easylinux/kube-prometheus
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

manifests: regenerate

Browse Source
This commit is contained in:
paulfantom
2021-05-25 16:03:49 +02:00
parent ce7e86b93a
commit 888443e447
6 changed files with 1841 additions and 1064 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2628
manifests/grafana-dashboardDefinitions.yaml
View File

File diff suppressed because it is too large Load Diff

6
manifests/grafana-deployment.yaml
View File

@@ -116,9 +116,6 @@ spec:
- mountPath: /grafana-dashboard-definitions/0/scheduler - mountPath: /grafana-dashboard-definitions/0/scheduler
name: grafana-dashboard-scheduler name: grafana-dashboard-scheduler
readOnly: false readOnly: false
- mountPath: /grafana-dashboard-definitions/0/statefulset
name: grafana-dashboard-statefulset
readOnly: false
- mountPath: /grafana-dashboard-definitions/0/workload-total - mountPath: /grafana-dashboard-definitions/0/workload-total
name: grafana-dashboard-workload-total name: grafana-dashboard-workload-total
readOnly: false readOnly: false
@@ -201,9 +198,6 @@ spec:
- configMap: - configMap:
name: grafana-dashboard-scheduler name: grafana-dashboard-scheduler
name: grafana-dashboard-scheduler name: grafana-dashboard-scheduler
- configMap:
name: grafana-dashboard-statefulset
name: grafana-dashboard-statefulset
- configMap: - configMap:
name: grafana-dashboard-workload-total name: grafana-dashboard-workload-total
name: grafana-dashboard-workload-total name: grafana-dashboard-workload-total

23
manifests/kube-state-metrics-prometheusRule.yaml
View File

@@ -40,3 +40,26 @@ spec:
for: 15m for: 15m
labels: labels:
severity: critical severity: critical
- alert: KubeStateMetricsShardingMismatch
annotations:
description: kube-state-metrics pods are running with different --total-shards configuration, some Kubernetes objects may be exposed multiple times or not exposed at all.
runbook_url: https://github.com/prometheus-operator/kube-prometheus/wiki/kubestatemetricsshardingmismatch
summary: kube-state-metrics sharding is misconfigured.
expr: |
stdvar (kube_state_metrics_total_shards{job="kube-state-metrics"}) != 0
for: 15m
labels:
severity: critical
- alert: KubeStateMetricsShardsMissing
annotations:
description: kube-state-metrics shards are missing, some Kubernetes objects are not being exposed.
runbook_url: https://github.com/prometheus-operator/kube-prometheus/wiki/kubestatemetricsshardsmissing
summary: kube-state-metrics shards are missing.
expr: |
2^max(kube_state_metrics_total_shards{job="kube-state-metrics"}) - 1
-
sum( 2 ^ max by (shard_ordinal) (kube_state_metrics_shard_ordinal{job="kube-state-metrics"}) )
!= 0
for: 15m
labels:
severity: critical

236
manifests/kubernetes-prometheusRule.yaml
View File

@@ -699,26 +699,26 @@ spec:
( (
( (
# too slow # too slow
sum(rate(apiserver_request_duration_seconds_count{job="apiserver",verb=~"LIST|GET"}[1d])) sum by (cluster) (rate(apiserver_request_duration_seconds_count{job="apiserver",verb=~"LIST|GET"}[1d]))
- -
( (
( (
sum(rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"LIST|GET",scope=~"resource|",le="0.1"}[1d])) sum by (cluster) (rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"LIST|GET",scope=~"resource|",le="0.1"}[1d]))
or or
vector(0) vector(0)
) )
+ +
sum(rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"LIST|GET",scope="namespace",le="0.5"}[1d])) sum by (cluster) (rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"LIST|GET",scope="namespace",le="0.5"}[1d]))
+ +
sum(rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"LIST|GET",scope="cluster",le="5"}[1d])) sum by (cluster) (rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"LIST|GET",scope="cluster",le="5"}[1d]))
) )
) )
+ +
# errors # errors
sum(rate(apiserver_request_total{job="apiserver",verb=~"LIST|GET",code=~"5.."}[1d])) sum by (cluster) (rate(apiserver_request_total{job="apiserver",verb=~"LIST|GET",code=~"5.."}[1d]))
) )
/ /
sum(rate(apiserver_request_total{job="apiserver",verb=~"LIST|GET"}[1d])) sum by (cluster) (rate(apiserver_request_total{job="apiserver",verb=~"LIST|GET"}[1d]))
labels: labels:
verb: read verb: read
record: apiserver_request:burnrate1d record: apiserver_request:burnrate1d
@@ -726,26 +726,26 @@ spec:
( (
( (
# too slow # too slow
sum(rate(apiserver_request_duration_seconds_count{job="apiserver",verb=~"LIST|GET"}[1h])) sum by (cluster) (rate(apiserver_request_duration_seconds_count{job="apiserver",verb=~"LIST|GET"}[1h]))
- -
( (
( (
sum(rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"LIST|GET",scope=~"resource|",le="0.1"}[1h])) sum by (cluster) (rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"LIST|GET",scope=~"resource|",le="0.1"}[1h]))
