diff --git a/manifests/prometheus/prometheus-k8s-rules.yaml b/manifests/prometheus/prometheus-k8s-rules.yaml
index 14284560ced4093f7ce3f28b5d4d4630f24222f7..cb062db18e6fd87199ffe409df1a493e8023becc 100644
--- a/manifests/prometheus/prometheus-k8s-rules.yaml
+++ b/manifests/prometheus/prometheus-k8s-rules.yaml
@@ -225,14 +225,14 @@ data:
### Up Alerting ###
Alert TargetDown
- IF up == 0
+ IF 100 * (count(up == 0) / count(up)) > 3
FOR 10m
LABELS {
severity = "warning"
}
ANNOTATIONS {
- summary = "target is down",
- description = "A target of type {{ $labels.job }} is down."
+ summary = "Targets are down",
+ description = "More than {{ $value }}% of targets are down."
}
### Dead man's switch ###
@@ -249,26 +249,15 @@ data:
### File descriptor alerts ###
- ALERT TooManyOpenFiles
- IF 100*process_open_fds / process_max_fds > 50
- FOR 10m
- LABELS {
- severity = "warning"
- }
- ANNOTATIONS {
- summary = "too many open file descriptors",
- description = "{{ $labels.job }}: {{ $labels.namespace }}/{{ $labels.pod }} is using {{ $value }}% of the available file/socket descriptors.",
- }
-
- ALERT K8STooManyOpenFiles
- IF 100*process_open_fds / process_max_fds > 80
+ ALERT TooManyOpenFileDescriptors
+ IF 100 * (process_open_fds / process_max_fds) > 95
FOR 10m
LABELS {
severity = "critical"
}
ANNOTATIONS {
summary = "too many open file descriptors",
- description = "{{ $labels.job }}: {{ $labels.namespace }}/{{ $labels.pod }} is using {{ $value }}% of the available file/socket descriptors.",
+ description = "{{ $labels.job }}: {{ $labels.namespace }}/{{ $labels.pod }} ({{ $labels.instance }}) is using {{ $value }}% of the available file/socket descriptors.",
}
instance:fd_utilization = process_open_fds / process_max_fds
@@ -282,7 +271,7 @@ data:
}
ANNOTATIONS {
summary = "file descriptors soon exhausted",
- description = "{{ $labels.job }}: {{ $labels.namespace }}/{{ $labels.pod }} instance {{ $labels.namespace }}/{{ $labels.pod }} will exhaust in file descriptors soon",
+ description = "{{ $labels.job }}: {{ $labels.namespace }}/{{ $labels.pod }} ({{ $labels.instance }}) instance will exhaust in file/socket descriptors soon",
}
# alert if file descriptors are likely to exhaust within the next hour
@@ -294,56 +283,9 @@ data:
}
ANNOTATIONS {
summary = "file descriptors soon exhausted",
- description = "{{ $labels.job }}: {{ $labels.namespace }}/{{ $labels.pod }} instance {{ $labels.namespace }}/{{ $labels.pod }} will exhaust in file descriptors soon",
- }
-
- ### Contrack alerts ###
-
- # To catch the conntrack sysctl de-tuning when it happens
- ALERT ConntrackTuningMissing
- IF node_nf_conntrack_udp_timeout > 10
- FOR 10m
- LABELS {
- severity = "warning",
- }
- ANNOTATIONS {
- summary = "Node does not have the correct conntrack tunings",
- description = "Nodes keep un-setting the correct tunings, investigate when it happens.",
- }
-
- ALERT ConntrackTableFull
- IF 100*node_nf_conntrack_entries / node_nf_conntrack_entries_limit > 50
- FOR 10m
- LABELS {
- severity = "warning"
- }
- ANNOTATIONS {
- summary = "Number of tracked connections is near the limit",
- description = "The nf_conntrack table is {{ $value }}% full.",
- }
-
- ALERT ConntrackTableFull
- IF 100*node_nf_conntrack_entries / node_nf_conntrack_entries_limit > 90
- LABELS {
- severity = "critical"
- }
- ANNOTATIONS {
- summary = "Number of tracked connections is near the limit",
- description = "The nf_conntrack table is {{ $value }}% full.",
+ description = "{{ $labels.job }}: {{ $labels.namespace }}/{{ $labels.pod }} ({{ $labels.instance }}) instance will exhaust in file/socket descriptors soon",
}
kube-apiserver.rules: |+
- ALERT K8SApiserverDown
- IF up{job="apiserver"} == 0
- FOR 15m
- LABELS {
- severity = "warning"
- }
- ANNOTATIONS {
- summary = "API server unreachable",
- description = "An API server could not be scraped.",
- }
-
- # Disable for non HA kubernetes setups.
