(kube_node_status_capacity_pods) - sum(kube_pod_info)) / sum(kube_node_status_capacity_pods) * 100",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "",
- "refId": "A",
- "step": 60,
- "target": ""
- }
- ],
- "thresholds": "80, 90",
- "title": "Pod Utilization",
- "transparent": false,
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "N/A",
- "value": "null"
- }
- ],
- "valueName": "current"
- }
- ],
- "showTitle": false,
- "title": "New Row",
- "titleSize": "h6"
- }
- ],
- "schemaVersion": 14,
- "sharedCrosshair": false,
- "style": "dark",
- "tags": [],
- "templating": {
- "list": []
- },
- "time": {
- "from": "now-1h",
- "to": "now"
- },
- "timepicker": {
- "refresh_intervals": [
- "5s",
- "10s",
- "30s",
- "1m",
- "5m",
- "15m",
- "30m",
- "1h",
- "2h",
- "1d"
- ],
- "time_options": [
- "5m",
- "15m",
- "1h",
- "6h",
- "12h",
- "24h",
- "2d",
- "7d",
- "30d"
- ]
- },
- "timezone": "browser",
- "title": "Kubernetes Capacity Planning",
- "version": 4
- }
- ,
- "inputs": [
- {
- "name": "DS_PROMETHEUS",
- "pluginId": "prometheus",
- "type": "datasource",
- "value": "prometheus"
- }
- ],
- "overwrite": true
- }
- kubernetes-cluster-health-dashboard.json: |+
- {
- "dashboard":
- {
- "__inputs": [
- {
- "description": "",
- "label": "prometheus",
- "name": "DS_PROMETHEUS",
- "pluginId": "prometheus",
- "pluginName": "Prometheus",
- "type": "datasource"
- }
- ],
- "annotations": {
- "list": []
- },
- "editable": false,
- "graphTooltip": 0,
- "hideControls": false,
- "links": [],
- "refresh": "10s",
- "rows": [
- {
- "collapse": false,
- "editable": false,
- "height": "254px",
- "panels": [
- {
- "colorBackground": false,
- "colorValue": true,
- "colors": [
- "rgba(50, 172, 45, 0.97)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(245, 54, 54, 0.9)"
- ],
- "datasource": "${DS_PROMETHEUS}",
- "editable": false,
- "format": "none",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": false,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "hideTimeOverride": false,
- "id": 1,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "postfix": "",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 3,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false
- },
- "targets": [
- {
- "expr": "sum(up{job=~\"apiserver|kube-scheduler|kube-controller-manager\"} == 0)",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "",
- "refId": "A",
- "step": 600
- }
- ],
- "thresholds": "1, 3",
- "title": "Control Plane Components Down",
- "transparent": false,
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "Everything UP and healthy",
- "value": "null"
- },
- {
- "op": "=",
- "text": "",
- "value": ""
- }
- ],
- "valueName": "avg"
- },
- {
- "colorBackground": false,
- "colorValue": true,
- "colors": [
- "rgba(50, 172, 45, 0.97)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(245, 54, 54, 0.9)"
- ],
- "datasource": "${DS_PROMETHEUS}",
- "editable": false,
- "format": "none",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": false,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "hideTimeOverride": false,
- "id": 2,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "postfix": "",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 3,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false
- },
- "targets": [
- {
- "expr": "sum(ALERTS{alertstate=\"firing\",alertname!=\"DeadMansSwitch\"})",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "",
- "refId": "A",
- "step": 600
- }
- ],
- "thresholds": "1, 3",
- "title": "Alerts Firing",
- "transparent": false,
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "0",
- "value": "null"
- }
- ],
- "valueName": "current"
- },
- {
- "colorBackground": false,
- "colorValue": true,
- "colors": [
- "rgba(50, 172, 45, 0.97)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(245, 54, 54, 0.9)"
- ],
- "datasource": "${DS_PROMETHEUS}",
- "editable": false,
- "format": "none",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": false,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "hideTimeOverride": false,
- "id": 3,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "postfix": "",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 3,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false
- },
- "targets": [
- {
- "expr": "sum(ALERTS{alertstate=\"pending\",alertname!=\"DeadMansSwitch\"})",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "",
- "refId": "A",
- "step": 600
- }
- ],
- "thresholds": "3, 5",
- "title": "Alerts Pending",
- "transparent": false,
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "0",
- "value": "null"
- }
- ],
- "valueName": "current"
- },
- {
- "colorBackground": false,
- "colorValue": true,
- "colors": [
- "rgba(50, 172, 45, 0.97)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(245, 54, 54, 0.9)"
- ],
- "datasource": "${DS_PROMETHEUS}",
- "editable": false,
- "format": "none",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": false,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "hideTimeOverride": false,
- "id": 4,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "postfix": "",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 3,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false
- },
- "targets": [
- {
- "expr": "count(increase(kube_pod_container_status_restarts[1h]) > 5)",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "",
- "refId": "A",
- "step": 600
- }
- ],
- "thresholds": "1, 3",
- "title": "Crashlooping Pods",
- "transparent": false,
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "0",
- "value": "null"
- }
- ],
- "valueName": "current"
- }
- ],
- "showTitle": false,
- "title": "Row",
- "titleSize": "h6"
- },
- {
- "collapse": false,
- "editable": false,
- "height": "250px",
- "panels": [
- {
- "colorBackground": false,
- "colorValue": true,
- "colors": [
- "rgba(50, 172, 45, 0.97)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(245, 54, 54, 0.9)"
- ],
- "datasource": "${DS_PROMETHEUS}",
- "editable": false,
- "format": "none",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": false,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "hideTimeOverride": false,
- "id": 5,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "postfix": "",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 3,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false
- },
- "targets": [
- {
- "expr": "sum(kube_node_status_condition{condition=\"Ready\",status!=\"true\"})",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "",
- "refId": "A",
- "step": 600
- }
- ],
- "thresholds": "1, 3",
- "title": "Node Not Ready",
- "transparent": false,
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "N/A",
- "value": "null"
- }
- ],
- "valueName": "current"
- },
- {
- "colorBackground": false,
- "colorValue": true,
- "colors": [
- "rgba(50, 172, 45, 0.97)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(245, 54, 54, 0.9)"
- ],
- "datasource": "${DS_PROMETHEUS}",
- "editable": false,
- "format": "none",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": false,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "hideTimeOverride": false,
- "id": 6,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "postfix": "",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 3,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false
- },
- "targets": [
- {
- "expr": "sum(kube_node_status_condition{condition=\"DiskPressure\",status=\"true\"})",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "",
- "refId": "A",
- "step": 600
- }
- ],
- "thresholds": "1, 3",
- "title": "Node Disk Pressure",
- "transparent": false,
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "N/A",
- "value": "null"
- }
- ],
- "valueName": "current"
- },
- {
- "colorBackground": false,
- "colorValue": true,
- "colors": [
- "rgba(50, 172, 45, 0.97)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(245, 54, 54, 0.9)"
- ],
- "datasource": "${DS_PROMETHEUS}",
- "editable": false,
- "format": "none",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": false,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "hideTimeOverride": false,
- "id": 7,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "postfix": "",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 3,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false
- },
- "targets": [
- {
- "expr": "sum(kube_node_status_condition{condition=\"MemoryPressure\",status=\"true\"})",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "",
- "refId": "A",
- "step": 600
- }
- ],
- "thresholds": "1, 3",
- "title": "Node Memory Pressure",
- "transparent": false,
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "N/A",
- "value": "null"
- }
- ],
- "valueName": "current"
- },
- {
- "colorBackground": false,
- "colorValue": true,
- "colors": [
- "rgba(50, 172, 45, 0.97)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(245, 54, 54, 0.9)"
- ],
- "datasource": "${DS_PROMETHEUS}",
- "editable": false,
- "format": "none",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": false,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "hideTimeOverride": false,
- "id": 8,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "postfix": "",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 3,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false