or or
vector(0) vector(0)
) )
+ +
sum(rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"LIST|GET",scope="namespace",le="0.5"}[1h])) sum by (cluster) (rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"LIST|GET",scope="namespace",le="0.5"}[1h]))
+ +
sum(rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"LIST|GET",scope="cluster",le="5"}[1h])) sum by (cluster) (rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"LIST|GET",scope="cluster",le="5"}[1h]))
) )
) )
+ +
# errors # errors
sum(rate(apiserver_request_total{job="apiserver",verb=~"LIST|GET",code=~"5.."}[1h])) sum by (cluster) (rate(apiserver_request_total{job="apiserver",verb=~"LIST|GET",code=~"5.."}[1h]))
) )
/ /
sum(rate(apiserver_request_total{job="apiserver",verb=~"LIST|GET"}[1h])) sum by (cluster) (rate(apiserver_request_total{job="apiserver",verb=~"LIST|GET"}[1h]))
labels: labels:
verb: read verb: read
record: apiserver_request:burnrate1h record: apiserver_request:burnrate1h
@@ -753,26 +753,26 @@ spec:
( (
( (
# too slow # too slow
sum(rate(apiserver_request_duration_seconds_count{job="apiserver",verb=~"LIST|GET"}[2h])) sum by (cluster) (rate(apiserver_request_duration_seconds_count{job="apiserver",verb=~"LIST|GET"}[2h]))
- -
( (
( (
sum(rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"LIST|GET",scope=~"resource|",le="0.1"}[2h])) sum by (cluster) (rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"LIST|GET",scope=~"resource|",le="0.1"}[2h]))
or or
vector(0) vector(0)
) )
+ +
sum(rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"LIST|GET",scope="namespace",le="0.5"}[2h])) sum by (cluster) (rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"LIST|GET",scope="namespace",le="0.5"}[2h]))
+ +
sum(rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"LIST|GET",scope="cluster",le="5"}[2h])) sum by (cluster) (rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"LIST|GET",scope="cluster",le="5"}[2h]))
) )
) )
+ +
# errors # errors
sum(rate(apiserver_request_total{job="apiserver",verb=~"LIST|GET",code=~"5.."}[2h])) sum by (cluster) (rate(apiserver_request_total{job="apiserver",verb=~"LIST|GET",code=~"5.."}[2h]))
) )
/ /
sum(rate(apiserver_request_total{job="apiserver",verb=~"LIST|GET"}[2h])) sum by (cluster) (rate(apiserver_request_total{job="apiserver",verb=~"LIST|GET"}[2h]))
labels: labels:
verb: read verb: read
record: apiserver_request:burnrate2h record: apiserver_request:burnrate2h
@@ -780,26 +780,26 @@ spec:
( (
( (
# too slow # too slow
sum(rate(apiserver_request_duration_seconds_count{job="apiserver",verb=~"LIST|GET"}[30m])) sum by (cluster) (rate(apiserver_request_duration_seconds_count{job="apiserver",verb=~"LIST|GET"}[30m]))
- -
( (
( (
sum(rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"LIST|GET",scope=~"resource|",le="0.1"}[30m])) sum by (cluster) (rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"LIST|GET",scope=~"resource|",le="0.1"}[30m]))
or or
vector(0) vector(0)
) )
+ +
sum(rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"LIST|GET",scope="namespace",le="0.5"}[30m])) sum by (cluster) (rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"LIST|GET",scope="namespace",le="0.5"}[30m]))
+ +
sum(rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"LIST|GET",scope="cluster",le="5"}[30m])) sum by (cluster) (rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"LIST|GET",scope="cluster",le="5"}[30m]))
) )
) )
+ +
# errors # errors
sum(rate(apiserver_request_total{job="apiserver",verb=~"LIST|GET",code=~"5.."}[30m])) sum by (cluster) (rate(apiserver_request_total{job="apiserver",verb=~"LIST|GET",code=~"5.."}[30m]))
) )
/ /
sum(rate(apiserver_request_total{job="apiserver",verb=~"LIST|GET"}[30m])) sum by (cluster) (rate(apiserver_request_total{job="apiserver",verb=~"LIST|GET"}[30m]))
labels: labels:
verb: read verb: read
record: apiserver_request:burnrate30m record: apiserver_request:burnrate30m
@@ -807,26 +807,26 @@ spec:
( (
( (
# too slow # too slow
sum(rate(apiserver_request_duration_seconds_count{job="apiserver",verb=~"LIST|GET"}[3d])) sum by (cluster) (rate(apiserver_request_duration_seconds_count{job="apiserver",verb=~"LIST|GET"}[3d]))
- -
( (
( (
sum(rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"LIST|GET",scope=~"resource|",le="0.1"}[3d])) sum by (cluster) (rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"LIST|GET",scope=~"resource|",le="0.1"}[3d]))
or or
vector(0) vector(0)
) )
+ +
