diff --git a/manifests/grafana-dashboardDefinitions.yaml b/manifests/grafana-dashboardDefinitions.yaml
index 56e92f0bbf4eeae86a2de72e58ae382382fe1ad5..7e0b4844fbd622ca52892c6602c654710618180c 100644
--- a/manifests/grafana-dashboardDefinitions.yaml
+++ b/manifests/grafana-dashboardDefinitions.yaml
@@ -3697,7 +3697,7 @@ items:
"tableColumn": "",
"targets": [
{
- "expr": "sum(up{job=\"kube-controller-manager\"})",
+ "expr": "sum(up{cluster=\"$cluster\", job=\"kube-controller-manager\"})",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "",
@@ -3766,7 +3766,7 @@ items:
"steppedLine": false,
"targets": [
{
- "expr": "sum(rate(workqueue_adds_total{job=\"kube-controller-manager\", instance=~\"$instance\"}[5m])) by (instance, name)",
+ "expr": "sum(rate(workqueue_adds_total{cluster=\"$cluster\", job=\"kube-controller-manager\", instance=~\"$instance\"}[5m])) by (instance, name)",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "{{instance}} {{name}}",
@@ -3872,7 +3872,7 @@ items:
"steppedLine": false,
"targets": [
{
- "expr": "sum(rate(workqueue_depth{job=\"kube-controller-manager\", instance=~\"$instance\"}[5m])) by (instance, name)",
+ "expr": "sum(rate(workqueue_depth{cluster=\"$cluster\", job=\"kube-controller-manager\", instance=~\"$instance\"}[5m])) by (instance, name)",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "{{instance}} {{name}}",
@@ -3978,7 +3978,7 @@ items:
"steppedLine": false,
"targets": [
{
- "expr": "histogram_quantile(0.99, sum(rate(workqueue_queue_duration_seconds_bucket{job=\"kube-controller-manager\", instance=~\"$instance\"}[5m])) by (instance, name, le))",
+ "expr": "histogram_quantile(0.99, sum(rate(workqueue_queue_duration_seconds_bucket{cluster=\"$cluster\", job=\"kube-controller-manager\", instance=~\"$instance\"}[5m])) by (instance, name, le))",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "{{instance}} {{name}}",
@@ -4198,7 +4198,7 @@ items:
"steppedLine": false,
"targets": [
{
- "expr": "histogram_quantile(0.99, sum(rate(rest_client_request_duration_seconds_bucket{job=\"kube-controller-manager\", instance=~\"$instance\", verb=\"POST\"}[5m])) by (verb, url, le))",
+ "expr": "histogram_quantile(0.99, sum(rate(rest_client_request_duration_seconds_bucket{cluster=\"$cluster\", job=\"kube-controller-manager\", instance=~\"$instance\", verb=\"POST\"}[5m])) by (verb, url, le))",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "{{verb}} {{url}}",
@@ -4304,7 +4304,7 @@ items:
"steppedLine": false,
"targets": [
{
- "expr": "histogram_quantile(0.99, sum(rate(rest_client_request_duration_seconds_bucket{job=\"kube-controller-manager\", instance=~\"$instance\", verb=\"GET\"}[5m])) by (verb, url, le))",
+ "expr": "histogram_quantile(0.99, sum(rate(rest_client_request_duration_seconds_bucket{cluster=\"$cluster\", job=\"kube-controller-manager\", instance=~\"$instance\", verb=\"GET\"}[5m])) by (verb, url, le))",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "{{verb}} {{url}}",
@@ -4410,7 +4410,7 @@ items:
"steppedLine": false,
"targets": [
{
- "expr": "process_resident_memory_bytes{job=\"kube-controller-manager\",instance=~\"$instance\"}",
+ "expr": "process_resident_memory_bytes{cluster=\"$cluster\", job=\"kube-controller-manager\",instance=~\"$instance\"}",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "{{instance}}",
@@ -4503,7 +4503,7 @@ items:
"steppedLine": false,
"targets": [
{
- "expr": "rate(process_cpu_seconds_total{job=\"kube-controller-manager\",instance=~\"$instance\"}[5m])",
+ "expr": "rate(process_cpu_seconds_total{cluster=\"$cluster\", job=\"kube-controller-manager\",instance=~\"$instance\"}[5m])",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "{{instance}}",
@@ -4596,7 +4596,7 @@ items:
"steppedLine": false,
"targets": [
{
- "expr": "go_goroutines{job=\"kube-controller-manager\",instance=~\"$instance\"}",
+ "expr": "go_goroutines{cluster=\"$cluster\", job=\"kube-controller-manager\",instance=~\"$instance\"}",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "{{instance}}",
@@ -4680,6 +4680,32 @@ items:
"allValue": null,
"current": {
+ },
+ "datasource": "$datasource",
+ "hide": 2,
+ "includeAll": false,
+ "label": "cluster",
+ "multi": false,
+ "name": "cluster",
+ "options": [
+
+ ],
+ "query": "label_values(kube_pod_info, cluster)",
+ "refresh": 2,
+ "regex": "",
+ "sort": 1,
+ "tagValuesQuery": "",
+ "tags": [
+
+ ],
+ "tagsQuery": "",
+ "type": "query",
+ "useTags": false
+ },
+ {
+ "allValue": null,
+ "current": {
+
},
"datasource": "$datasource",
"hide": 0,
@@ -4690,7 +4716,7 @@ items:
"options": [
],
- "query": "label_values(process_cpu_seconds_total{job=\"kube-controller-manager\"}, instance)",
+ "query": "label_values(process_cpu_seconds_total{cluster=\"$cluster\", job=\"kube-controller-manager\"}, instance)",
"refresh": 2,
"regex": "",
"sort": 1,
@@ -4893,7 +4919,7 @@ items:
"steppedLine": false,
"targets": [
{
- "expr": "sum(kube_pod_container_resource_requests_cpu_cores{cluster=\"$cluster\"}) / sum(kube_node_status_allocatable_cpu_cores{cluster=\"$cluster\"})",
+ "expr": "sum(namespace:kube_pod_container_resource_requests_cpu_cores:sum{cluster=\"$cluster\"}) / sum(kube_node_status_allocatable_cpu_cores{cluster=\"$cluster\"})",
"format": "time_series",
"instant": true,
"intervalFactor": 2,
@@ -5145,7 +5171,7 @@ items:
"steppedLine": false,
"targets": [
{
- "expr": "sum(kube_pod_container_resource_requests_memory_bytes{cluster=\"$cluster\"}) / sum(kube_node_status_allocatable_memory_bytes{cluster=\"$cluster\"})",
+ "expr": "sum(namespace:kube_pod_container_resource_requests_memory_bytes:sum{cluster=\"$cluster\"}) / sum(kube_node_status_allocatable_memory_bytes{cluster=\"$cluster\"})",
"format": "time_series",
"instant": true,
"intervalFactor": 2,
@@ -5624,7 +5650,7 @@ items:
"step": 10
},
{
- "expr": "sum(kube_pod_container_resource_requests_cpu_cores{cluster=\"$cluster\"}) by (namespace)",