- },
- "targets": [
- {
- "expr": "sum(kube_node_spec_unschedulable)",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "",
- "refId": "A",
- "step": 600
- }
- ],
- "thresholds": "1, 3",
- "title": "Nodes Unschedulable",
- "transparent": false,
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "N/A",
- "value": "null"
- }
- ],
- "valueName": "current"
- }
- ],
- "showTitle": false,
- "title": "Row",
- "titleSize": "h6"
- }
- ],
- "schemaVersion": 14,
- "sharedCrosshair": false,
- "style": "dark",
- "tags": [],
- "templating": {
- "list": []
- },
- "time": {
- "from": "now-6h",
- "to": "now"
- },
- "timepicker": {
- "refresh_intervals": [
- "5s",
- "10s",
- "30s",
- "1m",
- "5m",
- "15m",
- "30m",
- "1h",
- "2h",
- "1d"
- ],
- "time_options": [
- "5m",
- "15m",
- "1h",
- "6h",
- "12h",
- "24h",
- "2d",
- "7d",
- "30d"
- ]
- },
- "timezone": "browser",
- "title": "Kubernetes Cluster Health",
- "version": 9
- }
- ,
- "inputs": [
- {
- "name": "DS_PROMETHEUS",
- "pluginId": "prometheus",
- "type": "datasource",
- "value": "prometheus"
- }
- ],
- "overwrite": true
- }
- kubernetes-cluster-status-dashboard.json: |+
- {
- "dashboard":
- {
- "__inputs": [
- {
- "description": "",
- "label": "prometheus",
- "name": "DS_PROMETHEUS",
- "pluginId": "prometheus",
- "pluginName": "Prometheus",
- "type": "datasource"
- }
- ],
- "annotations": {
- "list": []
- },
- "editable": false,
- "graphTooltip": 0,
- "hideControls": false,
- "links": [],
- "rows": [
- {
- "collapse": false,
- "editable": false,
- "height": "129px",
- "panels": [
- {
- "colorBackground": false,
- "colorValue": true,
- "colors": [
- "rgba(50, 172, 45, 0.97)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(245, 54, 54, 0.9)"
- ],
- "datasource": "${DS_PROMETHEUS}",
- "editable": false,
- "format": "none",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": false,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "id": 5,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 6,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false
- },
- "targets": [
- {
- "expr": "sum(up{job=~\"apiserver|kube-scheduler|kube-controller-manager\"} == 0)",
- "format": "time_series",
- "intervalFactor": 2,
- "refId": "A",
- "step": 600
- }
- ],
- "thresholds": "1, 3",
- "title": "Control Plane UP",
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "UP",
- "value": "null"
- }
- ],
- "valueName": "total"
- },
- {
- "colorBackground": false,
- "colorValue": true,
- "colors": [
- "rgba(50, 172, 45, 0.97)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(245, 54, 54, 0.9)"
- ],
- "datasource": "${DS_PROMETHEUS}",
- "editable": false,
- "format": "none",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": false,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "id": 6,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 6,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false
- },
- "targets": [
- {
- "expr": "sum(ALERTS{alertstate=\"firing\",alertname!=\"DeadMansSwitch\"})",
- "format": "time_series",
- "intervalFactor": 2,
- "refId": "A",
- "step": 600
- }
- ],
- "thresholds": "3, 5",
- "title": "Alerts Firing",
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "0",
- "value": "null"
- }
- ],
- "valueName": "current"
- }
- ],
- "showTitle": true,
- "title": "Cluster Health",
- "titleSize": "h6"
- },
- {
- "collapse": false,
- "editable": false,
- "height": "168px",
- "panels": [
- {
- "colorBackground": false,
- "colorValue": false,
- "colors": [
- "rgba(245, 54, 54, 0.9)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(50, 172, 45, 0.97)"
- ],
- "datasource": "${DS_PROMETHEUS}",
- "editable": false,
- "format": "percent",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": true,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "id": 1,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 3,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false
- },
- "targets": [
- {
- "expr": "(sum(up{job=\"apiserver\"} == 1) / count(up{job=\"apiserver\"})) * 100",
- "format": "time_series",
- "intervalFactor": 2,
- "refId": "A",
- "step": 600
- }
- ],
- "thresholds": "50, 80",
- "title": "API Servers UP",
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "N/A",
- "value": "null"
- }
- ],
- "valueName": "current"
- },
- {
- "colorBackground": false,
- "colorValue": false,
- "colors": [
- "rgba(245, 54, 54, 0.9)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(50, 172, 45, 0.97)"
- ],
- "datasource": "${DS_PROMETHEUS}",
- "editable": false,
- "format": "percent",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": true,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "id": 2,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 3,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false
- },
- "targets": [
- {
- "expr": "(sum(up{job=\"kube-controller-manager\"} == 1) / count(up{job=\"kube-controller-manager\"})) * 100",
- "format": "time_series",
- "intervalFactor": 2,
- "refId": "A",
- "step": 600
- }
- ],
- "thresholds": "50, 80",
- "title": "Controller Managers UP",
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "N/A",
- "value": "null"
- }
- ],
- "valueName": "current"
- },
- {
- "colorBackground": false,
- "colorValue": false,
- "colors": [
- "rgba(245, 54, 54, 0.9)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(50, 172, 45, 0.97)"
- ],
- "datasource": "${DS_PROMETHEUS}",
- "editable": false,
- "format": "percent",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": true,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "id": 3,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 3,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false
- },
- "targets": [
- {
- "expr": "(sum(up{job=\"kube-scheduler\"} == 1) / count(up{job=\"kube-scheduler\"})) * 100",
- "format": "time_series",
- "intervalFactor": 2,
- "refId": "A",
- "step": 600
- }
- ],
- "thresholds": "50, 80",
- "title": "Schedulers UP",
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "N/A",
- "value": "null"
- }
- ],
- "valueName": "current"
- },
- {
- "colorBackground": false,
- "colorValue": true,
- "colors": [
- "rgba(50, 172, 45, 0.97)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(245, 54, 54, 0.9)"
- ],
- "datasource": "${DS_PROMETHEUS}",
- "editable": false,
- "format": "none",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": false,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "id": 4,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 3,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false
- },
- "targets": [
- {
- "expr": "count(increase(kube_pod_container_status_restarts{namespace=~\"kube-system|tectonic-system\"}[1h]) > 5)",
- "format": "time_series",
- "intervalFactor": 2,
- "refId": "A",
- "step": 600
- }
- ],
- "thresholds": "1, 3",
- "title": "Crashlooping Control Plane Pods",
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "0",
- "value": "null"
- }
- ],
- "valueName": "current"
- }
- ],
- "showTitle": true,
- "title": "Control Plane Status",
- "titleSize": "h6"
- },
- {
- "collapse": false,
- "editable": false,
- "height": "158px",
- "panels": [
- {
- "colorBackground": false,
- "colorValue": false,
- "colors": [
- "rgba(50, 172, 45, 0.97)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(245, 54, 54, 0.9)"
- ],
- "datasource": "${DS_PROMETHEUS}",
- "editable": false,
- "format": "percent",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": true,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "id": 8,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 3,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false
- },
- "targets": [
- {
- "expr": "sum(100 - (avg by (instance) (rate(node_cpu{job=\"node-exporter\",mode=\"idle\"}[5m])) * 100)) / count(node_cpu{job=\"node-exporter\",mode=\"idle\"})",
- "format": "time_series",
- "intervalFactor": 2,
- "refId": "A",
- "step": 600
- }
- ],
- "thresholds": "80, 90",
- "title": "CPU Utilization",
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "N/A",
- "value": "null"
- }
- ],
- "valueName": "avg"
- },
- {
- "colorBackground": false,
- "colorValue": false,
- "colors": [
- "rgba(50, 172, 45, 0.97)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(245, 54, 54, 0.9)"
- ],
- "datasource": "${DS_PROMETHEUS}",
- "editable": false,
- "format": "percent",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": true,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "id": 7,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 3,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false
- },
- "targets": [
- {
- "expr": "((sum(node_memory_MemTotal) - sum(node_memory_MemFree) - sum(node_memory_Buffers) - sum(node_memory_Cached)) / sum(node_memory_MemTotal)) * 100",
- "format": "time_series",
- "intervalFactor": 2,
- "refId": "A",
- "step": 600
- }
- ],
- "thresholds": "80, 90",
- "title": "Memory Utilization",
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "N/A",
- "value": "null"
- }
- ],
- "valueName": "avg"
- },
- {
- "colorBackground": false,
- "colorValue": false,
- "colors": [
- "rgba(50, 172, 45, 0.97)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(245, 54, 54, 0.9)"
- ],
- "datasource": "${DS_PROMETHEUS}",
- "editable": false,