sum(rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"LIST|GET",scope="namespace",le="0.5"}[3d])) sum by (cluster) (rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"LIST|GET",scope="namespace",le="0.5"}[3d]))
+ +
sum(rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"LIST|GET",scope="cluster",le="5"}[3d])) sum by (cluster) (rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"LIST|GET",scope="cluster",le="5"}[3d]))
) )
) )
+ +
# errors # errors
sum(rate(apiserver_request_total{job="apiserver",verb=~"LIST|GET",code=~"5.."}[3d])) sum by (cluster) (rate(apiserver_request_total{job="apiserver",verb=~"LIST|GET",code=~"5.."}[3d]))
) )
/ /
sum(rate(apiserver_request_total{job="apiserver",verb=~"LIST|GET"}[3d])) sum by (cluster) (rate(apiserver_request_total{job="apiserver",verb=~"LIST|GET"}[3d]))
labels: labels:
verb: read verb: read
record: apiserver_request:burnrate3d record: apiserver_request:burnrate3d
@@ -834,26 +834,26 @@ spec:
( (
( (
# too slow # too slow
sum(rate(apiserver_request_duration_seconds_count{job="apiserver",verb=~"LIST|GET"}[5m])) sum by (cluster) (rate(apiserver_request_duration_seconds_count{job="apiserver",verb=~"LIST|GET"}[5m]))
- -
( (
( (
sum(rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"LIST|GET",scope=~"resource|",le="0.1"}[5m])) sum by (cluster) (rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"LIST|GET",scope=~"resource|",le="0.1"}[5m]))
or or
vector(0) vector(0)
) )
+ +
sum(rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"LIST|GET",scope="namespace",le="0.5"}[5m])) sum by (cluster) (rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"LIST|GET",scope="namespace",le="0.5"}[5m]))
+ +
sum(rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"LIST|GET",scope="cluster",le="5"}[5m])) sum by (cluster) (rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"LIST|GET",scope="cluster",le="5"}[5m]))
) )
) )
+ +
# errors # errors
sum(rate(apiserver_request_total{job="apiserver",verb=~"LIST|GET",code=~"5.."}[5m])) sum by (cluster) (rate(apiserver_request_total{job="apiserver",verb=~"LIST|GET",code=~"5.."}[5m]))
) )
/ /
sum(rate(apiserver_request_total{job="apiserver",verb=~"LIST|GET"}[5m])) sum by (cluster) (rate(apiserver_request_total{job="apiserver",verb=~"LIST|GET"}[5m]))
labels: labels:
verb: read verb: read
record: apiserver_request:burnrate5m record: apiserver_request:burnrate5m
@@ -861,26 +861,26 @@ spec:
( (
( (
# too slow # too slow
sum(rate(apiserver_request_duration_seconds_count{job="apiserver",verb=~"LIST|GET"}[6h])) sum by (cluster) (rate(apiserver_request_duration_seconds_count{job="apiserver",verb=~"LIST|GET"}[6h]))
- -
( (
( (
sum(rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"LIST|GET",scope=~"resource|",le="0.1"}[6h])) sum by (cluster) (rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"LIST|GET",scope=~"resource|",le="0.1"}[6h]))
or or
vector(0) vector(0)
) )
+ +
sum(rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"LIST|GET",scope="namespace",le="0.5"}[6h])) sum by (cluster) (rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"LIST|GET",scope="namespace",le="0.5"}[6h]))
+ +
sum(rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"LIST|GET",scope="cluster",le="5"}[6h])) sum by (cluster) (rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"LIST|GET",scope="cluster",le="5"}[6h]))
) )
) )
+ +
# errors # errors
sum(rate(apiserver_request_total{job="apiserver",verb=~"LIST|GET",code=~"5.."}[6h])) sum by (cluster) (rate(apiserver_request_total{job="apiserver",verb=~"LIST|GET",code=~"5.."}[6h]))
) )
/ /
sum(rate(apiserver_request_total{job="apiserver",verb=~"LIST|GET"}[6h])) sum by (cluster) (rate(apiserver_request_total{job="apiserver",verb=~"LIST|GET"}[6h]))
labels: labels:
verb: read verb: read
record: apiserver_request:burnrate6h record: apiserver_request:burnrate6h
@@ -888,15 +888,15 @@ spec:
( (
( (
# too slow # too slow
sum(rate(apiserver_request_duration_seconds_count{job="apiserver",verb=~"POST|PUT|PATCH|DELETE"}[1d])) sum by (cluster) (rate(apiserver_request_duration_seconds_count{job="apiserver",verb=~"POST|PUT|PATCH|DELETE"}[1d]))
- -
sum(rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"POST|PUT|PATCH|DELETE",le="1"}[1d])) sum by (cluster) (rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"POST|PUT|PATCH|DELETE",le="1"}[1d]))
) )
+ +
sum(rate(apiserver_request_total{job="apiserver",verb=~"POST|PUT|PATCH|DELETE",code=~"5.."}[1d])) sum by (cluster) (rate(apiserver_request_total{job="apiserver",verb=~"POST|PUT|PATCH|DELETE",code=~"5.."}[1d]))
) )
/ /
sum(rate(apiserver_request_total{job="apiserver",verb=~"POST|PUT|PATCH|DELETE"}[1d])) sum by (cluster) (rate(apiserver_request_total{job="apiserver",verb=~"POST|PUT|PATCH|DELETE"}[1d]))