+ "expr": "sum(namespace:kube_pod_container_resource_requests_cpu_cores:sum{cluster=\"$cluster\"}) by (namespace)",
"format": "table",
"instant": true,
"intervalFactor": 2,
@@ -5633,7 +5659,7 @@ items:
"step": 10
},
{
- "expr": "sum(node_namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\"}) by (namespace) / sum(kube_pod_container_resource_requests_cpu_cores{cluster=\"$cluster\"}) by (namespace)",
+ "expr": "sum(node_namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\"}) by (namespace) / sum(namespace:kube_pod_container_resource_requests_cpu_cores:sum{cluster=\"$cluster\"}) by (namespace)",
"format": "table",
"instant": true,
"intervalFactor": 2,
@@ -6435,7 +6461,7 @@ items:
"repeatIteration": null,
"repeatRowId": null,
"showTitle": true,
- "title": "Network",
+ "title": "Current Network Usage",
"titleSize": "h6"
},
{
@@ -6475,7 +6501,7 @@ items:
],
"spaceLength": 10,
- "span": 12,
+ "span": 6,
"stack": true,
"steppedLine": false,
"targets": [
@@ -6527,19 +6553,7 @@ items:
"show": false
}
]
- }
- ],
- "repeat": null,
- "repeatIteration": null,
- "repeatRowId": null,
- "showTitle": true,
- "title": "Network",
- "titleSize": "h6"
- },
- {
- "collapse": false,
- "height": "250px",
- "panels": [
+ },
{
"aliasColors": {
@@ -6573,7 +6587,7 @@ items:
],
"spaceLength": 10,
- "span": 12,
+ "span": 6,
"stack": true,
"steppedLine": false,
"targets": [
@@ -6631,7 +6645,7 @@ items:
"repeatIteration": null,
"repeatRowId": null,
"showTitle": true,
- "title": "Network",
+ "title": "Bandwidth",
"titleSize": "h6"
},
{
@@ -6671,7 +6685,7 @@ items:
],
"spaceLength": 10,
- "span": 12,
+ "span": 6,
"stack": true,
"steppedLine": false,
"targets": [
@@ -6723,19 +6737,7 @@ items:
"show": false
}
]
- }
- ],
- "repeat": null,
- "repeatIteration": null,
- "repeatRowId": null,
- "showTitle": true,
- "title": "Network",
- "titleSize": "h6"
- },
- {
- "collapse": false,
- "height": "250px",
- "panels": [
+ },
{
"aliasColors": {
@@ -6769,7 +6771,7 @@ items:
],
"spaceLength": 10,
- "span": 12,
+ "span": 6,
"stack": true,
"steppedLine": false,
"targets": [
@@ -6827,7 +6829,7 @@ items:
"repeatIteration": null,
"repeatRowId": null,
"showTitle": true,
- "title": "Network",
+ "title": "Average Container Bandwidth by Namespace",
"titleSize": "h6"
},
{
@@ -6867,7 +6869,7 @@ items:
],
"spaceLength": 10,
- "span": 12,
+ "span": 6,
"stack": true,
"steppedLine": false,
"targets": [
@@ -6919,19 +6921,7 @@ items:
"show": false
}
]
- }
- ],
- "repeat": null,
- "repeatIteration": null,
- "repeatRowId": null,
- "showTitle": true,
- "title": "Network",
- "titleSize": "h6"
- },
- {
- "collapse": false,
- "height": "250px",
- "panels": [
+ },
{
"aliasColors": {
@@ -6965,7 +6955,7 @@ items:
],
"spaceLength": 10,
- "span": 12,
+ "span": 6,
"stack": true,
"steppedLine": false,
"targets": [
@@ -7023,7 +7013,7 @@ items:
"repeatIteration": null,
"repeatRowId": null,
"showTitle": true,
- "title": "Network",
+ "title": "Rate of Packets",
"titleSize": "h6"
},
{
@@ -7063,7 +7053,7 @@ items:
],
"spaceLength": 10,
- "span": 12,
+ "span": 6,
"stack": true,
"steppedLine": false,
"targets": [
@@ -7115,19 +7105,7 @@ items:
"show": false
}
]
- }
- ],
- "repeat": null,
- "repeatIteration": null,
- "repeatRowId": null,
- "showTitle": true,
- "title": "Network",
- "titleSize": "h6"
- },
- {
- "collapse": false,
- "height": "250px",
- "panels": [
+ },
{
"aliasColors": {
@@ -7161,7 +7139,7 @@ items:
],
"spaceLength": 10,
- "span": 12,
+ "span": 6,
"stack": true,
"steppedLine": false,
"targets": [
@@ -7219,7 +7197,7 @@ items:
"repeatIteration": null,
"repeatRowId": null,
"showTitle": true,
- "title": "Network",
+ "title": "Rate of Packets Dropped",
"titleSize": "h6"
}
],
@@ -7723,8 +7701,9 @@ items:
"color": "#F2495C",
"dashes": true,
"fill": 0,
+ "hiddenSeries": true,
"hideTooltip": true,
- "legend": false,
+ "legend": true,
"linewidth": 2,
"stack": false
},
@@ -7733,8 +7712,9 @@ items:
"color": "#FF9830",
"dashes": true,
"fill": 0,
+ "hiddenSeries": true,
"hideTooltip": true,
- "legend": false,
+ "legend": true,
"linewidth": 2,
"stack": false
}
@@ -8129,8 +8109,9 @@ items:
"color": "#F2495C",
"dashes": true,
"fill": 0,
+ "hiddenSeries": true,
"hideTooltip": true,
- "legend": false,
+ "legend": true,
"linewidth": 2,
"stack": false
},
@@ -8139,8 +8120,9 @@ items:
"color": "#FF9830",
"dashes": true,
"fill": 0,
+ "hiddenSeries": true,
"hideTooltip": true,
- "legend": false,
+ "legend": true,
"linewidth": 2,
"stack": false
}
@@ -8879,7 +8861,7 @@ items:
"repeatIteration": null,
"repeatRowId": null,
"showTitle": true,
- "title": "Network",
+ "title": "Current Network Usage",
"titleSize": "h6"
},
{
@@ -8919,7 +8901,7 @@ items:
],
"spaceLength": 10,
- "span": 12,
+ "span": 6,
"stack": true,
"steppedLine": false,
"targets": [
@@ -8971,19 +8953,7 @@ items:
"show": false
}
]
- }
- ],
- "repeat": null,
- "repeatIteration": null,
- "repeatRowId": null,
- "showTitle": true,
- "title": "Network",
- "titleSize": "h6"
- },
- {
- "collapse": false,
- "height": "250px",
- "panels": [
+ },
{
"aliasColors": {
@@ -9017,7 +8987,7 @@ items:
],
"spaceLength": 10,
- "span": 12,
+ "span": 6,
"stack": true,
"steppedLine": false,
"targets": [
@@ -9075,7 +9045,7 @@ items:
"repeatIteration": null,
"repeatRowId": null,
"showTitle": true,
- "title": "Network",
+ "title": "Bandwidth",
"titleSize": "h6"
},
{
@@ -9115,7 +9085,7 @@ items:
],
"spaceLength": 10,
- "span": 12,
+ "span": 6,
"stack": true,
"steppedLine": false,
"targets": [
@@ -9167,19 +9137,7 @@ items:
"show": false
}
]
- }
- ],
- "repeat": null,
- "repeatIteration": null,
- "repeatRowId": null,
- "showTitle": true,
- "title": "Network",
- "titleSize": "h6"
- },
- {
- "collapse": false,
- "height": "250px",
- "panels": [
+ },
{
"aliasColors": {
@@ -9213,7 +9171,7 @@ items:
],
"spaceLength": 10,
- "span": 12,
+ "span": 6,