- "format": "percent",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": true,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "id": 9,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 3,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false
- },
- "targets": [
- {
- "expr": "(sum(node_filesystem_size{device!=\"rootfs\"}) - sum(node_filesystem_free{device!=\"rootfs\"})) / sum(node_filesystem_size{device!=\"rootfs\"})",
- "format": "time_series",
- "intervalFactor": 2,
- "refId": "A",
- "step": 600
- }
- ],
- "thresholds": "80, 90",
- "title": "Filesystem Utilization",
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "N/A",
- "value": "null"
- }
- ],
- "valueName": "avg"
- },
- {
- "colorBackground": false,
- "colorValue": false,
- "colors": [
- "rgba(50, 172, 45, 0.97)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(245, 54, 54, 0.9)"
- ],
- "datasource": "${DS_PROMETHEUS}",
- "editable": false,
- "format": "percent",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": true,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "id": 10,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 3,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false
- },
- "targets": [
- {
- "expr": "100 - (sum(kube_node_status_capacity_pods) - sum(kube_pod_info)) / sum(kube_node_status_capacity_pods) * 100",
- "format": "time_series",
- "intervalFactor": 2,
- "refId": "A",
- "step": 600
- }
- ],
- "thresholds": "80, 90",
- "title": "Pod Utilization",
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "N/A",
- "value": "null"
- }
- ],
- "valueName": "avg"
- }
- ],
- "showTitle": true,
- "title": "Capacity Planning",
- "titleSize": "h6"
- }
- ],
- "schemaVersion": 14,
- "sharedCrosshair": false,
- "style": "dark",
- "tags": [],
- "templating": {
- "list": []
- },
- "time": {
- "from": "now-6h",
- "to": "now"
- },
- "timepicker": {
- "refresh_intervals": [
- "5s",
- "10s",
- "30s",
- "1m",
- "5m",
- "15m",
- "30m",
- "1h",
- "2h",
- "1d"
- ],
- "time_options": [
- "5m",
- "15m",
- "1h",
- "6h",
- "12h",
- "24h",
- "2d",
- "7d",
- "30d"
- ]
- },
- "timezone": "browser",
- "title": "Kubernetes Cluster Status",
- "version": 3
- }
- ,
- "inputs": [
- {
- "name": "DS_PROMETHEUS",
- "pluginId": "prometheus",
- "type": "datasource",
- "value": "prometheus"
- }
- ],
- "overwrite": true
- }
- kubernetes-control-plane-status-dashboard.json: |+
- {
- "dashboard":
- {
- "__inputs": [
- {
- "description": "",
- "label": "prometheus",
- "name": "DS_PROMETHEUS",
- "pluginId": "prometheus",
- "pluginName": "Prometheus",
- "type": "datasource"
- }
- ],
- "annotations": {
- "list": []
- },
- "editable": false,
- "graphTooltip": 0,
- "hideControls": false,
- "links": [],
- "rows": [
- {
- "collapse": false,
- "editable": false,
- "height": "250px",
- "panels": [
- {
- "colorBackground": false,
- "colorValue": false,
- "colors": [
- "rgba(245, 54, 54, 0.9)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(50, 172, 45, 0.97)"
- ],
- "datasource": "${DS_PROMETHEUS}",
- "editable": false,
- "format": "percent",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": true,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "hideTimeOverride": false,
- "id": 1,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "postfix": "",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 3,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false
- },
- "targets": [
- {
- "expr": "(sum(up{job=\"apiserver\"} == 1) / sum(up{job=\"apiserver\"})) * 100",
- "format": "time_series",
- "intervalFactor": 2,
- "refId": "A",
- "step": 600
- }
- ],
- "thresholds": "50, 80",
- "title": "API Servers UP",
- "transparent": false,
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "N/A",
- "value": "null"
- }
- ],
- "valueName": "avg"
- },
- {
- "colorBackground": false,
- "colorValue": false,
- "colors": [
- "rgba(245, 54, 54, 0.9)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(50, 172, 45, 0.97)"
- ],
- "datasource": "${DS_PROMETHEUS}",
- "editable": false,
- "format": "percent",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": true,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "hideTimeOverride": false,
- "id": 2,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "postfix": "",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 3,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false
- },
- "targets": [
- {
- "expr": "(sum(up{job=\"kube-controller-manager\"} == 1) / sum(up{job=\"kube-controller-manager\"})) * 100",
- "format": "time_series",
- "intervalFactor": 2,
- "refId": "A",
- "step": 600
- }
- ],
- "thresholds": "50, 80",
- "title": "Controller Managers UP",
- "transparent": false,
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "N/A",
- "value": "null"
- }
- ],
- "valueName": "avg"
- },
- {
- "colorBackground": false,
- "colorValue": false,
- "colors": [
- "rgba(245, 54, 54, 0.9)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(50, 172, 45, 0.97)"
- ],
- "datasource": "${DS_PROMETHEUS}",
- "editable": false,
- "format": "percent",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": true,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "hideTimeOverride": false,
- "id": 3,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "postfix": "",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 3,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false
- },
- "targets": [
- {
- "expr": "(sum(up{job=\"kube-scheduler\"} == 1) / sum(up{job=\"kube-scheduler\"})) * 100",
- "format": "time_series",
- "intervalFactor": 2,
- "refId": "A",
- "step": 600
- }
- ],
- "thresholds": "50, 80",
- "title": "Schedulers UP",
- "transparent": false,
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "N/A",
- "value": "null"
- }
- ],
- "valueName": "avg"
- },
- {
- "colorBackground": false,
- "colorValue": false,
- "colors": [
- "rgba(50, 172, 45, 0.97)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(245, 54, 54, 0.9)"
- ],
- "datasource": "${DS_PROMETHEUS}",
- "editable": false,
- "format": "percent",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": true,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "hideTimeOverride": false,
- "id": 4,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "postfix": "",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 3,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false
- },
- "targets": [
- {
- "expr": "max(sum by(instance) (rate(apiserver_request_count{code=~\"5..\"}[5m])) / sum by(instance) (rate(apiserver_request_count[5m]))) * 100",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "",
- "refId": "A",
- "step": 600
- }
- ],
- "thresholds": "5, 10",
- "title": "API Server Request Error Rate",
- "transparent": false,
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "0",
- "value": "null"
- }
- ],
- "valueName": "avg"
- }
- ],
- "showTitle": false,
- "title": "Dashboard Row",
- "titleSize": "h6"
- },
- {
- "collapse": false,
- "editable": false,
- "height": "250px",
- "panels": [
- {
- "aliasColors": {},
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": "${DS_PROMETHEUS}",
- "editable": false,
- "error": false,
- "fill": 1,
- "grid": {
- "threshold1Color": "rgba(216, 200, 27, 0.27)",
- "threshold2Color": "rgba(234, 112, 112, 0.22)"
- },
- "id": 7,
- "isNew": false,
- "legend": {
- "alignAsTable": false,
- "avg": false,
- "current": false,
- "hideEmpty": false,
- "hideZero": false,
- "max": false,
- "min": false,
- "rightSide": false,
- "show": true,
- "total": false
- },
- "lines": true,
- "linewidth": 1,
- "links": [],
- "nullPointMode": "null",
- "percentage": false,
- "pointradius": 5,
- "points": false,
- "renderer": "flot",
- "seriesOverrides": [],
- "spaceLength": 10,
- "span": 12,
- "stack": false,
- "steppedLine": false,
- "targets": [
- {
- "expr": "sum by(verb) (rate(apiserver_latency_seconds:quantile[5m]) >= 0)",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "",
- "refId": "A",
- "step": 30
- }
- ],
- "title": "API Server Request Latency",
- "tooltip": {
- "msResolution": false,
- "shared": true,
- "sort": 0,
- "value_type": "individual"
- },
- "type": "graph",
- "xaxis": {
- "mode": "time",
- "show": true,
- "values": []
- },
- "yaxes": [
- {
- "format": "short",
- "logBase": 1,
- "show": true
- },
- {
- "format": "short",
- "logBase": 1,
- "show": true
- }
- ]
- }
- ],
- "showTitle": false,
- "title": "Dashboard Row",
- "titleSize": "h6"
- },
- {
- "collapse": false,
- "editable": false,
- "height": "250px",
- "panels": [
- {
- "aliasColors": {},
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": "${DS_PROMETHEUS}",
- "editable": false,
- "error": false,
- "fill": 1,
- "grid": {
- "threshold1Color": "rgba(216, 200, 27, 0.27)",
- "threshold2Color": "rgba(234, 112, 112, 0.22)"
- },
- "id": 5,
- "isNew": false,
- "legend": {
- "alignAsTable": false,
- "avg": false,
- "current": false,
- "hideEmpty": false,
- "hideZero": false,
- "max": false,
- "min": false,
- "rightSide": false,
- "show": true,
- "total": false
- },
- "lines": true,
- "linewidth": 1,
- "links": [],
- "nullPointMode": "null",
- "percentage": false,
- "pointradius": 5,
- "points": false,
- "renderer": "flot",
- "seriesOverrides": [],
- "spaceLength": 10,
- "span": 6,
- "stack": false,
- "steppedLine": false,
- "targets": [
- {
- "expr": "cluster:scheduler_e2e_scheduling_latency_seconds:quantile",
- "format": "time_series",