labels: labels:
verb: write verb: write
record: apiserver_request:burnrate1d record: apiserver_request:burnrate1d
@@ -904,15 +904,15 @@ spec:
( (
( (
# too slow # too slow
sum(rate(apiserver_request_duration_seconds_count{job="apiserver",verb=~"POST|PUT|PATCH|DELETE"}[1h])) sum by (cluster) (rate(apiserver_request_duration_seconds_count{job="apiserver",verb=~"POST|PUT|PATCH|DELETE"}[1h]))
- -
sum(rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"POST|PUT|PATCH|DELETE",le="1"}[1h])) sum by (cluster) (rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"POST|PUT|PATCH|DELETE",le="1"}[1h]))
) )
+ +
sum(rate(apiserver_request_total{job="apiserver",verb=~"POST|PUT|PATCH|DELETE",code=~"5.."}[1h])) sum by (cluster) (rate(apiserver_request_total{job="apiserver",verb=~"POST|PUT|PATCH|DELETE",code=~"5.."}[1h]))
) )
/ /
sum(rate(apiserver_request_total{job="apiserver",verb=~"POST|PUT|PATCH|DELETE"}[1h])) sum by (cluster) (rate(apiserver_request_total{job="apiserver",verb=~"POST|PUT|PATCH|DELETE"}[1h]))
labels: labels:
verb: write verb: write
record: apiserver_request:burnrate1h record: apiserver_request:burnrate1h
@@ -920,15 +920,15 @@ spec:
( (
( (
# too slow # too slow
sum(rate(apiserver_request_duration_seconds_count{job="apiserver",verb=~"POST|PUT|PATCH|DELETE"}[2h])) sum by (cluster) (rate(apiserver_request_duration_seconds_count{job="apiserver",verb=~"POST|PUT|PATCH|DELETE"}[2h]))
- -
sum(rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"POST|PUT|PATCH|DELETE",le="1"}[2h])) sum by (cluster) (rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"POST|PUT|PATCH|DELETE",le="1"}[2h]))
) )
+ +
sum(rate(apiserver_request_total{job="apiserver",verb=~"POST|PUT|PATCH|DELETE",code=~"5.."}[2h])) sum by (cluster) (rate(apiserver_request_total{job="apiserver",verb=~"POST|PUT|PATCH|DELETE",code=~"5.."}[2h]))
) )
/ /
sum(rate(apiserver_request_total{job="apiserver",verb=~"POST|PUT|PATCH|DELETE"}[2h])) sum by (cluster) (rate(apiserver_request_total{job="apiserver",verb=~"POST|PUT|PATCH|DELETE"}[2h]))
labels: labels:
verb: write verb: write
record: apiserver_request:burnrate2h record: apiserver_request:burnrate2h
@@ -936,15 +936,15 @@ spec:
( (
( (
# too slow # too slow
sum(rate(apiserver_request_duration_seconds_count{job="apiserver",verb=~"POST|PUT|PATCH|DELETE"}[30m])) sum by (cluster) (rate(apiserver_request_duration_seconds_count{job="apiserver",verb=~"POST|PUT|PATCH|DELETE"}[30m]))
- -
sum(rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"POST|PUT|PATCH|DELETE",le="1"}[30m])) sum by (cluster) (rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"POST|PUT|PATCH|DELETE",le="1"}[30m]))
) )
+ +
sum(rate(apiserver_request_total{job="apiserver",verb=~"POST|PUT|PATCH|DELETE",code=~"5.."}[30m])) sum by (cluster) (rate(apiserver_request_total{job="apiserver",verb=~"POST|PUT|PATCH|DELETE",code=~"5.."}[30m]))
) )
/ /
sum(rate(apiserver_request_total{job="apiserver",verb=~"POST|PUT|PATCH|DELETE"}[30m])) sum by (cluster) (rate(apiserver_request_total{job="apiserver",verb=~"POST|PUT|PATCH|DELETE"}[30m]))
labels: labels:
verb: write verb: write
record: apiserver_request:burnrate30m record: apiserver_request:burnrate30m
@@ -952,15 +952,15 @@ spec:
( (
( (
# too slow # too slow
sum(rate(apiserver_request_duration_seconds_count{job="apiserver",verb=~"POST|PUT|PATCH|DELETE"}[3d])) sum by (cluster) (rate(apiserver_request_duration_seconds_count{job="apiserver",verb=~"POST|PUT|PATCH|DELETE"}[3d]))
- -
sum(rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"POST|PUT|PATCH|DELETE",le="1"}[3d])) sum by (cluster) (rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"POST|PUT|PATCH|DELETE",le="1"}[3d]))
) )
+ +
sum(rate(apiserver_request_total{job="apiserver",verb=~"POST|PUT|PATCH|DELETE",code=~"5.."}[3d])) sum by (cluster) (rate(apiserver_request_total{job="apiserver",verb=~"POST|PUT|PATCH|DELETE",code=~"5.."}[3d]))
) )
/ /
sum(rate(apiserver_request_total{job="apiserver",verb=~"POST|PUT|PATCH|DELETE"}[3d])) sum by (cluster) (rate(apiserver_request_total{job="apiserver",verb=~"POST|PUT|PATCH|DELETE"}[3d]))
labels: labels:
verb: write verb: write
record: apiserver_request:burnrate3d record: apiserver_request:burnrate3d
@@ -968,15 +968,15 @@ spec:
( (
( (
# too slow # too slow
sum(rate(apiserver_request_duration_seconds_count{job="apiserver",verb=~"POST|PUT|PATCH|DELETE"}[5m])) sum by (cluster) (rate(apiserver_request_duration_seconds_count{job="apiserver",verb=~"POST|PUT|PATCH|DELETE"}[5m]))
- -