"stack": true,
"steppedLine": false,
"targets": [
@@ -9271,7 +9229,7 @@ items:
"repeatIteration": null,
"repeatRowId": null,
"showTitle": true,
- "title": "Network",
+ "title": "Rate of Packets",
"titleSize": "h6"
},
{
@@ -9311,7 +9269,7 @@ items:
],
"spaceLength": 10,
- "span": 12,
+ "span": 6,
"stack": true,
"steppedLine": false,
"targets": [
@@ -9363,19 +9321,7 @@ items:
"show": false
}
]
- }
- ],
- "repeat": null,
- "repeatIteration": null,
- "repeatRowId": null,
- "showTitle": true,
- "title": "Network",
- "titleSize": "h6"
- },
- {
- "collapse": false,
- "height": "250px",
- "panels": [
+ },
{
"aliasColors": {
@@ -9409,7 +9355,7 @@ items:
],
"spaceLength": 10,
- "span": 12,
+ "span": 6,
"stack": true,
"steppedLine": false,
"targets": [
@@ -9467,7 +9413,7 @@ items:
"repeatIteration": null,
"repeatRowId": null,
"showTitle": true,
- "title": "Network",
+ "title": "Rate of Packets Dropped",
"titleSize": "h6"
}
],
@@ -10637,7 +10583,7 @@ items:
"steppedLine": false,
"targets": [
{
- "expr": "sum(node_namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{namespace=\"$namespace\", pod=\"$pod\", container!=\"POD\", cluster=\"$cluster\"}) by (container)",
+ "expr": "sum(node_namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{namespace=\"$namespace\", pod=\"$pod\", cluster=\"$cluster\"}) by (container)",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "{{container}}",
@@ -10751,7 +10697,7 @@ items:
"steppedLine": false,
"targets": [
{
- "expr": "sum(increase(container_cpu_cfs_throttled_periods_total{namespace=\"$namespace\", pod=\"$pod\", container!=\"POD\", container!=\"\", cluster=\"$cluster\"}[5m])) by (container) /sum(increase(container_cpu_cfs_periods_total{namespace=\"$namespace\", pod=\"$pod\", container!=\"POD\", container!=\"\", cluster=\"$cluster\"}[5m])) by (container)",
+ "expr": "sum(increase(container_cpu_cfs_throttled_periods_total{namespace=\"$namespace\", pod=\"$pod\", container!=\"\", cluster=\"$cluster\"}[5m])) by (container) /sum(increase(container_cpu_cfs_periods_total{namespace=\"$namespace\", pod=\"$pod\", container!=\"\", cluster=\"$cluster\"}[5m])) by (container)",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "{{container}}",
@@ -10993,7 +10939,7 @@ items:
],
"targets": [
{
- "expr": "sum(node_namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\", container!=\"POD\"}) by (container)",
+ "expr": "sum(node_namespace_pod_container:container_cpu_usage_seconds_total:sum_rate{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\"}) by (container)",
"format": "table",
"instant": true,
"intervalFactor": 2,
@@ -11127,7 +11073,7 @@ items:
"dashes": true,
"fill": 0,
"hideTooltip": true,
- "legend": false,
+ "legend": true,
"linewidth": 2,
"stack": false
},
@@ -11137,7 +11083,7 @@ items:
"dashes": true,
"fill": 0,
"hideTooltip": true,
- "legend": false,
+ "legend": true,
"linewidth": 2,
"stack": false
}
@@ -11148,7 +11094,7 @@ items:
"steppedLine": false,
"targets": [
{
- "expr": "sum(container_memory_working_set_bytes{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\", container!=\"POD\", container!=\"\", image!=\"\"}) by (container)",
+ "expr": "sum(container_memory_working_set_bytes{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\", container!=\"\", image!=\"\"}) by (container)",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "{{container}}",
@@ -11456,7 +11402,7 @@ items:
],
"targets": [
{
- "expr": "sum(container_memory_working_set_bytes{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\", container!=\"POD\", container!=\"\", image!=\"\"}) by (container)",
+ "expr": "sum(container_memory_working_set_bytes{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\", container!=\"\", image!=\"\"}) by (container)",
"format": "table",
"instant": true,
"intervalFactor": 2,
@@ -11501,7 +11447,7 @@ items:
"step": 10
},
{
- "expr": "sum(container_memory_rss{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\", container != \"\", container != \"POD\"}) by (container)",
+ "expr": "sum(container_memory_rss{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\", container != \"\"}) by (container)",
"format": "table",
"instant": true,
"intervalFactor": 2,
@@ -11510,7 +11456,7 @@ items:
"step": 10
},
{
- "expr": "sum(container_memory_cache{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\", container != \"\", container != \"POD\"}) by (container)",
+ "expr": "sum(container_memory_cache{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\", container != \"\"}) by (container)",
"format": "table",
"instant": true,
"intervalFactor": 2,
@@ -11519,7 +11465,7 @@ items:
"step": 10
},
{
- "expr": "sum(container_memory_swap{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\", container != \"\", container != \"POD\"}) by (container)",
+ "expr": "sum(container_memory_swap{cluster=\"$cluster\", namespace=\"$namespace\", pod=\"$pod\", container != \"\"}) by (container)",
"format": "table",
"instant": true,
"intervalFactor": 2,
@@ -11615,12 +11561,12 @@ items:
],
"spaceLength": 10,
- "span": 12,
+ "span": 6,
"stack": true,
"steppedLine": false,
"targets": [
{
- "expr": "sum(irate(container_network_receive_bytes_total{namespace=~\"$namespace\", pod=~\"$pod\"}[$__rate_interval])) by (pod)",
+ "expr": "sum(irate(container_network_receive_bytes_total{cluster=\"$cluster\", namespace=~\"$namespace\", pod=~\"$pod\"}[$__rate_interval])) by (pod)",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "{{pod}}",
@@ -11667,19 +11613,7 @@ items:
"show": false
}
]
- }
- ],
- "repeat": null,
- "repeatIteration": null,
- "repeatRowId": null,
- "showTitle": true,
- "title": "Network",
- "titleSize": "h6"
- },
- {
- "collapse": false,
- "height": "250px",
- "panels": [
+ },
{
"aliasColors": {
@@ -11714,12 +11648,12 @@ items:
],
"spaceLength": 10,
- "span": 12,
+ "span": 6,
"stack": true,
"steppedLine": false,