- "intervalFactor": 2,
- "refId": "A",
- "step": 60
- }
- ],
- "title": "End to End Scheduling Latency",
- "tooltip": {
- "msResolution": false,
- "shared": true,
- "sort": 0,
- "value_type": "individual"
- },
- "type": "graph",
- "xaxis": {
- "mode": "time",
- "show": true,
- "values": []
- },
- "yaxes": [
- {
- "format": "short",
- "logBase": 1,
- "show": true
- },
- {
- "format": "dtdurations",
- "logBase": 1,
- "show": true
- }
- ]
- },
- {
- "aliasColors": {},
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": "${DS_PROMETHEUS}",
- "editable": false,
- "error": false,
- "fill": 1,
- "grid": {
- "threshold1Color": "rgba(216, 200, 27, 0.27)",
- "threshold2Color": "rgba(234, 112, 112, 0.22)"
- },
- "id": 6,
- "isNew": false,
- "legend": {
- "alignAsTable": false,
- "avg": false,
- "current": false,
- "hideEmpty": false,
- "hideZero": false,
- "max": false,
- "min": false,
- "rightSide": false,
- "show": true,
- "total": false
- },
- "lines": true,
- "linewidth": 1,
- "links": [],
- "nullPointMode": "null",
- "percentage": false,
- "pointradius": 5,
- "points": false,
- "renderer": "flot",
- "seriesOverrides": [],
- "spaceLength": 10,
- "span": 6,
- "stack": false,
- "steppedLine": false,
- "targets": [
- {
- "expr": "sum by(instance) (rate(apiserver_request_count{code!~\"2..\"}[5m]))",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "Error Rate",
- "refId": "A",
- "step": 60
- },
- {
- "expr": "sum by(instance) (rate(apiserver_request_count[5m]))",
- "format": "time_series",
- "intervalFactor": 2,
- "legendFormat": "Request Rate",
- "refId": "B",
- "step": 60
- }
- ],
- "title": "API Server Request Rates",
- "tooltip": {
- "msResolution": false,
- "shared": true,
- "sort": 0,
- "value_type": "individual"
- },
- "type": "graph",
- "xaxis": {
- "mode": "time",
- "show": true,
- "values": []
- },
- "yaxes": [
- {
- "format": "short",
- "logBase": 1,
- "show": true
- },
- {
- "format": "short",
- "logBase": 1,
- "show": true
- }
- ]
- }
- ],
- "showTitle": false,
- "title": "Dashboard Row",
- "titleSize": "h6"
- }
- ],
- "schemaVersion": 14,
- "sharedCrosshair": false,
- "style": "dark",
- "tags": [],
- "templating": {
- "list": []
- },
- "time": {
- "from": "now-6h",
- "to": "now"
- },
- "timepicker": {
- "refresh_intervals": [
- "5s",
- "10s",
- "30s",
- "1m",
- "5m",
- "15m",
- "30m",
- "1h",
- "2h",
- "1d"
- ],
- "time_options": [
- "5m",
- "15m",
- "1h",
- "6h",
- "12h",
- "24h",
- "2d",
- "7d",
- "30d"
- ]
- },
- "timezone": "browser",
- "title": "Kubernetes Control Plane Status",
- "version": 3
- }
- ,
- "inputs": [
- {
- "name": "DS_PROMETHEUS",
- "pluginId": "prometheus",
- "type": "datasource",
- "value": "prometheus"
- }
- ],
- "overwrite": true
- }
- kubernetes-resource-requests-dashboard.json: |+
- {
- "dashboard":
- {
- "__inputs": [
- {
- "description": "",
- "label": "prometheus",
- "name": "DS_PROMETHEUS",
- "pluginId": "prometheus",
- "pluginName": "Prometheus",
- "type": "datasource"
- }
- ],
- "annotations": {
- "list": []
- },
- "editable": false,
- "graphTooltip": 0,
- "hideControls": false,
- "links": [],
- "refresh": false,
- "rows": [
- {
- "collapse": false,
- "editable": false,
- "height": "300px",
- "panels": [
- {
- "aliasColors": {},
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": "${DS_PROMETHEUS}",
- "description": "This represents the total [CPU resource requests](https://kubernetes.io/docs/concepts/configuration/manage-compute-resources-container/#meaning-of-cpu) in the cluster.\nFor comparison the total [allocatable CPU cores](https://github.com/kubernetes/community/blob/master/contributors/design-proposals/node-allocatable.md) is also shown.",
- "editable": false,
- "error": false,
- "fill": 1,
- "grid": {
- "threshold1Color": "rgba(216, 200, 27, 0.27)",
- "threshold2Color": "rgba(234, 112, 112, 0.22)"
- },
- "id": 1,
- "isNew": false,
- "legend": {
- "alignAsTable": false,
- "avg": false,
- "current": false,
- "hideEmpty": false,
- "hideZero": false,
- "max": false,
- "min": false,
- "rightSide": false,
- "show": true,
- "total": false
- },
- "lines": true,
- "linewidth": 1,
- "links": [],
- "nullPointMode": "null",
- "percentage": false,
- "pointradius": 5,
- "points": false,
- "renderer": "flot",
- "seriesOverrides": [],
- "spaceLength": 10,
- "span": 9,
- "stack": false,
- "steppedLine": false,
- "targets": [
- {
- "expr": "min(sum(kube_node_status_allocatable_cpu_cores) by (instance))",
- "hide": false,
- "intervalFactor": 2,
- "legendFormat": "Allocatable CPU Cores",
- "refId": "A",
- "step": 20
- },
- {
- "expr": "max(sum(kube_pod_container_resource_requests_cpu_cores) by (instance))",
- "hide": false,
- "intervalFactor": 2,
- "legendFormat": "Requested CPU Cores",
- "refId": "B",
- "step": 20
- }
- ],
- "title": "CPU Cores",
- "tooltip": {
- "msResolution": false,
- "shared": true,
- "sort": 0,
- "value_type": "individual"
- },
- "type": "graph",
- "xaxis": {
- "mode": "time",
- "show": true,
- "values": []
- },
- "yaxes": [
- {
- "format": "short",
- "label": "CPU Cores",
- "logBase": 1,
- "show": true
- },
- {
- "format": "short",
- "logBase": 1,
- "show": true
- }
- ]
- },
- {
- "colorBackground": false,
- "colorValue": false,
- "colors": [
- "rgba(50, 172, 45, 0.97)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(245, 54, 54, 0.9)"
- ],
- "datasource": "${DS_PROMETHEUS}",
- "editable": false,
- "format": "percent",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": true,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "hideTimeOverride": false,
- "id": 2,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "postfix": "",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 3,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": true
- },
- "targets": [
- {
- "expr": "max(sum(kube_pod_container_resource_requests_cpu_cores) by (instance)) / min(sum(kube_node_status_allocatable_cpu_cores) by (instance)) * 100",
- "intervalFactor": 2,
- "legendFormat": "",
- "refId": "A",
- "step": 240
- }
- ],
- "thresholds": "80, 90",
- "title": "CPU Cores",
- "transparent": false,
- "type": "singlestat",
- "valueFontSize": "110%",
- "valueMaps": [
- {
- "op": "=",
- "text": "N/A",
- "value": "null"
- }
- ],
- "valueName": "avg"
- }
- ],
- "showTitle": false,
- "title": "CPU Cores",
- "titleSize": "h6"
- },
- {
- "collapse": false,
- "editable": false,
- "height": "300px",
- "panels": [
- {
- "aliasColors": {},
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": "${DS_PROMETHEUS}",
- "description": "This represents the total [memory resource requests](https://kubernetes.io/docs/concepts/configuration/manage-compute-resources-container/#meaning-of-memory) in the cluster.\nFor comparison the total [allocatable memory](https://github.com/kubernetes/community/blob/master/contributors/design-proposals/node-allocatable.md) is also shown.",
- "editable": false,
- "error": false,
- "fill": 1,
- "grid": {
- "threshold1Color": "rgba(216, 200, 27, 0.27)",
- "threshold2Color": "rgba(234, 112, 112, 0.22)"
- },
- "id": 3,
- "isNew": false,
- "legend": {
- "alignAsTable": false,
- "avg": false,
- "current": false,
- "hideEmpty": false,
- "hideZero": false,
- "max": false,
- "min": false,
- "rightSide": false,
- "show": true,
- "total": false
- },
- "lines": true,
- "linewidth": 1,
- "links": [],
- "nullPointMode": "null",
- "percentage": false,
- "pointradius": 5,
- "points": false,
- "renderer": "flot",
- "seriesOverrides": [],
- "spaceLength": 10,
- "span": 9,
- "stack": false,
- "steppedLine": false,
- "targets": [
- {
- "expr": "min(sum(kube_node_status_allocatable_memory_bytes) by (instance))",
- "hide": false,
- "intervalFactor": 2,
- "legendFormat": "Allocatable Memory",
- "refId": "A",
- "step": 20
- },
- {
- "expr": "max(sum(kube_pod_container_resource_requests_memory_bytes) by (instance))",
- "hide": false,
- "intervalFactor": 2,
- "legendFormat": "Requested Memory",
- "refId": "B",
- "step": 20
- }
- ],
- "title": "Memory",
- "tooltip": {
- "msResolution": false,
- "shared": true,
- "sort": 0,
- "value_type": "individual"
- },
- "type": "graph",
- "xaxis": {
- "mode": "time",
- "show": true,
- "values": []
- },
- "yaxes": [
- {
- "format": "bytes",
- "label": "Memory",
- "logBase": 1,
- "show": true
- },
- {
- "format": "short",
- "logBase": 1,
- "show": true
- }
- ]
- },
- {
- "colorBackground": false,
- "colorValue": false,
- "colors": [
- "rgba(50, 172, 45, 0.97)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(245, 54, 54, 0.9)"
- ],
- "datasource": "${DS_PROMETHEUS}",
- "editable": false,
- "format": "percent",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": true,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "hideTimeOverride": false,
- "id": 4,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "postfix": "",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 3,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": true
- },
- "targets": [
- {
- "expr": "max(sum(kube_pod_container_resource_requests_memory_bytes) by (instance)) / min(sum(kube_node_status_allocatable_memory_bytes) by (instance)) * 100",
- "intervalFactor": 2,
- "legendFormat": "",
- "refId": "A",
- "step": 240
- }
- ],
- "thresholds": "80, 90",
- "title": "Memory",
- "transparent": false,
- "type": "singlestat",
- "valueFontSize": "110%",
- "valueMaps": [
- {