sum(rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"POST|PUT|PATCH|DELETE",le="1"}[5m])) sum by (cluster) (rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"POST|PUT|PATCH|DELETE",le="1"}[5m]))
) )
+ +
sum(rate(apiserver_request_total{job="apiserver",verb=~"POST|PUT|PATCH|DELETE",code=~"5.."}[5m])) sum by (cluster) (rate(apiserver_request_total{job="apiserver",verb=~"POST|PUT|PATCH|DELETE",code=~"5.."}[5m]))
) )
/ /
sum(rate(apiserver_request_total{job="apiserver",verb=~"POST|PUT|PATCH|DELETE"}[5m])) sum by (cluster) (rate(apiserver_request_total{job="apiserver",verb=~"POST|PUT|PATCH|DELETE"}[5m]))
labels: labels:
verb: write verb: write
record: apiserver_request:burnrate5m record: apiserver_request:burnrate5m
@@ -984,36 +984,36 @@ spec:
( (
( (
# too slow # too slow
sum(rate(apiserver_request_duration_seconds_count{job="apiserver",verb=~"POST|PUT|PATCH|DELETE"}[6h])) sum by (cluster) (rate(apiserver_request_duration_seconds_count{job="apiserver",verb=~"POST|PUT|PATCH|DELETE"}[6h]))
- -
sum(rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"POST|PUT|PATCH|DELETE",le="1"}[6h])) sum by (cluster) (rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"POST|PUT|PATCH|DELETE",le="1"}[6h]))
) )
+ +
sum(rate(apiserver_request_total{job="apiserver",verb=~"POST|PUT|PATCH|DELETE",code=~"5.."}[6h])) sum by (cluster) (rate(apiserver_request_total{job="apiserver",verb=~"POST|PUT|PATCH|DELETE",code=~"5.."}[6h]))
) )
/ /
sum(rate(apiserver_request_total{job="apiserver",verb=~"POST|PUT|PATCH|DELETE"}[6h])) sum by (cluster) (rate(apiserver_request_total{job="apiserver",verb=~"POST|PUT|PATCH|DELETE"}[6h]))
labels: labels:
verb: write verb: write
record: apiserver_request:burnrate6h record: apiserver_request:burnrate6h
- expr: | - expr: |
sum by (code,resource) (rate(apiserver_request_total{job="apiserver",verb=~"LIST|GET"}[5m])) sum by (cluster,code,resource) (rate(apiserver_request_total{job="apiserver",verb=~"LIST|GET"}[5m]))
labels: labels:
verb: read verb: read
record: code_resource:apiserver_request_total:rate5m record: code_resource:apiserver_request_total:rate5m
- expr: | - expr: |
sum by (code,resource) (rate(apiserver_request_total{job="apiserver",verb=~"POST|PUT|PATCH|DELETE"}[5m])) sum by (cluster,code,resource) (rate(apiserver_request_total{job="apiserver",verb=~"POST|PUT|PATCH|DELETE"}[5m]))
labels: labels:
verb: write verb: write
record: code_resource:apiserver_request_total:rate5m record: code_resource:apiserver_request_total:rate5m
- expr: | - expr: |
histogram_quantile(0.99, sum by (le, resource) (rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"LIST|GET"}[5m]))) > 0 histogram_quantile(0.99, sum by (cluster, le, resource) (rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"LIST|GET"}[5m]))) > 0
labels: labels:
quantile: "0.99" quantile: "0.99"
verb: read verb: read
record: cluster_quantile:apiserver_request_duration_seconds:histogram_quantile record: cluster_quantile:apiserver_request_duration_seconds:histogram_quantile
- expr: | - expr: |
histogram_quantile(0.99, sum by (le, resource) (rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"POST|PUT|PATCH|DELETE"}[5m]))) > 0 histogram_quantile(0.99, sum by (cluster, le, resource) (rate(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"POST|PUT|PATCH|DELETE"}[5m]))) > 0
labels: labels:
quantile: "0.99" quantile: "0.99"
verb: write verb: write
@@ -1040,56 +1040,56 @@ spec:
1 - ( 1 - (
( (
# write too slow # write too slow
sum(increase(apiserver_request_duration_seconds_count{verb=~"POST|PUT|PATCH|DELETE"}[30d])) sum by (cluster) (increase(apiserver_request_duration_seconds_count{verb=~"POST|PUT|PATCH|DELETE"}[30d]))
- -
sum(increase(apiserver_request_duration_seconds_bucket{verb=~"POST|PUT|PATCH|DELETE",le="1"}[30d])) sum by (cluster) (increase(apiserver_request_duration_seconds_bucket{verb=~"POST|PUT|PATCH|DELETE",le="1"}[30d]))
) + ) +
( (
# read too slow # read too slow
sum(increase(apiserver_request_duration_seconds_count{verb=~"LIST|GET"}[30d])) sum by (cluster) (increase(apiserver_request_duration_seconds_count{verb=~"LIST|GET"}[30d]))
- -
( (
( (
sum(increase(apiserver_request_duration_seconds_bucket{verb=~"LIST|GET",scope=~"resource|",le="0.1"}[30d])) sum by (cluster) (increase(apiserver_request_duration_seconds_bucket{verb=~"LIST|GET",scope=~"resource|",le="0.1"}[30d]))
or or
vector(0) vector(0)
) )
+ +
sum(increase(apiserver_request_duration_seconds_bucket{verb=~"LIST|GET",scope="namespace",le="0.5"}[30d])) sum by (cluster) (increase(apiserver_request_duration_seconds_bucket{verb=~"LIST|GET",scope="namespace",le="0.5"}[30d]))
+ +