"targets": [
{
- "expr": "sum(irate(container_network_transmit_bytes_total{namespace=~\"$namespace\", pod=~\"$pod\"}[$__rate_interval])) by (pod)",
+ "expr": "sum(irate(container_network_transmit_bytes_total{cluster=\"$cluster\", namespace=~\"$namespace\", pod=~\"$pod\"}[$__rate_interval])) by (pod)",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "{{pod}}",
@@ -11772,7 +11706,7 @@ items:
"repeatIteration": null,
"repeatRowId": null,
"showTitle": true,
- "title": "Network",
+ "title": "Bandwidth",
"titleSize": "h6"
},
{
@@ -11813,12 +11747,12 @@ items:
],
"spaceLength": 10,
- "span": 12,
+ "span": 6,
"stack": true,
"steppedLine": false,
"targets": [
{
- "expr": "sum(irate(container_network_receive_packets_total{namespace=~\"$namespace\", pod=~\"$pod\"}[$__rate_interval])) by (pod)",
+ "expr": "sum(irate(container_network_receive_packets_total{cluster=\"$cluster\", namespace=~\"$namespace\", pod=~\"$pod\"}[$__rate_interval])) by (pod)",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "{{pod}}",
@@ -11865,19 +11799,7 @@ items:
"show": false
}
]
- }
- ],
- "repeat": null,
- "repeatIteration": null,
- "repeatRowId": null,
- "showTitle": true,
- "title": "Network",
- "titleSize": "h6"
- },
- {
- "collapse": false,
- "height": "250px",
- "panels": [
+ },
{
"aliasColors": {
@@ -11912,12 +11834,12 @@ items:
],
"spaceLength": 10,
- "span": 12,
+ "span": 6,
"stack": true,
"steppedLine": false,
"targets": [
{
- "expr": "sum(irate(container_network_transmit_packets_total{namespace=~\"$namespace\", pod=~\"$pod\"}[$__rate_interval])) by (pod)",
+ "expr": "sum(irate(container_network_transmit_packets_total{cluster=\"$cluster\", namespace=~\"$namespace\", pod=~\"$pod\"}[$__rate_interval])) by (pod)",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "{{pod}}",
@@ -11970,7 +11892,7 @@ items:
"repeatIteration": null,
"repeatRowId": null,
"showTitle": true,
- "title": "Network",
+ "title": "Rate of Packets",
"titleSize": "h6"
},
{
@@ -12011,12 +11933,12 @@ items:
],
"spaceLength": 10,
- "span": 12,
+ "span": 6,
"stack": true,
"steppedLine": false,
"targets": [
{
- "expr": "sum(irate(container_network_receive_packets_dropped_total{namespace=~\"$namespace\", pod=~\"$pod\"}[$__rate_interval])) by (pod)",
+ "expr": "sum(irate(container_network_receive_packets_dropped_total{cluster=\"$cluster\", namespace=~\"$namespace\", pod=~\"$pod\"}[$__rate_interval])) by (pod)",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "{{pod}}",
@@ -12063,19 +11985,7 @@ items:
"show": false
}
]
- }
- ],
- "repeat": null,
- "repeatIteration": null,
- "repeatRowId": null,
- "showTitle": true,
- "title": "Network",
- "titleSize": "h6"
- },
- {
- "collapse": false,
- "height": "250px",
- "panels": [
+ },
{
"aliasColors": {
@@ -12110,12 +12020,12 @@ items:
],
"spaceLength": 10,
- "span": 12,
+ "span": 6,
"stack": true,
"steppedLine": false,
"targets": [
{
- "expr": "sum(irate(container_network_transmit_packets_dropped_total{namespace=~\"$namespace\", pod=~\"$pod\"}[$__rate_interval])) by (pod)",
+ "expr": "sum(irate(container_network_transmit_packets_dropped_total{cluster=\"$cluster\", namespace=~\"$namespace\", pod=~\"$pod\"}[$__rate_interval])) by (pod)",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "{{pod}}",
@@ -12168,7 +12078,7 @@ items:
"repeatIteration": null,
"repeatRowId": null,
"showTitle": true,
- "title": "Network",
+ "title": "Rate of Packets Dropped",
"titleSize": "h6"
}
],
@@ -13380,7 +13290,7 @@ items:
"repeatIteration": null,
"repeatRowId": null,
"showTitle": true,
- "title": "Network",
+ "title": "Current Network Usage",
"titleSize": "h6"
},
{
@@ -13420,7 +13330,7 @@ items:
],
"spaceLength": 10,
- "span": 12,
+ "span": 6,
"stack": true,
"steppedLine": false,
"targets": [
@@ -13472,19 +13382,7 @@ items:
"show": false
}
]
- }
- ],
- "repeat": null,
- "repeatIteration": null,
- "repeatRowId": null,
- "showTitle": true,
- "title": "Network",
- "titleSize": "h6"
- },
- {
- "collapse": false,
- "height": "250px",
- "panels": [
+ },
{
"aliasColors": {
@@ -13518,7 +13416,7 @@ items:
],
"spaceLength": 10,
- "span": 12,
+ "span": 6,
"stack": true,
"steppedLine": false,
"targets": [
@@ -13576,7 +13474,7 @@ items:
"repeatIteration": null,
"repeatRowId": null,
"showTitle": true,
- "title": "Network",
+ "title": "Bandwidth",
"titleSize": "h6"
},
{
@@ -13616,7 +13514,7 @@ items:
],
"spaceLength": 10,
- "span": 12,
+ "span": 6,
"stack": true,
"steppedLine": false,
"targets": [
@@ -13668,19 +13566,7 @@ items:
"show": false
}
]
- }
- ],
- "repeat": null,
- "repeatIteration": null,
- "repeatRowId": null,
- "showTitle": true,
- "title": "Network",
- "titleSize": "h6"
- },
- {
- "collapse": false,
- "height": "250px",
- "panels": [
+ },
{
"aliasColors": {
@@ -13714,7 +13600,7 @@ items:
],
"spaceLength": 10,
- "span": 12,
+ "span": 6,
"stack": true,
"steppedLine": false,
"targets": [
@@ -13772,7 +13658,7 @@ items:
"repeatIteration": null,
"repeatRowId": null,
"showTitle": true,
- "title": "Network",
+ "title": "Average Container Bandwidth by Pod",
"titleSize": "h6"
},
{
@@ -13812,7 +13698,7 @@ items:
],
"spaceLength": 10,
- "span": 12,
+ "span": 6,
"stack": true,
"steppedLine": false,
"targets": [
@@ -13864,19 +13750,7 @@ items:
"show": false
}
]
- }
- ],
- "repeat": null,
- "repeatIteration": null,
- "repeatRowId": null,
- "showTitle": true,
- "title": "Network",
- "titleSize": "h6"
- },
- {
- "collapse": false,
- "height": "250px",
- "panels": [
+ },
{
"aliasColors": {
@@ -13910,7 +13784,7 @@ items:
],
"spaceLength": 10,
- "span": 12,
+ "span": 6,
"stack": true,
"steppedLine": false,
"targets": [
@@ -13968,7 +13842,7 @@ items:
"repeatIteration": null,
"repeatRowId": null,
"showTitle": true,
- "title": "Network",
+ "title": "Rate of Packets",
"titleSize": "h6"
},
{
@@ -14008,7 +13882,7 @@ items:
],
"spaceLength": 10,
- "span": 12,
+ "span": 6,
"stack": true,