- "op": "=",
- "text": "N/A",
- "value": "null"
- }
- ],
- "valueName": "avg"
- }
- ],
- "showTitle": false,
- "title": "Memory",
- "titleSize": "h6"
- }
- ],
- "schemaVersion": 14,
- "sharedCrosshair": false,
- "style": "dark",
- "tags": [],
- "templating": {
- "list": []
- },
- "time": {
- "from": "now-3h",
- "to": "now"
- },
- "timepicker": {
- "refresh_intervals": [
- "5s",
- "10s",
- "30s",
- "1m",
- "5m",
- "15m",
- "30m",
- "1h",
- "2h",
- "1d"
- ],
- "time_options": [
- "5m",
- "15m",
- "1h",
- "6h",
- "12h",
- "24h",
- "2d",
- "7d",
- "30d"
- ]
- },
- "timezone": "browser",
- "title": "Kubernetes Resource Requests",
- "version": 2
- }
- ,
- "inputs": [
- {
- "name": "DS_PROMETHEUS",
- "pluginId": "prometheus",
- "type": "datasource",
- "value": "prometheus"
- }
- ],
- "overwrite": true
- }
- nodes-dashboard.json: |+
- {
- "dashboard":
- {
- "__inputs": [
- {
- "description": "",
- "label": "prometheus",
- "name": "DS_PROMETHEUS",
- "pluginId": "prometheus",
- "pluginName": "Prometheus",
- "type": "datasource"
- }
- ],
- "annotations": {
- "list": []
- },
- "description": "Dashboard to get an overview of one server",
- "editable": false,
- "gnetId": 22,
- "graphTooltip": 0,
- "hideControls": false,
- "links": [],
- "refresh": false,
- "rows": [
- {
- "collapse": false,
- "editable": false,
- "height": "250px",
- "panels": [
- {
- "aliasColors": {},
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": "${DS_PROMETHEUS}",
- "editable": false,
- "error": false,
- "fill": 1,
- "grid": {
- "threshold1Color": "rgba(216, 200, 27, 0.27)",
- "threshold2Color": "rgba(234, 112, 112, 0.22)"
- },
- "id": 3,
- "isNew": false,
- "legend": {
- "alignAsTable": false,
- "avg": false,
- "current": false,
- "hideEmpty": false,
- "hideZero": false,
- "max": false,
- "min": false,
- "rightSide": false,
- "show": true,
- "total": false
- },
- "lines": true,
- "linewidth": 2,
- "links": [],
- "nullPointMode": "connected",
- "percentage": false,
- "pointradius": 5,
- "points": false,
- "renderer": "flot",
- "seriesOverrides": [],
- "spaceLength": 10,
- "span": 6,
- "stack": false,
- "steppedLine": false,
- "targets": [
- {
- "expr": "100 - (avg by (cpu) (irate(node_cpu{mode=\"idle\", instance=\"$server\"}[5m])) * 100)",
- "hide": false,
- "intervalFactor": 10,
- "legendFormat": "{{cpu}}",
- "refId": "A",
- "step": 50
- }
- ],
- "title": "Idle CPU",
- "tooltip": {
- "msResolution": false,
- "shared": true,
- "sort": 0,
- "value_type": "cumulative"
- },
- "type": "graph",
- "xaxis": {
- "mode": "time",
- "show": true,
- "values": []
- },
- "yaxes": [
- {
- "format": "percent",
- "label": "cpu usage",
- "logBase": 1,
- "max": 100,
- "min": 0,
- "show": true
- },
- {
- "format": "short",
- "logBase": 1,
- "show": true
- }
- ]
- },
- {
- "aliasColors": {},
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": "${DS_PROMETHEUS}",
- "editable": false,
- "error": false,
- "fill": 1,
- "grid": {
- "threshold1Color": "rgba(216, 200, 27, 0.27)",
- "threshold2Color": "rgba(234, 112, 112, 0.22)"
- },
- "id": 9,
- "isNew": false,
- "legend": {
- "alignAsTable": false,
- "avg": false,
- "current": false,
- "hideEmpty": false,
- "hideZero": false,
- "max": false,
- "min": false,
- "rightSide": false,
- "show": true,
- "total": false
- },
- "lines": true,
- "linewidth": 2,
- "links": [],
- "nullPointMode": "connected",
- "percentage": false,
- "pointradius": 5,
- "points": false,
- "renderer": "flot",
- "seriesOverrides": [],
- "spaceLength": 10,
- "span": 6,
- "stack": false,
- "steppedLine": false,
- "targets": [
- {
- "expr": "node_load1{instance=\"$server\"}",
- "intervalFactor": 4,
- "legendFormat": "load 1m",
- "refId": "A",
- "step": 20,
- "target": ""
- },
- {
- "expr": "node_load5{instance=\"$server\"}",
- "intervalFactor": 4,
- "legendFormat": "load 5m",
- "refId": "B",
- "step": 20,
- "target": ""
- },
- {
- "expr": "node_load15{instance=\"$server\"}",
- "intervalFactor": 4,
- "legendFormat": "load 15m",
- "refId": "C",
- "step": 20,
- "target": ""
- }
- ],
- "title": "System Load",
- "tooltip": {
- "msResolution": false,
- "shared": true,
- "sort": 0,
- "value_type": "cumulative"
- },
- "type": "graph",
- "xaxis": {
- "mode": "time",
- "show": true,
- "values": []
- },
- "yaxes": [
- {
- "format": "percentunit",
- "logBase": 1,
- "show": true
- },
- {
- "format": "short",
- "logBase": 1,
- "show": true
- }
- ]
- }
- ],
- "showTitle": false,
- "title": "New Row",
- "titleSize": "h6"
- },
- {
- "collapse": false,
- "editable": false,
- "height": "250px",
- "panels": [
- {
- "aliasColors": {},
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": "${DS_PROMETHEUS}",
- "editable": false,
- "error": false,
- "fill": 1,
- "grid": {
- "threshold1Color": "rgba(216, 200, 27, 0.27)",
- "threshold2Color": "rgba(234, 112, 112, 0.22)"
- },
- "id": 4,
- "isNew": false,
- "legend": {
- "alignAsTable": false,
- "avg": false,
- "current": false,
- "hideEmpty": false,
- "hideZero": false,
- "max": false,
- "min": false,
- "rightSide": false,
- "show": true,
- "total": false
- },
- "lines": true,
- "linewidth": 2,
- "links": [],
- "nullPointMode": "connected",
- "percentage": false,
- "pointradius": 5,
- "points": false,
- "renderer": "flot",
- "seriesOverrides": [
- {
- "alias": "node_memory_SwapFree{instance=\"172.17.0.1:9100\",job=\"prometheus\"}",
- "yaxis": 2
- }
- ],
- "spaceLength": 10,
- "span": 9,
- "stack": true,
- "steppedLine": false,
- "targets": [
- {
- "expr": "node_memory_MemTotal{instance=\"$server\"} - node_memory_MemFree{instance=\"$server\"} - node_memory_Buffers{instance=\"$server\"} - node_memory_Cached{instance=\"$server\"}",
- "hide": false,
- "interval": "",
- "intervalFactor": 2,
- "legendFormat": "memory used",
- "metric": "",
- "refId": "C",
- "step": 10
- },
- {
- "expr": "node_memory_Buffers{instance=\"$server\"}",
- "interval": "",
- "intervalFactor": 2,
- "legendFormat": "memory buffers",
- "metric": "",
- "refId": "E",
- "step": 10
- },
- {
- "expr": "node_memory_Cached{instance=\"$server\"}",
- "intervalFactor": 2,
- "legendFormat": "memory cached",
- "metric": "",
- "refId": "F",
- "step": 10
- },
- {
- "expr": "node_memory_MemFree{instance=\"$server\"}",
- "intervalFactor": 2,
- "legendFormat": "memory free",
- "metric": "",
- "refId": "D",
- "step": 10
- }
- ],
- "title": "Memory Usage",
- "tooltip": {
- "msResolution": false,
- "shared": true,
- "sort": 0,
- "value_type": "individual"
- },
- "type": "graph",
- "xaxis": {
- "mode": "time",
- "show": true,
- "values": []
- },
- "yaxes": [
- {
- "format": "bytes",
- "logBase": 1,
- "min": "0",
- "show": true
- },
- {
- "format": "short",
- "logBase": 1,
- "show": true
- }
- ]
- },
- {
- "colorBackground": false,
- "colorValue": false,
- "colors": [
- "rgba(50, 172, 45, 0.97)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(245, 54, 54, 0.9)"
- ],
- "datasource": "${DS_PROMETHEUS}",
- "editable": false,
- "format": "percent",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": true,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "hideTimeOverride": false,
- "id": 5,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "postfix": "",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 3,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false
- },
- "targets": [
- {
- "expr": "((node_memory_MemTotal{instance=\"$server\"} - node_memory_MemFree{instance=\"$server\"} - node_memory_Buffers{instance=\"$server\"} - node_memory_Cached{instance=\"$server\"}) / node_memory_MemTotal{instance=\"$server\"}) * 100",
- "intervalFactor": 2,
- "refId": "A",
- "step": 60,
- "target": ""
- }
- ],
- "thresholds": "80, 90",
- "title": "Memory Usage",
- "transparent": false,
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "N/A",
- "value": "null"
- }
- ],
- "valueName": "avg"
- }
- ],
- "showTitle": false,
- "title": "New Row",
- "titleSize": "h6"
- },
- {
- "collapse": false,
- "editable": false,
- "height": "250px",
- "panels": [
- {
- "aliasColors": {},
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": "${DS_PROMETHEUS}",
- "editable": false,
- "error": false,
- "fill": 1,
- "grid": {
- "threshold1Color": "rgba(216, 200, 27, 0.27)",
- "threshold2Color": "rgba(234, 112, 112, 0.22)"
- },
- "id": 6,
- "isNew": true,
- "legend": {
- "alignAsTable": false,
- "avg": false,
- "current": false,
- "hideEmpty": false,
- "hideZero": false,
- "max": false,
- "min": false,
- "rightSide": false,
- "show": true,
- "total": false
- },
- "lines": true,
- "linewidth": 2,
- "links": [],
- "nullPointMode": "connected",
- "percentage": false,
- "pointradius": 5,
- "points": false,
- "renderer": "flot",
- "seriesOverrides": [
- {
- "alias": "read",
- "yaxis": 1
- },
- {
- "alias": "{instance=\"172.17.0.1:9100\"}",
- "yaxis": 2
- },
- {
- "alias": "io time",
- "yaxis": 2
- }
- ],
- "spaceLength": 10,
- "span": 9,
- "stack": false,
- "steppedLine": false,
- "targets": [
- {
- "expr": "sum by (instance) (rate(node_disk_bytes_read{instance=\"$server\"}[2m]))",
- "hide": false,
- "intervalFactor": 4,
- "legendFormat": "read",
- "refId": "A",
- "step": 20,
- "target": ""
- },
- {
- "expr": "sum by (instance) (rate(node_disk_bytes_written{instance=\"$server\"}[2m]))",
- "intervalFactor": 4,
- "legendFormat": "written",
- "refId": "B",
- "step": 20
- },
- {