ALERT K8SApiserverDown
IF absent({job="apiserver"}) or (count by(cluster) (up{job="apiserver"} == 1) < count by(cluster) (up{job="apiserver"}))
FOR 5m
@@ -352,11 +294,13 @@ data:
}
ANNOTATIONS {
summary = "API server unreachable",
- description = "Prometheus failed to scrape multiple API servers, or all API servers have disappeared from service discovery.",
+ description = "Prometheus failed to scrape API server(s), or all API servers have disappeared from service discovery.",
}
# Some verbs excluded because they are expected to be long-lasting:
# WATCHLIST is long-poll, CONNECT is `kubectl exec`.
+ #
+ # apiserver_request_latencies' unit is microseconds
ALERT K8SApiServerLatency
IF histogram_quantile(
0.99,
@@ -382,17 +326,6 @@ data:
description = "There is no running K8S controller manager. Deployments and replication controllers are not making progress.",
}
kubelet.rules: |+
- ALERT K8SNodeDown
- IF up{job="kubelet"} == 0
- FOR 1h
- LABELS {
- severity = "warning"
- }
- ANNOTATIONS {
- summary = "Kubelet cannot be scraped",
- description = "Prometheus could not scrape a {{ $labels.job }} for more than one hour",
- }
-
ALERT K8SNodeNotReady
IF kube_node_status_ready{condition="true"} == 0
FOR 1h
@@ -422,16 +355,26 @@ data:
description = "{{ $value }} K8s nodes (more than 10% of cluster {{ $labels.cluster }}) are in the NotReady state.",
}
+ ALERT K8SKubeletDown
+ IF count by (cluster) (up{job="kubelet"} == 0) / count by (cluster) (up{job="kubelet"}) > 0.03
+ FOR 1h
+ LABELS {
+ severity = "warning",
+ }
+ ANNOTATIONS {
+ summary = "Many Kubelets cannot be scraped",
+ description = "Prometheus failed to scrape {{ $value }}% of kubelets.",
+ }
+
ALERT K8SKubeletDown
IF absent(up{job="kubelet"}) or count by (cluster) (up{job="kubelet"} == 0) / count by (cluster) (up{job="kubelet"}) > 0.1
FOR 1h
LABELS {
- service = "k8s",
- severity = "critical"
+ severity = "critical",
}
ANNOTATIONS {
summary = "Many Kubelets cannot be scraped",
- description = "Prometheus failed to scrape more than 10% of kubelets, or all Kubelets have disappeared from service discovery.",
+ description = "Prometheus failed to scrape {{ $value }}% of kubelets, or all Kubelets have disappeared from service discovery.",
}
ALERT K8SKubeletTooManyPods
@@ -625,7 +568,6 @@ data:
ANNOTATIONS {
summary = "Scheduler is down",
description = "There is no running K8S scheduler. New pods are not being assigned to nodes.",
- runbook = "https://github.com/coreos/tectonic-installer/blob/master/Documentation/troubleshooting/controller-recovery.md#disaster-recovery-of-scheduler-and-controller-manager-pods"
}
node.rules: |+
ALERT NodeExporterDown