"steppedLine": false,
"targets": [
@@ -14060,19 +13934,7 @@ items:
"show": false
}
]
- }
- ],
- "repeat": null,
- "repeatIteration": null,
- "repeatRowId": null,
- "showTitle": true,
- "title": "Network",
- "titleSize": "h6"
- },
- {
- "collapse": false,
- "height": "250px",
- "panels": [
+ },
{
"aliasColors": {
@@ -14106,7 +13968,7 @@ items:
],
"spaceLength": 10,
- "span": 12,
+ "span": 6,
"stack": true,
"steppedLine": false,
"targets": [
@@ -14164,7 +14026,7 @@ items:
"repeatIteration": null,
"repeatRowId": null,
"showTitle": true,
- "title": "Network",
+ "title": "Rate of Packets Dropped",
"titleSize": "h6"
}
],
@@ -14401,8 +14263,9 @@ items:
"color": "#F2495C",
"dashes": true,
"fill": 0,
+ "hiddenSeries": true,
"hideTooltip": true,
- "legend": false,
+ "legend": true,
"linewidth": 2,
"stack": false
},
@@ -14411,8 +14274,9 @@ items:
"color": "#FF9830",
"dashes": true,
"fill": 0,
+ "hiddenSeries": true,
"hideTooltip": true,
- "legend": false,
+ "legend": true,
"linewidth": 2,
"stack": false
}
@@ -14854,8 +14718,9 @@ items:
"color": "#F2495C",
"dashes": true,
"fill": 0,
+ "hiddenSeries": true,
"hideTooltip": true,
- "legend": false,
+ "legend": true,
"linewidth": 2,
"stack": false
},
@@ -14864,8 +14729,9 @@ items:
"color": "#FF9830",
"dashes": true,
"fill": 0,
+ "hiddenSeries": true,
"hideTooltip": true,
- "legend": false,
+ "legend": true,
"linewidth": 2,
"stack": false
}
@@ -15586,7 +15452,7 @@ items:
"repeatIteration": null,
"repeatRowId": null,
"showTitle": true,
- "title": "Network",
+ "title": "Current Network Usage",
"titleSize": "h6"
},
{
@@ -15626,7 +15492,7 @@ items:
],
"spaceLength": 10,
- "span": 12,
+ "span": 6,
"stack": true,
"steppedLine": false,
"targets": [
@@ -15678,19 +15544,7 @@ items:
"show": false
}
]
- }
- ],
- "repeat": null,
- "repeatIteration": null,
- "repeatRowId": null,
- "showTitle": true,
- "title": "Network",
- "titleSize": "h6"
- },
- {
- "collapse": false,
- "height": "250px",
- "panels": [
+ },
{
"aliasColors": {
@@ -15724,7 +15578,7 @@ items:
],
"spaceLength": 10,
- "span": 12,
+ "span": 6,
"stack": true,
"steppedLine": false,
"targets": [
@@ -15782,7 +15636,7 @@ items:
"repeatIteration": null,
"repeatRowId": null,
"showTitle": true,
- "title": "Network",
+ "title": "Bandwidth",
"titleSize": "h6"
},
{
@@ -15822,7 +15676,7 @@ items:
],
"spaceLength": 10,
- "span": 12,
+ "span": 6,
"stack": true,
"steppedLine": false,
"targets": [
@@ -15874,19 +15728,7 @@ items:
"show": false
}
]
- }
- ],
- "repeat": null,
- "repeatIteration": null,
- "repeatRowId": null,
- "showTitle": true,
- "title": "Network",
- "titleSize": "h6"
- },
- {
- "collapse": false,
- "height": "250px",
- "panels": [
+ },
{
"aliasColors": {
@@ -15920,7 +15762,7 @@ items:
],
"spaceLength": 10,
- "span": 12,
+ "span": 6,
"stack": true,
"steppedLine": false,
"targets": [
@@ -15978,7 +15820,7 @@ items:
"repeatIteration": null,
"repeatRowId": null,
"showTitle": true,
- "title": "Network",
+ "title": "Average Container Bandwidth by Workload",
"titleSize": "h6"
},
{
@@ -16018,7 +15860,7 @@ items:
],
"spaceLength": 10,
- "span": 12,
+ "span": 6,
"stack": true,
"steppedLine": false,
"targets": [
@@ -16070,19 +15912,7 @@ items:
"show": false
}
]
- }
- ],
- "repeat": null,
- "repeatIteration": null,
- "repeatRowId": null,
- "showTitle": true,
- "title": "Network",
- "titleSize": "h6"
- },
- {
- "collapse": false,
- "height": "250px",
- "panels": [
+ },
{
"aliasColors": {
@@ -16116,7 +15946,7 @@ items:
],
"spaceLength": 10,
- "span": 12,
+ "span": 6,
"stack": true,
"steppedLine": false,
"targets": [
@@ -16174,7 +16004,7 @@ items:
"repeatIteration": null,
"repeatRowId": null,
"showTitle": true,
- "title": "Network",
+ "title": "Rate of Packets",
"titleSize": "h6"
},
{
@@ -16214,7 +16044,7 @@ items:
],
"spaceLength": 10,
- "span": 12,
+ "span": 6,
"stack": true,
"steppedLine": false,
"targets": [
@@ -16266,19 +16096,7 @@ items:
"show": false
}
]
- }
- ],
- "repeat": null,
- "repeatIteration": null,
- "repeatRowId": null,
- "showTitle": true,
- "title": "Network",
- "titleSize": "h6"
- },
- {
- "collapse": false,
- "height": "250px",
- "panels": [
+ },
{
"aliasColors": {
@@ -16312,7 +16130,7 @@ items:
],
"spaceLength": 10,
- "span": 12,
+ "span": 6,
"stack": true,
"steppedLine": false,
"targets": [
@@ -16370,7 +16188,7 @@ items:
"repeatIteration": null,
"repeatRowId": null,
"showTitle": true,
- "title": "Network",
+ "title": "Rate of Packets Dropped",
"titleSize": "h6"
}
],
@@ -16399,28 +16217,23 @@ items:
},
{
"allValue": null,
- "auto": false,
- "auto_count": 30,
- "auto_min": "10s",
"current": {
- "text": "deployment",
- "value": "deployment"
+ "text": "",
+ "value": ""
},
"datasource": "$datasource",
- "definition": "label_values(namespace_workload_pod:kube_pod_owner:relabel{namespace=~\"$namespace\", workload=~\".+\"}, workload_type)",
- "hide": 0,
+ "hide": 2,
"includeAll": false,
"label": null,
"multi": false,
- "name": "type",
+ "name": "cluster",
"options": [
],
- "query": "label_values(namespace_workload_pod:kube_pod_owner:relabel{namespace=~\"$namespace\", workload=~\".+\"}, workload_type)",
+ "query": "label_values(kube_pod_info, cluster)",
"refresh": 1,
"regex": "",
- "skipUrlSync": false,
- "sort": 0,
+ "sort": 1,
"tagValuesQuery": "",
"tags": [
@@ -16431,23 +16244,28 @@ items:
},
{
"allValue": null,
+ "auto": false,
+ "auto_count": 30,
+ "auto_min": "10s",
"current": {
- "text": "",
- "value": ""
+ "text": "deployment",
+ "value": "deployment"
},
"datasource": "$datasource",
- "hide": 2,
+ "definition": "label_values(namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=~\"$namespace\", workload=~\".+\"}, workload_type)",
+ "hide": 0,
"includeAll": false,
"label": null,
"multi": false,
- "name": "cluster",
+ "name": "type",
"options": [
],
- "query": "label_values(kube_pod_info, cluster)",