- "expr": "sum by (instance) (rate(node_disk_io_time_ms{instance=\"$server\"}[2m]))",
- "intervalFactor": 4,
- "legendFormat": "io time",
- "refId": "C",
- "step": 20
- }
- ],
- "title": "Disk I/O",
- "tooltip": {
- "msResolution": false,
- "shared": true,
- "sort": 0,
- "value_type": "cumulative"
- },
- "type": "graph",
- "xaxis": {
- "mode": "time",
- "show": true,
- "values": []
- },
- "yaxes": [
- {
- "format": "bytes",
- "logBase": 1,
- "show": true
- },
- {
- "format": "ms",
- "logBase": 1,
- "show": true
- }
- ]
- },
- {
- "colorBackground": false,
- "colorValue": false,
- "colors": [
- "rgba(50, 172, 45, 0.97)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(245, 54, 54, 0.9)"
- ],
- "datasource": "${DS_PROMETHEUS}",
- "editable": false,
- "format": "percentunit",
- "gauge": {
- "maxValue": 1,
- "minValue": 0,
- "show": true,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "hideTimeOverride": false,
- "id": 7,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "postfix": "",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 3,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false
- },
- "targets": [
- {
- "expr": "(sum(node_filesystem_size{device!=\"rootfs\",instance=\"$server\"}) - sum(node_filesystem_free{device!=\"rootfs\",instance=\"$server\"})) / sum(node_filesystem_size{device!=\"rootfs\",instance=\"$server\"})",
- "intervalFactor": 2,
- "refId": "A",
- "step": 60,
- "target": ""
- }
- ],
- "thresholds": "0.75, 0.9",
- "title": "Disk Space Usage",
- "transparent": false,
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "N/A",
- "value": "null"
- }
- ],
- "valueName": "current"
- }
- ],
- "showTitle": false,
- "title": "New Row",
- "titleSize": "h6"
- },
- {
- "collapse": false,
- "editable": false,
- "height": "250px",
- "panels": [
- {
- "aliasColors": {},
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": "${DS_PROMETHEUS}",
- "editable": false,
- "error": false,
- "fill": 1,
- "grid": {
- "threshold1Color": "rgba(216, 200, 27, 0.27)",
- "threshold2Color": "rgba(234, 112, 112, 0.22)"
- },
- "id": 8,
- "isNew": false,
- "legend": {
- "alignAsTable": false,
- "avg": false,
- "current": false,
- "hideEmpty": false,
- "hideZero": false,
- "max": false,
- "min": false,
- "rightSide": false,
- "show": true,
- "total": false
- },
- "lines": true,
- "linewidth": 2,
- "links": [],
- "nullPointMode": "connected",
- "percentage": false,
- "pointradius": 5,
- "points": false,
- "renderer": "flot",
- "seriesOverrides": [
- {
- "alias": "transmitted",
- "yaxis": 2
- }
- ],
- "spaceLength": 10,
- "span": 6,
- "stack": false,
- "steppedLine": false,
- "targets": [
- {
- "expr": "rate(node_network_receive_bytes{instance=\"$server\",device!~\"lo\"}[5m])",
- "hide": false,
- "intervalFactor": 2,
- "legendFormat": "{{device}}",
- "refId": "A",
- "step": 10,
- "target": ""
- }
- ],
- "title": "Network Received",
- "tooltip": {
- "msResolution": false,
- "shared": true,
- "sort": 0,
- "value_type": "cumulative"
- },
- "type": "graph",
- "xaxis": {
- "mode": "time",
- "show": true,
- "values": []
- },
- "yaxes": [
- {
- "format": "bytes",
- "logBase": 1,
- "show": true
- },
- {
- "format": "bytes",
- "logBase": 1,
- "show": true
- }
- ]
- },
- {
- "aliasColors": {},
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": "${DS_PROMETHEUS}",
- "editable": false,
- "error": false,
- "fill": 1,
- "grid": {
- "threshold1Color": "rgba(216, 200, 27, 0.27)",
- "threshold2Color": "rgba(234, 112, 112, 0.22)"
- },
- "id": 10,
- "isNew": false,
- "legend": {
- "alignAsTable": false,
- "avg": false,
- "current": false,
- "hideEmpty": false,
- "hideZero": false,
- "max": false,
- "min": false,
- "rightSide": false,
- "show": true,
- "total": false
- },
- "lines": true,
- "linewidth": 2,
- "links": [],
- "nullPointMode": "connected",
- "percentage": false,
- "pointradius": 5,
- "points": false,
- "renderer": "flot",
- "seriesOverrides": [
- {
- "alias": "transmitted",
- "yaxis": 2
- }
- ],
- "spaceLength": 10,
- "span": 6,
- "stack": false,
- "steppedLine": false,
- "targets": [
- {
- "expr": "rate(node_network_transmit_bytes{instance=\"$server\",device!~\"lo\"}[5m])",
- "hide": false,
- "intervalFactor": 2,
- "legendFormat": "{{device}}",
- "refId": "B",
- "step": 10,
- "target": ""
- }
- ],
- "title": "Network Transmitted",
- "tooltip": {
- "msResolution": false,
- "shared": true,
- "sort": 0,
- "value_type": "cumulative"
- },
- "type": "graph",
- "xaxis": {
- "mode": "time",
- "show": true,
- "values": []
- },
- "yaxes": [
- {
- "format": "bytes",
- "logBase": 1,
- "show": true
- },
- {
- "format": "bytes",
- "logBase": 1,
- "show": true
- }
- ]
- }
- ],
- "showTitle": false,
- "title": "New Row",
- "titleSize": "h6"
- }
- ],
- "schemaVersion": 14,
- "sharedCrosshair": false,
- "style": "dark",
- "tags": [],
- "templating": {
- "list": [
- {
- "allValue": null,
- "current": {},
- "datasource": "${DS_PROMETHEUS}",
- "hide": 0,
- "includeAll": false,
- "label": null,
- "multi": false,
- "name": "server",
- "options": [],
- "query": "label_values(node_boot_time, instance)",
- "refresh": 1,
- "regex": "",
- "sort": 0,
- "tagValuesQuery": "",
- "tags": [],
- "tagsQuery": "",
- "type": "query",
- "useTags": false
- }
- ]
- },
- "time": {
- "from": "now-1h",
- "to": "now"
- },
- "timepicker": {
- "refresh_intervals": [
- "5s",
- "10s",
- "30s",
- "1m",
- "5m",
- "15m",
- "30m",
- "1h",
- "2h",
- "1d"
- ],
- "time_options": [
- "5m",
- "15m",
- "1h",
- "6h",
- "12h",
- "24h",
- "2d",
- "7d",
- "30d"
- ]
- },
- "timezone": "browser",
- "title": "Nodes",
- "version": 2
- }
- ,
- "inputs": [
- {
- "name": "DS_PROMETHEUS",
- "pluginId": "prometheus",
- "type": "datasource",
- "value": "prometheus"
- }
- ],
- "overwrite": true
- }
- pods-dashboard.json: |+
- {
- "dashboard":
- {
- "__inputs": [
- {
- "description": "",
- "label": "prometheus",
- "name": "DS_PROMETHEUS",
- "pluginId": "prometheus",
- "pluginName": "Prometheus",
- "type": "datasource"
- }
- ],
- "annotations": {
- "list": []
- },
- "editable": false,
- "graphTooltip": 1,
- "hideControls": false,
- "links": [],
- "refresh": false,
- "rows": [
- {
- "collapse": false,
- "editable": false,
- "height": "250px",
- "panels": [
- {
- "aliasColors": {},
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": "${DS_PROMETHEUS}",
- "editable": false,
- "error": false,
- "fill": 1,
- "grid": {
- "threshold1Color": "rgba(216, 200, 27, 0.27)",
- "threshold2Color": "rgba(234, 112, 112, 0.22)"
- },
- "id": 1,
- "isNew": false,
- "legend": {
- "alignAsTable": true,
- "avg": true,
- "current": true,
- "hideEmpty": false,
- "hideZero": false,
- "max": false,
- "min": false,
- "rightSide": true,
- "show": true,
- "total": false,
- "values": true
- },
- "lines": true,
- "linewidth": 2,
- "links": [],
- "nullPointMode": "connected",
- "percentage": false,
- "pointradius": 5,
- "points": false,
- "renderer": "flot",
- "seriesOverrides": [],
- "spaceLength": 10,
- "span": 12,
- "stack": false,
- "steppedLine": false,
- "targets": [
- {
- "expr": "sum by(container_name) (container_memory_usage_bytes{pod_name=\"$pod\", container_name=~\"$container\", container_name!=\"POD\"})",
- "interval": "10s",
- "intervalFactor": 1,
- "legendFormat": "Current: {{ container_name }}",
- "metric": "container_memory_usage_bytes",
- "refId": "A",
- "step": 15
- },
- {
- "expr": "kube_pod_container_resource_requests_memory_bytes{pod=\"$pod\", container=~\"$container\"}",
- "interval": "10s",
- "intervalFactor": 2,
- "legendFormat": "Requested: {{ container }}",
- "metric": "kube_pod_container_resource_requests_memory_bytes",
- "refId": "B",
- "step": 20
- },
- {
- "expr": "kube_pod_container_resource_limits_memory_bytes{pod=\"$pod\", container=~\"$container\"}",
- "interval": "10s",
- "intervalFactor": 2,
- "legendFormat": "Limit: {{ container }}",
- "metric": "kube_pod_container_resource_limits_memory_bytes",
- "refId": "C",
- "step": 20
- }
- ],
- "title": "Memory Usage",
- "tooltip": {
- "msResolution": true,
- "shared": true,
- "sort": 0,
- "value_type": "cumulative"
- },
- "type": "graph",
- "xaxis": {
- "mode": "time",
- "show": true,
- "values": []
- },
- "yaxes": [
- {
- "format": "bytes",
- "logBase": 1,
- "show": true
- },
- {
- "format": "short",
- "logBase": 1,
- "show": true
- }
- ]
- }
- ],
- "showTitle": false,
- "title": "Row",
- "titleSize": "h6"
- },
- {
- "collapse": false,
- "editable": false,
- "height": "250px",
- "panels": [
- {
- "aliasColors": {},
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": "${DS_PROMETHEUS}",
- "editable": false,
- "error": false,
- "fill": 1,
- "grid": {
- "threshold1Color": "rgba(216, 200, 27, 0.27)",
- "threshold2Color": "rgba(234, 112, 112, 0.22)"
- },
- "id": 2,
- "isNew": false,
- "legend": {
- "alignAsTable": true,
- "avg": true,
- "current": true,
- "hideEmpty": false,
- "hideZero": false,
- "max": false,
- "min": false,
- "rightSide": true,
- "show": true,
- "total": false,
- "values": true
- },
- "lines": true,
- "linewidth": 2,
- "links": [],
- "nullPointMode": "connected",
- "percentage": false,
- "pointradius": 5,
- "points": false,
- "renderer": "flot",
- "seriesOverrides": [],
- "spaceLength": 10,
- "span": 12,
- "stack": false,
- "steppedLine": false,
- "targets": [
- {
- "expr": "sum by (container_name)(rate(container_cpu_usage_seconds_total{image!=\"\",container_name!=\"POD\",pod_name=\"$pod\"}[1m]))",