sum(increase(apiserver_request_duration_seconds_bucket{verb=~"LIST|GET",scope="cluster",le="5"}[30d])) sum by (cluster) (increase(apiserver_request_duration_seconds_bucket{verb=~"LIST|GET",scope="cluster",le="5"}[30d]))
) )
) + ) +
# errors # errors
sum(code:apiserver_request_total:increase30d{code=~"5.."} or vector(0)) sum by (cluster) (code:apiserver_request_total:increase30d{code=~"5.."} or vector(0))
) )
/ /
sum(code:apiserver_request_total:increase30d) sum by (cluster) (code:apiserver_request_total:increase30d)
labels: labels:
verb: all verb: all
record: apiserver_request:availability30d record: apiserver_request:availability30d
- expr: | - expr: |
1 - ( 1 - (
sum(increase(apiserver_request_duration_seconds_count{job="apiserver",verb=~"LIST|GET"}[30d])) sum by (cluster) (increase(apiserver_request_duration_seconds_count{job="apiserver",verb=~"LIST|GET"}[30d]))
- -
( (
# too slow # too slow
( (
sum(increase(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"LIST|GET",scope=~"resource|",le="0.1"}[30d])) sum by (cluster) (increase(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"LIST|GET",scope=~"resource|",le="0.1"}[30d]))
or or
vector(0) vector(0)
) )
+ +
sum(increase(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"LIST|GET",scope="namespace",le="0.5"}[30d])) sum by (cluster) (increase(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"LIST|GET",scope="namespace",le="0.5"}[30d]))
+ +
sum(increase(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"LIST|GET",scope="cluster",le="5"}[30d])) sum by (cluster) (increase(apiserver_request_duration_seconds_bucket{job="apiserver",verb=~"LIST|GET",scope="cluster",le="5"}[30d]))
) )
+ +
# errors # errors
sum(code:apiserver_request_total:increase30d{verb="read",code=~"5.."} or vector(0)) sum by (cluster) (code:apiserver_request_total:increase30d{verb="read",code=~"5.."} or vector(0))
) )
/ /
sum(code:apiserver_request_total:increase30d{verb="read"}) sum by (cluster) (code:apiserver_request_total:increase30d{verb="read"})
labels: labels:
verb: read verb: read
record: apiserver_request:availability30d record: apiserver_request:availability30d
@@ -1097,16 +1097,16 @@ spec:
1 - ( 1 - (
( (
# too slow # too slow
sum(increase(apiserver_request_duration_seconds_count{verb=~"POST|PUT|PATCH|DELETE"}[30d])) sum by (cluster) (increase(apiserver_request_duration_seconds_count{verb=~"POST|PUT|PATCH|DELETE"}[30d]))
- -
sum(increase(apiserver_request_duration_seconds_bucket{verb=~"POST|PUT|PATCH|DELETE",le="1"}[30d])) sum by (cluster) (increase(apiserver_request_duration_seconds_bucket{verb=~"POST|PUT|PATCH|DELETE",le="1"}[30d]))
) )
+ +
# errors # errors
sum(code:apiserver_request_total:increase30d{verb="write",code=~"5.."} or vector(0)) sum by (cluster) (code:apiserver_request_total:increase30d{verb="write",code=~"5.."} or vector(0))
) )
/ /
sum(code:apiserver_request_total:increase30d{verb="write"}) sum by (cluster) (code:apiserver_request_total:increase30d{verb="write"})
labels: labels:
verb: write verb: write
record: apiserver_request:availability30d record: apiserver_request:availability30d
@@ -1114,84 +1114,84 @@ spec:
avg_over_time(code_verb:apiserver_request_total:increase1h[30d]) * 24 * 30 avg_over_time(code_verb:apiserver_request_total:increase1h[30d]) * 24 * 30
record: code_verb:apiserver_request_total:increase30d record: code_verb:apiserver_request_total:increase30d
- expr: | - expr: |
sum by (code, verb) (increase(apiserver_request_total{job="apiserver",verb="LIST",code=~"2.."}[1h])) sum by (cluster, code, verb) (increase(apiserver_request_total{job="apiserver",verb="LIST",code=~"2.."}[1h]))
record: code_verb:apiserver_request_total:increase1h record: code_verb:apiserver_request_total:increase1h
- expr: | - expr: |
sum by (code, verb) (increase(apiserver_request_total{job="apiserver",verb="GET",code=~"2.."}[1h])) sum by (cluster, code, verb) (increase(apiserver_request_total{job="apiserver",verb="GET",code=~"2.."}[1h]))
record: code_verb:apiserver_request_total:increase1h record: code_verb:apiserver_request_total:increase1h
- expr: | - expr: |
sum by (code, verb) (increase(apiserver_request_total{job="apiserver",verb="POST",code=~"2.."}[1h])) sum by (cluster, code, verb) (increase(apiserver_request_total{job="apiserver",verb="POST",code=~"2.."}[1h]))
record: code_verb:apiserver_request_total:increase1h record: code_verb:apiserver_request_total:increase1h
- expr: | - expr: |
sum by (code, verb) (increase(apiserver_request_total{job="apiserver",verb="PUT",code=~"2.."}[1h])) sum by (cluster, code, verb) (increase(apiserver_request_total{job="apiserver",verb="PUT",code=~"2.."}[1h]))