+ "query": "label_values(namespace_workload_pod:kube_pod_owner:relabel{cluster=\"$cluster\", namespace=~\"$namespace\", workload=~\".+\"}, workload_type)",
"refresh": 1,
"regex": "",
- "sort": 1,
+ "skipUrlSync": false,
+ "sort": 0,
"tagValuesQuery": "",
"tags": [
@@ -29895,7 +29713,7 @@ items:
"tableColumn": "",
"targets": [
{
- "expr": "sum(up{job=\"kube-proxy\"})",
+ "expr": "sum(up{cluster=\"$cluster\", job=\"kube-proxy\"})",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "",
@@ -29964,7 +29782,7 @@ items:
"steppedLine": false,
"targets": [
{
- "expr": "sum(rate(kubeproxy_sync_proxy_rules_duration_seconds_count{job=\"kube-proxy\", instance=~\"$instance\"}[5m]))",
+ "expr": "sum(rate(kubeproxy_sync_proxy_rules_duration_seconds_count{cluster=\"$cluster\", job=\"kube-proxy\", instance=~\"$instance\"}[5m]))",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "rate",
@@ -30057,7 +29875,7 @@ items:
"steppedLine": false,
"targets": [
{
- "expr": "histogram_quantile(0.99,rate(kubeproxy_sync_proxy_rules_duration_seconds_bucket{job=\"kube-proxy\", instance=~\"$instance\"}[5m]))",
+ "expr": "histogram_quantile(0.99,rate(kubeproxy_sync_proxy_rules_duration_seconds_bucket{cluster=\"$cluster\", job=\"kube-proxy\", instance=~\"$instance\"}[5m]))",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "{{instance}}",
@@ -30163,7 +29981,7 @@ items:
"steppedLine": false,
"targets": [
{
- "expr": "sum(rate(kubeproxy_network_programming_duration_seconds_count{job=\"kube-proxy\", instance=~\"$instance\"}[5m]))",
+ "expr": "sum(rate(kubeproxy_network_programming_duration_seconds_count{cluster=\"$cluster\", job=\"kube-proxy\", instance=~\"$instance\"}[5m]))",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "rate",
@@ -30256,7 +30074,7 @@ items:
"steppedLine": false,
"targets": [
{
- "expr": "histogram_quantile(0.99, sum(rate(kubeproxy_network_programming_duration_seconds_bucket{job=\"kube-proxy\", instance=~\"$instance\"}[5m])) by (instance, le))",
+ "expr": "histogram_quantile(0.99, sum(rate(kubeproxy_network_programming_duration_seconds_bucket{cluster=\"$cluster\", job=\"kube-proxy\", instance=~\"$instance\"}[5m])) by (instance, le))",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "{{instance}}",
@@ -30362,28 +30180,28 @@ items:
"steppedLine": false,
"targets": [
{
- "expr": "sum(rate(rest_client_requests_total{job=\"kube-proxy\", instance=~\"$instance\",code=~\"2..\"}[5m]))",
+ "expr": "sum(rate(rest_client_requests_total{cluster=\"$cluster\", job=\"kube-proxy\", instance=~\"$instance\",code=~\"2..\"}[5m]))",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "2xx",
"refId": "A"
},
{
- "expr": "sum(rate(rest_client_requests_total{job=\"kube-proxy\", instance=~\"$instance\",code=~\"3..\"}[5m]))",
+ "expr": "sum(rate(rest_client_requests_total{cluster=\"$cluster\", job=\"kube-proxy\", instance=~\"$instance\",code=~\"3..\"}[5m]))",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "3xx",
"refId": "B"
},
{
- "expr": "sum(rate(rest_client_requests_total{job=\"kube-proxy\", instance=~\"$instance\",code=~\"4..\"}[5m]))",
+ "expr": "sum(rate(rest_client_requests_total{cluster=\"$cluster\", job=\"kube-proxy\", instance=~\"$instance\",code=~\"4..\"}[5m]))",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "4xx",
"refId": "C"
},
{
- "expr": "sum(rate(rest_client_requests_total{job=\"kube-proxy\", instance=~\"$instance\",code=~\"5..\"}[5m]))",
+ "expr": "sum(rate(rest_client_requests_total{cluster=\"$cluster\", job=\"kube-proxy\", instance=~\"$instance\",code=~\"5..\"}[5m]))",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "5xx",
@@ -30476,7 +30294,7 @@ items:
"steppedLine": false,
"targets": [
{
- "expr": "histogram_quantile(0.99, sum(rate(rest_client_request_duration_seconds_bucket{job=\"kube-proxy\",instance=~\"$instance\",verb=\"POST\"}[5m])) by (verb, url, le))",
+ "expr": "histogram_quantile(0.99, sum(rate(rest_client_request_duration_seconds_bucket{cluster=\"$cluster\", job=\"kube-proxy\",instance=~\"$instance\",verb=\"POST\"}[5m])) by (verb, url, le))",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "{{verb}} {{url}}",
@@ -30582,7 +30400,7 @@ items:
"steppedLine": false,
"targets": [
{
- "expr": "histogram_quantile(0.99, sum(rate(rest_client_request_duration_seconds_bucket{job=\"kube-proxy\", instance=~\"$instance\", verb=\"GET\"}[5m])) by (verb, url, le))",
+ "expr": "histogram_quantile(0.99, sum(rate(rest_client_request_duration_seconds_bucket{cluster=\"$cluster\", job=\"kube-proxy\", instance=~\"$instance\", verb=\"GET\"}[5m])) by (verb, url, le))",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "{{verb}} {{url}}",
@@ -30688,7 +30506,7 @@ items:
"steppedLine": false,
"targets": [
{
- "expr": "process_resident_memory_bytes{job=\"kube-proxy\",instance=~\"$instance\"}",
+ "expr": "process_resident_memory_bytes{cluster=\"$cluster\", job=\"kube-proxy\",instance=~\"$instance\"}",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "{{instance}}",
@@ -30781,7 +30599,7 @@ items:
"steppedLine": false,
"targets": [
{
- "expr": "rate(process_cpu_seconds_total{job=\"kube-proxy\",instance=~\"$instance\"}[5m])",
+ "expr": "rate(process_cpu_seconds_total{cluster=\"$cluster\", job=\"kube-proxy\",instance=~\"$instance\"}[5m])",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "{{instance}}",
@@ -30874,7 +30692,7 @@ items:
"steppedLine": false,
"targets": [
{
- "expr": "go_goroutines{job=\"kube-proxy\",instance=~\"$instance\"}",
+ "expr": "go_goroutines{cluster=\"$cluster\", job=\"kube-proxy\",instance=~\"$instance\"}",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "{{instance}}",
@@ -30958,6 +30776,32 @@ items:
"allValue": null,
"current": {
+ },
+ "datasource": "$datasource",
+ "hide": 2,
+ "includeAll": false,
+ "label": "cluster",
+ "multi": false,
+ "name": "cluster",
+ "options": [
+
+ ],
+ "query": "label_values(kube_pod_info, cluster)",
+ "refresh": 2,