- "intervalFactor": 2,
- "legendFormat": "{{ container_name }}",
- "refId": "A",
- "step": 30
- },
- {
- "expr": "kube_pod_container_resource_requests_cpu_cores{pod=\"$pod\", container=~\"$container\"}",
- "interval": "10s",
- "intervalFactor": 2,
- "legendFormat": "Requested: {{ container }}",
- "metric": "kube_pod_container_resource_requests_cpu_cores",
- "refId": "B",
- "step": 20
- },
- {
- "expr": "kube_pod_container_resource_limits_cpu_cores{pod=\"$pod\", container=~\"$container\"}",
- "interval": "10s",
- "intervalFactor": 2,
- "legendFormat": "Limit: {{ container }}",
- "metric": "kube_pod_container_resource_limits_memory_bytes",
- "refId": "C",
- "step": 20
- }
- ],
- "title": "CPU Usage",
- "tooltip": {
- "msResolution": true,
- "shared": true,
- "sort": 0,
- "value_type": "cumulative"
- },
- "type": "graph",
- "xaxis": {
- "mode": "time",
- "show": true,
- "values": []
- },
- "yaxes": [
- {
- "format": "short",
- "logBase": 1,
- "show": true
- },
- {
- "format": "short",
- "logBase": 1,
- "show": true
- }
- ]
- }
- ],
- "showTitle": false,
- "title": "Row",
- "titleSize": "h6"
- },
- {
- "collapse": false,
- "editable": false,
- "height": "250px",
- "panels": [
- {
- "aliasColors": {},
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": "${DS_PROMETHEUS}",
- "editable": false,
- "error": false,
- "fill": 1,
- "grid": {
- "threshold1Color": "rgba(216, 200, 27, 0.27)",
- "threshold2Color": "rgba(234, 112, 112, 0.22)"
- },
- "id": 3,
- "isNew": false,
- "legend": {
- "alignAsTable": true,
- "avg": true,
- "current": true,
- "hideEmpty": false,
- "hideZero": false,
- "max": false,
- "min": false,
- "rightSide": true,
- "show": true,
- "total": false,
- "values": true
- },
- "lines": true,
- "linewidth": 2,
- "links": [],
- "nullPointMode": "connected",
- "percentage": false,
- "pointradius": 5,
- "points": false,
- "renderer": "flot",
- "seriesOverrides": [],
- "spaceLength": 10,
- "span": 12,
- "stack": false,
- "steppedLine": false,
- "targets": [
- {
- "expr": "sort_desc(sum by (pod_name) (rate(container_network_receive_bytes_total{pod_name=\"$pod\"}[1m])))",
- "intervalFactor": 2,
- "legendFormat": "{{ pod_name }}",
- "refId": "A",
- "step": 30
- }
- ],
- "title": "Network I/O",
- "tooltip": {
- "msResolution": true,
- "shared": true,
- "sort": 0,
- "value_type": "cumulative"
- },
- "type": "graph",
- "xaxis": {
- "mode": "time",
- "show": true,
- "values": []
- },
- "yaxes": [
- {
- "format": "bytes",
- "logBase": 1,
- "show": true
- },
- {
- "format": "short",
- "logBase": 1,
- "show": true
- }
- ]
- }
- ],
- "showTitle": false,
- "title": "New Row",
- "titleSize": "h6"
- }
- ],
- "schemaVersion": 14,
- "sharedCrosshair": false,
- "style": "dark",
- "tags": [],
- "templating": {
- "list": [
- {
- "allValue": ".*",
- "current": {},
- "datasource": "${DS_PROMETHEUS}",
- "hide": 0,
- "includeAll": true,
- "label": "Namespace",
- "multi": false,
- "name": "namespace",
- "options": [],
- "query": "label_values(kube_pod_info, namespace)",
- "refresh": 1,
- "regex": "",
- "sort": 0,
- "tagValuesQuery": "",
- "tags": [],
- "tagsQuery": "",
- "type": "query",
- "useTags": false
- },
- {
- "allValue": null,
- "current": {},
- "datasource": "${DS_PROMETHEUS}",
- "hide": 0,
- "includeAll": false,
- "label": "Pod",
- "multi": false,
- "name": "pod",
- "options": [],
- "query": "label_values(kube_pod_info{namespace=~\"$namespace\"}, pod)",
- "refresh": 1,
- "regex": "",
- "sort": 0,
- "tagValuesQuery": "",
- "tags": [],
- "tagsQuery": "",
- "type": "query",
- "useTags": false
- },
- {
- "allValue": ".*",
- "current": {},
- "datasource": "${DS_PROMETHEUS}",
- "hide": 0,
- "includeAll": true,
- "label": "Container",
- "multi": false,
- "name": "container",
- "options": [],
- "query": "label_values(kube_pod_container_info{namespace=\"$namespace\", pod=\"$pod\"}, container)",
- "refresh": 1,
- "regex": "",
- "sort": 0,
- "tagValuesQuery": "",
- "tags": [],
- "tagsQuery": "",
- "type": "query",
- "useTags": false
- }
- ]
- },
- "time": {
- "from": "now-6h",
- "to": "now"
- },
- "timepicker": {
- "refresh_intervals": [
- "5s",
- "10s",
- "30s",
- "1m",
- "5m",
- "15m",
- "30m",
- "1h",
- "2h",
- "1d"
- ],
- "time_options": [
- "5m",
- "15m",
- "1h",
- "6h",
- "12h",
- "24h",
- "2d",
- "7d",
- "30d"
- ]
- },
- "timezone": "browser",
- "title": "Pods",
- "version": 1
- }
- ,
- "inputs": [
- {
- "name": "DS_PROMETHEUS",
- "pluginId": "prometheus",
- "type": "datasource",
- "value": "prometheus"
- }
- ],
- "overwrite": true
- }
- statefulset-dashboard.json: |+
- {
- "dashboard":
- {
- "__inputs": [
- {
- "description": "",
- "label": "prometheus",
- "name": "DS_PROMETHEUS",
- "pluginId": "prometheus",
- "pluginName": "Prometheus",
- "type": "datasource"
- }
- ],
- "annotations": {
- "list": []
- },
- "editable": false,
- "graphTooltip": 1,
- "hideControls": false,
- "links": [],
- "rows": [
- {
- "collapse": false,
- "editable": false,
- "height": "200px",
- "panels": [
- {
- "colorBackground": false,
- "colorValue": false,
- "colors": [
- "rgba(245, 54, 54, 0.9)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(50, 172, 45, 0.97)"
- ],
- "datasource": "${DS_PROMETHEUS}",
- "editable": false,
- "format": "none",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": false,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "id": 8,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "postfix": "cores",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 4,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": true
- },
- "targets": [
- {
- "expr": "sum(rate(container_cpu_usage_seconds_total{namespace=\"$statefulset_namespace\",pod_name=~\"$statefulset_name.*\"}[3m]))",
- "intervalFactor": 2,
- "refId": "A",
- "step": 600
- }
- ],
- "title": "CPU",
- "type": "singlestat",
- "valueFontSize": "110%",
- "valueMaps": [
- {
- "op": "=",
- "text": "N/A",
- "value": "null"
- }
- ],
- "valueName": "avg"
- },
- {
- "colorBackground": false,
- "colorValue": false,
- "colors": [
- "rgba(245, 54, 54, 0.9)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(50, 172, 45, 0.97)"
- ],
- "datasource": "${DS_PROMETHEUS}",
- "editable": false,
- "format": "none",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": false,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "id": 9,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "postfix": "GB",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "80%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 4,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": true
- },
- "targets": [
- {
- "expr": "sum(container_memory_usage_bytes{namespace=\"$statefulset_namespace\",pod_name=~\"$statefulset_name.*\"}) / 1024^3",
- "intervalFactor": 2,
- "refId": "A",
- "step": 600
- }
- ],
- "title": "Memory",
- "type": "singlestat",
- "valueFontSize": "110%",
- "valueMaps": [
- {
- "op": "=",
- "text": "N/A",
- "value": "null"
- }
- ],
- "valueName": "avg"
- },
- {
- "colorBackground": false,
- "colorValue": false,
- "colors": [
- "rgba(245, 54, 54, 0.9)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(50, 172, 45, 0.97)"
- ],
- "datasource": "${DS_PROMETHEUS}",
- "editable": false,
- "format": "Bps",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": false,
- "thresholdLabels": false,
- "thresholdMarkers": false
- },
- "id": 7,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "postfix": "",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 4,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": true
- },
- "targets": [
- {
- "expr": "sum(rate(container_network_transmit_bytes_total{namespace=\"$statefulset_namespace\",pod_name=~\"$statefulset_name.*\"}[3m])) + sum(rate(container_network_receive_bytes_total{namespace=\"$statefulset_namespace\",pod_name=~\"$statefulset_name.*\"}[3m]))",
- "intervalFactor": 2,
- "refId": "A",
- "step": 600
- }
- ],
- "title": "Network",
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "N/A",
- "value": "null"
- }
- ],
- "valueName": "avg"
- }
- ],
- "showTitle": false,
- "title": "Dashboard Row",
- "titleSize": "h6"
- },
- {
- "collapse": false,
- "editable": false,
- "height": "100px",
- "panels": [
- {
- "colorBackground": false,
- "colorValue": false,
- "colors": [
- "rgba(245, 54, 54, 0.9)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(50, 172, 45, 0.97)"
- ],
- "datasource": "${DS_PROMETHEUS}",
- "editable": false,
- "format": "none",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": false,
- "thresholdLabels": false,
- "thresholdMarkers": false
- },
- "id": 5,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 3,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false
- },
- "targets": [
- {
- "expr": "max(kube_statefulset_replicas{statefulset=\"$statefulset_name\",namespace=\"$statefulset_namespace\"}) without (instance, pod)",
- "intervalFactor": 2,
- "metric": "kube_statefulset_replicas",
- "refId": "A",
- "step": 600
- }
- ],
- "title": "Desired Replicas",
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "N/A",
- "value": "null"
- }
- ],
- "valueName": "avg"
- },
- {
- "colorBackground": false,
- "colorValue": false,
- "colors": [
- "rgba(245, 54, 54, 0.9)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(50, 172, 45, 0.97)"
- ],