record: code_verb:apiserver_request_total:increase1h record: code_verb:apiserver_request_total:increase1h
- expr: | - expr: |
sum by (code, verb) (increase(apiserver_request_total{job="apiserver",verb="PATCH",code=~"2.."}[1h])) sum by (cluster, code, verb) (increase(apiserver_request_total{job="apiserver",verb="PATCH",code=~"2.."}[1h]))
record: code_verb:apiserver_request_total:increase1h record: code_verb:apiserver_request_total:increase1h
- expr: | - expr: |
sum by (code, verb) (increase(apiserver_request_total{job="apiserver",verb="DELETE",code=~"2.."}[1h])) sum by (cluster, code, verb) (increase(apiserver_request_total{job="apiserver",verb="DELETE",code=~"2.."}[1h]))
record: code_verb:apiserver_request_total:increase1h record: code_verb:apiserver_request_total:increase1h
- expr: | - expr: |
sum by (code, verb) (increase(apiserver_request_total{job="apiserver",verb="LIST",code=~"3.."}[1h])) sum by (cluster, code, verb) (increase(apiserver_request_total{job="apiserver",verb="LIST",code=~"3.."}[1h]))
record: code_verb:apiserver_request_total:increase1h record: code_verb:apiserver_request_total:increase1h
- expr: | - expr: |
sum by (code, verb) (increase(apiserver_request_total{job="apiserver",verb="GET",code=~"3.."}[1h])) sum by (cluster, code, verb) (increase(apiserver_request_total{job="apiserver",verb="GET",code=~"3.."}[1h]))
record: code_verb:apiserver_request_total:increase1h record: code_verb:apiserver_request_total:increase1h
- expr: | - expr: |
sum by (code, verb) (increase(apiserver_request_total{job="apiserver",verb="POST",code=~"3.."}[1h])) sum by (cluster, code, verb) (increase(apiserver_request_total{job="apiserver",verb="POST",code=~"3.."}[1h]))
record: code_verb:apiserver_request_total:increase1h record: code_verb:apiserver_request_total:increase1h
- expr: | - expr: |
sum by (code, verb) (increase(apiserver_request_total{job="apiserver",verb="PUT",code=~"3.."}[1h])) sum by (cluster, code, verb) (increase(apiserver_request_total{job="apiserver",verb="PUT",code=~"3.."}[1h]))
record: code_verb:apiserver_request_total:increase1h record: code_verb:apiserver_request_total:increase1h
- expr: | - expr: |
sum by (code, verb) (increase(apiserver_request_total{job="apiserver",verb="PATCH",code=~"3.."}[1h])) sum by (cluster, code, verb) (increase(apiserver_request_total{job="apiserver",verb="PATCH",code=~"3.."}[1h]))
record: code_verb:apiserver_request_total:increase1h record: code_verb:apiserver_request_total:increase1h
- expr: | - expr: |
sum by (code, verb) (increase(apiserver_request_total{job="apiserver",verb="DELETE",code=~"3.."}[1h])) sum by (cluster, code, verb) (increase(apiserver_request_total{job="apiserver",verb="DELETE",code=~"3.."}[1h]))
record: code_verb:apiserver_request_total:increase1h record: code_verb:apiserver_request_total:increase1h
- expr: | - expr: |
sum by (code, verb) (increase(apiserver_request_total{job="apiserver",verb="LIST",code=~"4.."}[1h])) sum by (cluster, code, verb) (increase(apiserver_request_total{job="apiserver",verb="LIST",code=~"4.."}[1h]))
record: code_verb:apiserver_request_total:increase1h record: code_verb:apiserver_request_total:increase1h
- expr: | - expr: |
sum by (code, verb) (increase(apiserver_request_total{job="apiserver",verb="GET",code=~"4.."}[1h])) sum by (cluster, code, verb) (increase(apiserver_request_total{job="apiserver",verb="GET",code=~"4.."}[1h]))
record: code_verb:apiserver_request_total:increase1h record: code_verb:apiserver_request_total:increase1h
- expr: | - expr: |
sum by (code, verb) (increase(apiserver_request_total{job="apiserver",verb="POST",code=~"4.."}[1h])) sum by (cluster, code, verb) (increase(apiserver_request_total{job="apiserver",verb="POST",code=~"4.."}[1h]))
record: code_verb:apiserver_request_total:increase1h record: code_verb:apiserver_request_total:increase1h
- expr: | - expr: |
sum by (code, verb) (increase(apiserver_request_total{job="apiserver",verb="PUT",code=~"4.."}[1h])) sum by (cluster, code, verb) (increase(apiserver_request_total{job="apiserver",verb="PUT",code=~"4.."}[1h]))
record: code_verb:apiserver_request_total:increase1h record: code_verb:apiserver_request_total:increase1h
- expr: | - expr: |
sum by (code, verb) (increase(apiserver_request_total{job="apiserver",verb="PATCH",code=~"4.."}[1h])) sum by (cluster, code, verb) (increase(apiserver_request_total{job="apiserver",verb="PATCH",code=~"4.."}[1h]))
record: code_verb:apiserver_request_total:increase1h record: code_verb:apiserver_request_total:increase1h
- expr: | - expr: |
sum by (code, verb) (increase(apiserver_request_total{job="apiserver",verb="DELETE",code=~"4.."}[1h])) sum by (cluster, code, verb) (increase(apiserver_request_total{job="apiserver",verb="DELETE",code=~"4.."}[1h]))