+ "regex": "",
+ "sort": 1,
+ "tagValuesQuery": "",
+ "tags": [
+
+ ],
+ "tagsQuery": "",
+ "type": "query",
+ "useTags": false
+ },
+ {
+ "allValue": null,
+ "current": {
+
},
"datasource": "$datasource",
"hide": 0,
@@ -30968,7 +30812,7 @@ items:
"options": [
],
- "query": "label_values(kubeproxy_network_programming_duration_seconds_bucket{job=\"kube-proxy\"}, instance)",
+ "query": "label_values(kubeproxy_network_programming_duration_seconds_bucket{cluster=\"$cluster\", job=\"kube-proxy\"}, instance)",
"refresh": 2,
"regex": "",
"sort": 1,
@@ -31115,7 +30959,7 @@ items:
"tableColumn": "",
"targets": [
{
- "expr": "sum(up{job=\"kube-scheduler\"})",
+ "expr": "sum(up{cluster=\"$cluster\", job=\"kube-scheduler\"})",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "",
@@ -31184,28 +31028,28 @@ items:
"steppedLine": false,
"targets": [
{
- "expr": "sum(rate(scheduler_e2e_scheduling_duration_seconds_count{job=\"kube-scheduler\", instance=~\"$instance\"}[5m])) by (instance)",
+ "expr": "sum(rate(scheduler_e2e_scheduling_duration_seconds_count{cluster=\"$cluster\", job=\"kube-scheduler\", instance=~\"$instance\"}[5m])) by (instance)",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "{{instance}} e2e",
"refId": "A"
},
{
- "expr": "sum(rate(scheduler_binding_duration_seconds_count{job=\"kube-scheduler\", instance=~\"$instance\"}[5m])) by (instance)",
+ "expr": "sum(rate(scheduler_binding_duration_seconds_count{cluster=\"$cluster\", job=\"kube-scheduler\", instance=~\"$instance\"}[5m])) by (instance)",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "{{instance}} binding",
"refId": "B"
},
{
- "expr": "sum(rate(scheduler_scheduling_algorithm_duration_seconds_count{job=\"kube-scheduler\", instance=~\"$instance\"}[5m])) by (instance)",
+ "expr": "sum(rate(scheduler_scheduling_algorithm_duration_seconds_count{cluster=\"$cluster\", job=\"kube-scheduler\", instance=~\"$instance\"}[5m])) by (instance)",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "{{instance}} scheduling algorithm",
"refId": "C"
},
{
- "expr": "sum(rate(scheduler_volume_scheduling_duration_seconds_count{job=\"kube-scheduler\", instance=~\"$instance\"}[5m])) by (instance)",
+ "expr": "sum(rate(scheduler_volume_scheduling_duration_seconds_count{cluster=\"$cluster\", job=\"kube-scheduler\", instance=~\"$instance\"}[5m])) by (instance)",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "{{instance}} volume",
@@ -31298,28 +31142,28 @@ items:
"steppedLine": false,
"targets": [
{
- "expr": "histogram_quantile(0.99, sum(rate(scheduler_e2e_scheduling_duration_seconds_bucket{job=\"kube-scheduler\",instance=~\"$instance\"}[5m])) by (instance, le))",
+ "expr": "histogram_quantile(0.99, sum(rate(scheduler_e2e_scheduling_duration_seconds_bucket{cluster=\"$cluster\", job=\"kube-scheduler\",instance=~\"$instance\"}[5m])) by (instance, le))",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "{{instance}} e2e",
"refId": "A"
},
{
- "expr": "histogram_quantile(0.99, sum(rate(scheduler_binding_duration_seconds_bucket{job=\"kube-scheduler\",instance=~\"$instance\"}[5m])) by (instance, le))",
+ "expr": "histogram_quantile(0.99, sum(rate(scheduler_binding_duration_seconds_bucket{cluster=\"$cluster\", job=\"kube-scheduler\",instance=~\"$instance\"}[5m])) by (instance, le))",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "{{instance}} binding",
"refId": "B"
},
{
- "expr": "histogram_quantile(0.99, sum(rate(scheduler_scheduling_algorithm_duration_seconds_bucket{job=\"kube-scheduler\",instance=~\"$instance\"}[5m])) by (instance, le))",
+ "expr": "histogram_quantile(0.99, sum(rate(scheduler_scheduling_algorithm_duration_seconds_bucket{cluster=\"$cluster\", job=\"kube-scheduler\",instance=~\"$instance\"}[5m])) by (instance, le))",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "{{instance}} scheduling algorithm",
"refId": "C"
},
{
- "expr": "histogram_quantile(0.99, sum(rate(scheduler_volume_scheduling_duration_seconds_bucket{job=\"kube-scheduler\",instance=~\"$instance\"}[5m])) by (instance, le))",
+ "expr": "histogram_quantile(0.99, sum(rate(scheduler_volume_scheduling_duration_seconds_bucket{cluster=\"$cluster\", job=\"kube-scheduler\",instance=~\"$instance\"}[5m])) by (instance, le))",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "{{instance}} volume",
@@ -31425,28 +31269,28 @@ items:
"steppedLine": false,
"targets": [
{
- "expr": "sum(rate(rest_client_requests_total{job=\"kube-scheduler\", instance=~\"$instance\",code=~\"2..\"}[5m]))",
+ "expr": "sum(rate(rest_client_requests_total{cluster=\"$cluster\", job=\"kube-scheduler\", instance=~\"$instance\",code=~\"2..\"}[5m]))",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "2xx",
"refId": "A"
},
{
- "expr": "sum(rate(rest_client_requests_total{job=\"kube-scheduler\", instance=~\"$instance\",code=~\"3..\"}[5m]))",
+ "expr": "sum(rate(rest_client_requests_total{cluster=\"$cluster\", job=\"kube-scheduler\", instance=~\"$instance\",code=~\"3..\"}[5m]))",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "3xx",
"refId": "B"
},
{
- "expr": "sum(rate(rest_client_requests_total{job=\"kube-scheduler\", instance=~\"$instance\",code=~\"4..\"}[5m]))",
+ "expr": "sum(rate(rest_client_requests_total{cluster=\"$cluster\", job=\"kube-scheduler\", instance=~\"$instance\",code=~\"4..\"}[5m]))",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "4xx",
"refId": "C"
},
{
- "expr": "sum(rate(rest_client_requests_total{job=\"kube-scheduler\", instance=~\"$instance\",code=~\"5..\"}[5m]))",
+ "expr": "sum(rate(rest_client_requests_total{cluster=\"$cluster\", job=\"kube-scheduler\", instance=~\"$instance\",code=~\"5..\"}[5m]))",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "5xx",
@@ -31539,7 +31383,7 @@ items:
"steppedLine": false,
"targets": [
{