- "datasource": "${DS_PROMETHEUS}",
- "editable": false,
- "format": "none",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": false,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "id": 6,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 3,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false
- },
- "targets": [
- {
- "expr": "min(kube_statefulset_status_replicas{statefulset=\"$statefulset_name\",namespace=\"$statefulset_namespace\"}) without (instance, pod)",
- "intervalFactor": 2,
- "refId": "A",
- "step": 600
- }
- ],
- "title": "Available Replicas",
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "N/A",
- "value": "null"
- }
- ],
- "valueName": "avg"
- },
- {
- "colorBackground": false,
- "colorValue": false,
- "colors": [
- "rgba(245, 54, 54, 0.9)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(50, 172, 45, 0.97)"
- ],
- "datasource": "${DS_PROMETHEUS}",
- "editable": false,
- "format": "none",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": false,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "id": 3,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 3,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false
- },
- "targets": [
- {
- "expr": "max(kube_statefulset_status_observed_generation{statefulset=\"$statefulset_name\",namespace=\"$statefulset_namespace\"}) without (instance, pod)",
- "intervalFactor": 2,
- "refId": "A",
- "step": 600
- }
- ],
- "title": "Observed Generation",
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "N/A",
- "value": "null"
- }
- ],
- "valueName": "avg"
- },
- {
- "colorBackground": false,
- "colorValue": false,
- "colors": [
- "rgba(245, 54, 54, 0.9)",
- "rgba(237, 129, 40, 0.89)",
- "rgba(50, 172, 45, 0.97)"
- ],
- "datasource": "${DS_PROMETHEUS}",
- "editable": false,
- "format": "none",
- "gauge": {
- "maxValue": 100,
- "minValue": 0,
- "show": false,
- "thresholdLabels": false,
- "thresholdMarkers": true
- },
- "id": 2,
- "links": [],
- "mappingType": 1,
- "mappingTypes": [
- {
- "name": "value to text",
- "value": 1
- },
- {
- "name": "range to text",
- "value": 2
- }
- ],
- "maxDataPoints": 100,
- "nullPointMode": "connected",
- "postfixFontSize": "50%",
- "prefix": "",
- "prefixFontSize": "50%",
- "rangeMaps": [
- {
- "from": "null",
- "text": "N/A",
- "to": "null"
- }
- ],
- "span": 3,
- "sparkline": {
- "fillColor": "rgba(31, 118, 189, 0.18)",
- "full": false,
- "lineColor": "rgb(31, 120, 193)",
- "show": false
- },
- "targets": [
- {
- "expr": "max(kube_statefulset_metadata_generation{statefulset=\"$statefulset_name\",namespace=\"$statefulset_namespace\"}) without (instance, pod)",
- "intervalFactor": 2,
- "refId": "A",
- "step": 600
- }
- ],
- "title": "Metadata Generation",
- "type": "singlestat",
- "valueFontSize": "80%",
- "valueMaps": [
- {
- "op": "=",
- "text": "N/A",
- "value": "null"
- }
- ],
- "valueName": "avg"
- }
- ],
- "showTitle": false,
- "title": "Dashboard Row",
- "titleSize": "h6"
- },
- {
- "collapse": false,
- "editable": false,
- "height": "350px",
- "panels": [
- {
- "aliasColors": {},
- "bars": false,
- "dashLength": 10,
- "dashes": false,
- "datasource": "${DS_PROMETHEUS}",
- "editable": false,
- "error": false,
- "fill": 1,
- "grid": {
- "threshold1Color": "rgba(216, 200, 27, 0.27)",
- "threshold2Color": "rgba(234, 112, 112, 0.22)"
- },
- "id": 1,
- "isNew": true,
- "legend": {
- "alignAsTable": false,
- "avg": false,
- "current": false,
- "hideEmpty": false,
- "hideZero": false,
- "max": false,
- "min": false,
- "rightSide": false,
- "show": true,
- "total": false
- },
- "lines": true,
- "linewidth": 2,
- "links": [],
- "nullPointMode": "connected",
- "percentage": false,
- "pointradius": 5,
- "points": false,
- "renderer": "flot",
- "seriesOverrides": [],
- "spaceLength": 10,
- "span": 12,
- "stack": false,
- "steppedLine": false,
- "targets": [
- {
- "expr": "min(kube_statefulset_status_replicas{statefulset=\"$statefulset_name\",namespace=\"$statefulset_namespace\"}) without (instance, pod)",
- "intervalFactor": 2,
- "legendFormat": "available",
- "refId": "B",
- "step": 30
- },
- {
- "expr": "max(kube_statefulset_replicas{statefulset=\"$statefulset_name\",namespace=\"$statefulset_namespace\"}) without (instance, pod)",
- "intervalFactor": 2,
- "legendFormat": "desired",
- "refId": "E",
- "step": 30
- }
- ],
- "title": "Replicas",
- "tooltip": {
- "msResolution": true,
- "shared": true,
- "sort": 0,
- "value_type": "cumulative"
- },
- "type": "graph",
- "xaxis": {
- "mode": "time",
- "show": true,
- "values": []
- },
- "yaxes": [
- {
- "format": "none",
- "label": "",
- "logBase": 1,
- "show": true
- },
- {
- "format": "short",
- "label": "",
- "logBase": 1,
- "show": false
- }
- ]
- }
- ],
- "showTitle": false,
- "title": "Dashboard Row",
- "titleSize": "h6"
- }
- ],
- "schemaVersion": 14,
- "sharedCrosshair": false,
- "style": "dark",
- "tags": [],
- "templating": {
- "list": [
- {
- "allValue": ".*",
- "current": {},
- "datasource": "${DS_PROMETHEUS}",
- "hide": 0,
- "includeAll": false,
- "label": "Namespace",
- "multi": false,
- "name": "statefulset_namespace",
- "options": [],
- "query": "label_values(kube_statefulset_metadata_generation, namespace)",
- "refresh": 1,
- "regex": "",
- "sort": 0,
- "tagValuesQuery": null,
- "tags": [],
- "tagsQuery": "",
- "type": "query",
- "useTags": false
- },
- {
- "allValue": null,
- "current": {},
- "datasource": "${DS_PROMETHEUS}",
- "hide": 0,
- "includeAll": false,
- "label": "StatefulSet",
- "multi": false,
- "name": "statefulset_name",
- "options": [],
- "query": "label_values(kube_statefulset_metadata_generation{namespace=\"$statefulset_namespace\"}, statefulset)",
- "refresh": 1,
- "regex": "",
- "sort": 0,
- "tagValuesQuery": "",
- "tags": [],
- "tagsQuery": "statefulset",
- "type": "query",
- "useTags": false
- }
- ]
- },
- "time": {
- "from": "now-6h",
- "to": "now"
- },
- "timepicker": {
- "refresh_intervals": [
- "5s",
- "10s",
- "30s",
- "1m",
- "5m",
- "15m",
- "30m",
- "1h",
- "2h",
- "1d"
- ],
- "time_options": [
- "5m",
- "15m",
- "1h",
- "6h",
- "12h",
- "24h",
- "2d",
- "7d",
- "30d"
- ]
- },
- "timezone": "browser",
- "title": "StatefulSet",
- "version": 1
- }
- ,
- "inputs": [
- {
- "name": "DS_PROMETHEUS",
- "pluginId": "prometheus",
- "type": "datasource",
- "value": "prometheus"
- }
- ],
- "overwrite": true
- }
- prometheus-datasource.json: |+
- {
- "access": "proxy",
- "basicAuth": false,
- "name": "prometheus",
- "type": "prometheus",
- "url": "http://prometheus-k8s.monitoring.svc:9090"
- }
----
+ dashboards.yaml: |+
+ - name: 'default'
+ org_id: 1
+ folder: ''
+ type: file
+ options:
+ folder: /grafana-dashboard-definitions/0
diff --git a/manifests/grafana/grafana-datasources.yaml b/manifests/grafana/grafana-datasources.yaml
new file mode 100644
index 0000000000000000000000000000000000000000..33c3081f7497e91daba85b737908c8e30314adec
--- /dev/null
+++ b/manifests/grafana/grafana-datasources.yaml
@@ -0,0 +1,15 @@
+apiVersion: v1
+kind: ConfigMap
+metadata:
+ name: grafana-datasources
+data:
+ prometheus.yaml: |+
+ datasources:
+ - name: prometheus
+ type: prometheus
+ access: proxy
+ org_id: 1
+ url: http://prometheus-k8s.monitoring.svc:9090
+ version: 1
+ editable: false
+
diff --git a/manifests/grafana/grafana-deployment.yaml b/manifests/grafana/grafana-deployment.yaml
index 804657a072a83e0158bd7382b5e250e5aecc85d9..a937b138f80b39f378190e6d3508742cfc1c3788 100644
--- a/manifests/grafana/grafana-deployment.yaml
+++ b/manifests/grafana/grafana-deployment.yaml
@@ -14,25 +14,16 @@ spec:
runAsUser: 65534
containers:
- name: grafana
- image: quay.io/coreos/monitoring-grafana:4.6.3-non-root.1
- env:
- - name: GF_AUTH_BASIC_ENABLED
- value: "true"
- - name: GF_AUTH_ANONYMOUS_ENABLED
- value: "true"
- - name: GF_SECURITY_ADMIN_USER
- valueFrom:
- secretKeyRef:
- name: grafana-credentials
- key: user
- - name: GF_SECURITY_ADMIN_PASSWORD
- valueFrom:
- secretKeyRef:
- name: grafana-credentials
- key: password
+ image: quay.io/coreos/monitoring-grafana:5.0.0-beta1
volumeMounts:
- name: grafana-storage
mountPath: /data
+ - name: grafana-datasources
+ mountPath: /grafana/conf/provisioning/datasources
+ - name: grafana-dashboards
+ mountPath: /grafana/conf/provisioning/dashboards
+ - name: grafana-dashboard-definitions-0
+ mountPath: /grafana-dashboard-definitions/0
ports:
- name: web
containerPort: 3000
@@ -43,35 +34,15 @@ spec:
limits:
memory: 200Mi
cpu: 200m
- - name: grafana-watcher
- image: quay.io/coreos/grafana-watcher:v0.0.8
- args:
- - '--watch-dir=/var/grafana-dashboards-0'
- - '--grafana-url=http://localhost:3000'
- env:
- - name: GRAFANA_USER
- valueFrom:
- secretKeyRef:
- name: grafana-credentials
- key: user
- - name: GRAFANA_PASSWORD
- valueFrom:
- secretKeyRef:
- name: grafana-credentials
- key: password
- resources:
- requests:
- memory: "16Mi"
- cpu: "50m"
- limits:
- memory: "32Mi"
- cpu: "100m"
- volumeMounts:
- - name: grafana-dashboards-0
- mountPath: /var/grafana-dashboards-0
volumes:
- name: grafana-storage
emptyDir: {}
- - name: grafana-dashboards-0
+ - name: grafana-datasources
+ configMap:
+ name: grafana-datasources
+ - name: grafana-dashboards
+ configMap:
+ name: grafana-dashboards
+ - name: grafana-dashboard-definitions-0
configMap:
- name: grafana-dashboards-0
+ name: grafana-dashboard-definitions-0