record: code_verb:apiserver_request_total:increase1h record: code_verb:apiserver_request_total:increase1h
- expr: | - expr: |
sum by (code, verb) (increase(apiserver_request_total{job="apiserver",verb="LIST",code=~"5.."}[1h])) sum by (cluster, code, verb) (increase(apiserver_request_total{job="apiserver",verb="LIST",code=~"5.."}[1h]))
record: code_verb:apiserver_request_total:increase1h record: code_verb:apiserver_request_total:increase1h
- expr: | - expr: |
sum by (code, verb) (increase(apiserver_request_total{job="apiserver",verb="GET",code=~"5.."}[1h])) sum by (cluster, code, verb) (increase(apiserver_request_total{job="apiserver",verb="GET",code=~"5.."}[1h]))
record: code_verb:apiserver_request_total:increase1h record: code_verb:apiserver_request_total:increase1h
- expr: | - expr: |
sum by (code, verb) (increase(apiserver_request_total{job="apiserver",verb="POST",code=~"5.."}[1h])) sum by (cluster, code, verb) (increase(apiserver_request_total{job="apiserver",verb="POST",code=~"5.."}[1h]))
record: code_verb:apiserver_request_total:increase1h record: code_verb:apiserver_request_total:increase1h
- expr: | - expr: |
sum by (code, verb) (increase(apiserver_request_total{job="apiserver",verb="PUT",code=~"5.."}[1h])) sum by (cluster, code, verb) (increase(apiserver_request_total{job="apiserver",verb="PUT",code=~"5.."}[1h]))
record: code_verb:apiserver_request_total:increase1h record: code_verb:apiserver_request_total:increase1h
- expr: | - expr: |
sum by (code, verb) (increase(apiserver_request_total{job="apiserver",verb="PATCH",code=~"5.."}[1h])) sum by (cluster, code, verb) (increase(apiserver_request_total{job="apiserver",verb="PATCH",code=~"5.."}[1h]))
record: code_verb:apiserver_request_total:increase1h record: code_verb:apiserver_request_total:increase1h
- expr: | - expr: |
sum by (code, verb) (increase(apiserver_request_total{job="apiserver",verb="DELETE",code=~"5.."}[1h])) sum by (cluster, code, verb) (increase(apiserver_request_total{job="apiserver",verb="DELETE",code=~"5.."}[1h]))
record: code_verb:apiserver_request_total:increase1h record: code_verb:apiserver_request_total:increase1h
- expr: | - expr: |
sum by (code) (code_verb:apiserver_request_total:increase30d{verb=~"LIST|GET"}) sum by (cluster, code) (code_verb:apiserver_request_total:increase30d{verb=~"LIST|GET"})
labels: labels:
verb: read verb: read
record: code:apiserver_request_total:increase30d record: code:apiserver_request_total:increase30d
- expr: | - expr: |
sum by (code) (code_verb:apiserver_request_total:increase30d{verb=~"POST|PUT|PATCH|DELETE"}) sum by (cluster, code) (code_verb:apiserver_request_total:increase30d{verb=~"POST|PUT|PATCH|DELETE"})
labels: labels:
verb: write verb: write
record: code:apiserver_request_total:increase30d record: code:apiserver_request_total:increase30d

10
manifests/prometheus-prometheusRule.yaml
View File

@@ -211,6 +211,16 @@ spec:
for: 15m for: 15m
labels: labels:
severity: warning severity: warning
- alert: PrometheusLabelLimitHit
annotations:
description: Prometheus {{$labels.namespace}}/{{$labels.pod}} has dropped {{ printf "%.0f" $value }} targets because some samples exceeded the configured label_limit, label_name_length_limit or label_value_length_limit.
runbook_url: https://github.com/prometheus-operator/kube-prometheus/wiki/prometheuslabellimithit
summary: Prometheus has dropped targets because some scrape configs have exceeded the labels limit.
expr: |
increase(prometheus_target_scrape_pool_exceeded_label_limits_total{job="prometheus-k8s",namespace="monitoring"}[5m]) > 0
for: 15m
labels:
severity: warning
- alert: PrometheusErrorSendingAlertsToAnyAlertmanager - alert: PrometheusErrorSendingAlertsToAnyAlertmanager
annotations: annotations:
description: '{{ printf "%.1f" $value }}% minimum errors while sending alerts from Prometheus {{$labels.namespace}}/{{$labels.pod}} to any Alertmanager.' description: '{{ printf "%.1f" $value }}% minimum errors while sending alerts from Prometheus {{$labels.namespace}}/{{$labels.pod}} to any Alertmanager.'

2
manifests/setup/prometheus-operator-deployment.yaml
View File

@@ -17,6 +17,8 @@ spec:
app.kubernetes.io/part-of: kube-prometheus app.kubernetes.io/part-of: kube-prometheus
template: template:
metadata: metadata:
annotations:
kubectl.kubernetes.io/default-container: prometheus-operator
labels: labels:
app.kubernetes.io/component: controller app.kubernetes.io/component: controller
app.kubernetes.io/name: prometheus-operator app.kubernetes.io/name: prometheus-operator