- "expr": "histogram_quantile(0.99, sum(rate(rest_client_request_duration_seconds_bucket{job=\"kube-scheduler\", instance=~\"$instance\", verb=\"POST\"}[5m])) by (verb, url, le))",
+ "expr": "histogram_quantile(0.99, sum(rate(rest_client_request_duration_seconds_bucket{cluster=\"$cluster\", job=\"kube-scheduler\", instance=~\"$instance\", verb=\"POST\"}[5m])) by (verb, url, le))",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "{{verb}} {{url}}",
@@ -31645,7 +31489,7 @@ items:
"steppedLine": false,
"targets": [
{
- "expr": "histogram_quantile(0.99, sum(rate(rest_client_request_duration_seconds_bucket{job=\"kube-scheduler\", instance=~\"$instance\", verb=\"GET\"}[5m])) by (verb, url, le))",
+ "expr": "histogram_quantile(0.99, sum(rate(rest_client_request_duration_seconds_bucket{cluster=\"$cluster\", job=\"kube-scheduler\", instance=~\"$instance\", verb=\"GET\"}[5m])) by (verb, url, le))",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "{{verb}} {{url}}",
@@ -31751,7 +31595,7 @@ items:
"steppedLine": false,
"targets": [
{
- "expr": "process_resident_memory_bytes{job=\"kube-scheduler\", instance=~\"$instance\"}",
+ "expr": "process_resident_memory_bytes{cluster=\"$cluster\", job=\"kube-scheduler\", instance=~\"$instance\"}",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "{{instance}}",
@@ -31844,7 +31688,7 @@ items:
"steppedLine": false,
"targets": [
{
- "expr": "rate(process_cpu_seconds_total{job=\"kube-scheduler\", instance=~\"$instance\"}[5m])",
+ "expr": "rate(process_cpu_seconds_total{cluster=\"$cluster\", job=\"kube-scheduler\", instance=~\"$instance\"}[5m])",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "{{instance}}",
@@ -31937,7 +31781,7 @@ items:
"steppedLine": false,
"targets": [
{
- "expr": "go_goroutines{job=\"kube-scheduler\",instance=~\"$instance\"}",
+ "expr": "go_goroutines{cluster=\"$cluster\", job=\"kube-scheduler\",instance=~\"$instance\"}",
"format": "time_series",
"intervalFactor": 2,
"legendFormat": "{{instance}}",
@@ -32021,6 +31865,32 @@ items:
"allValue": null,
"current": {
+ },
+ "datasource": "$datasource",
+ "hide": 2,
+ "includeAll": false,
+ "label": "cluster",
+ "multi": false,
+ "name": "cluster",
+ "options": [
+
+ ],
+ "query": "label_values(kube_pod_info, cluster)",
+ "refresh": 2,
+ "regex": "",
+ "sort": 1,
+ "tagValuesQuery": "",
+ "tags": [
+
+ ],
+ "tagsQuery": "",
+ "type": "query",
+ "useTags": false
+ },
+ {
+ "allValue": null,
+ "current": {
+
},
"datasource": "$datasource",
"hide": 0,
@@ -32031,7 +31901,7 @@ items:
"options": [
],
- "query": "label_values(process_cpu_seconds_total{job=\"kube-scheduler\"}, instance)",
+ "query": "label_values(process_cpu_seconds_total{cluster=\"$cluster\", job=\"kube-scheduler\"}, instance)",
"refresh": 2,
"regex": "",
"sort": 1,
diff --git a/manifests/kubernetes-prometheusRule.yaml b/manifests/kubernetes-prometheusRule.yaml
index 4dc01f2f49d9561a9ae3d4b54c0a5617eb50e0a1..968f5f6142d30ff9dba3a59cdc73ea9e7b84aa19 100644
--- a/manifests/kubernetes-prometheusRule.yaml
+++ b/manifests/kubernetes-prometheusRule.yaml
@@ -61,7 +61,7 @@ spec:
!=
kube_deployment_status_replicas_available{job="kube-state-metrics"}
) and (
- changes(kube_deployment_status_replicas_updated{job="kube-state-metrics"}[5m])
+ changes(kube_deployment_status_replicas_updated{job="kube-state-metrics"}[10m])
==
0
)
@@ -79,7 +79,7 @@ spec:
!=
kube_statefulset_status_replicas{job="kube-state-metrics"}
) and (
- changes(kube_statefulset_status_replicas_updated{job="kube-state-metrics"}[5m])
+ changes(kube_statefulset_status_replicas_updated{job="kube-state-metrics"}[10m])
==
0
)
@@ -401,7 +401,7 @@ spec:
runbook_url: https://github.com/prometheus-operator/kube-prometheus/wiki/kubeversionmismatch
summary: Different semantic versions of Kubernetes components running.
expr: |
- count(count by (gitVersion) (label_replace(kubernetes_build_info{job!~"kube-dns|coredns"},"gitVersion","$1","gitVersion","(v[0-9]*.[0-9]*).*"))) > 1
+ count(count by (git_version) (label_replace(kubernetes_build_info{job!~"kube-dns|coredns"},"git_version","$1","git_version","(v[0-9]*.[0-9]*).*"))) > 1
for: 15m
labels:
severity: warning
@@ -1199,7 +1199,7 @@ spec:
rules:
- expr: |
sum by (cluster, namespace, pod, container) (
- rate(container_cpu_usage_seconds_total{job="kubelet", metrics_path="/metrics/cadvisor", image!="", container!="POD"}[5m])
+ rate(container_cpu_usage_seconds_total{job="kubelet", metrics_path="/metrics/cadvisor", image!=""}[5m])
) * on (cluster, namespace, pod) group_left(node) topk by (cluster, namespace, pod) (
1, max by(cluster, namespace, pod, node) (kube_pod_info{node!=""})
)
@@ -1229,9 +1229,9 @@ spec:
)
record: node_namespace_pod_container:container_memory_swap
- expr: |
- sum by (namespace) (
- sum by (namespace, pod) (
- max by (namespace, pod, container) (
+ sum by (namespace, cluster) (
+ sum by (namespace, pod, cluster) (
+ max by (namespace, pod, container, cluster) (
kube_pod_container_resource_requests_memory_bytes{job="kube-state-metrics"}
) * on(namespace, pod) group_left() max by (namespace, pod) (
kube_pod_status_phase{phase=~"Pending|Running"} == 1
@@ -1240,9 +1240,9 @@ spec:
)
record: namespace:kube_pod_container_resource_requests_memory_bytes:sum
- expr: |
- sum by (namespace) (
- sum by (namespace, pod) (
- max by (namespace, pod, container) (
+ sum by (namespace, cluster) (
+ sum by (namespace, pod, cluster) (
+ max by (namespace, pod, container, cluster) (
kube_pod_container_resource_requests_cpu_cores{job="kube-state-metrics"}
) * on(namespace, pod) group_left() max by (namespace, pod) (
kube_pod_status_phase{phase=~"Pending|Running"} == 1
@@ -1346,7 +1346,7 @@ spec:
count by (cluster, node) (sum by (node, cpu) (
node_cpu_seconds_total{job="node-exporter"}
* on (namespace, pod) group_left(node)
- node_namespace_pod:kube_pod_info:
+ topk by(namespace, pod) (1, node_namespace_pod:kube_pod_info:)
))
record: node:node_num_cpu:sum
- expr: |