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

Merge pull request #186 from LiliC/fix-many2many

Browse Source 
*: Fix AlertmanagerConfigInconsistent many-to-many matching errors
This commit is contained in:
Frederic Branczyk
2019-08-16 16:25:27 +02:00
committed by GitHub
parent a2875bc637 c6e6f2e74f
commit c4b89c9414
2 changed files with 4 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
jsonnet/kube-prometheus/alerts/alertmanager.libsonnet
View File

@@ -10,7 +10,7 @@
message: 'The configuration of the instances of the Alertmanager cluster `{{$labels.service}}` are out of sync.', message: 'The configuration of the instances of the Alertmanager cluster `{{$labels.service}}` are out of sync.',
}, },
expr: ||| expr: |||
count_values("config_hash", alertmanager_config_hash{%(alertmanagerSelector)s}) BY (service) / ON(service) GROUP_LEFT() label_replace(prometheus_operator_spec_replicas{%(prometheusOperatorSelector)s,controller="alertmanager"}, "service", "alertmanager-$1", "name", "(.*)") != 1 count_values("config_hash", alertmanager_config_hash{%(alertmanagerSelector)s}) BY (service) / ON(service) GROUP_LEFT() label_replace(max(prometheus_operator_spec_replicas{%(prometheusOperatorSelector)s,controller="alertmanager"}) by (name, job, namespace, controller), "service", "alertmanager-$1", "name", "(.*)") != 1
||| % $._config, ||| % $._config,
'for': '5m', 'for': '5m',
labels: { labels: {
@@ -31,8 +31,8 @@
}, },
}, },
{ {
alert:'AlertmanagerMembersInconsistent', alert: 'AlertmanagerMembersInconsistent',
annotations:{ annotations: {
message: 'Alertmanager has not found all other members of the cluster.', message: 'Alertmanager has not found all other members of the cluster.',
}, },
expr: ||| expr: |||

2
manifests/prometheus-rules.yaml
View File

@@ -1054,7 +1054,7 @@ spec:
message: The configuration of the instances of the Alertmanager cluster `{{$labels.service}}` message: The configuration of the instances of the Alertmanager cluster `{{$labels.service}}`
are out of sync. are out of sync.
expr: | expr: |
count_values("config_hash", alertmanager_config_hash{job="alertmanager-main",namespace="monitoring"}) BY (service) / ON(service) GROUP_LEFT() label_replace(prometheus_operator_spec_replicas{job="prometheus-operator",namespace="monitoring",controller="alertmanager"}, "service", "alertmanager-$1", "name", "(.*)") != 1 count_values("config_hash", alertmanager_config_hash{job="alertmanager-main",namespace="monitoring"}) BY (service) / ON(service) GROUP_LEFT() label_replace(max(prometheus_operator_spec_replicas{job="prometheus-operator",namespace="monitoring",controller="alertmanager"}) by (name, job, namespace, controller), "service", "alertmanager-$1", "name", "(.*)") != 1
for: 5m for: 5m
labels: labels:
severity: critical severity: critical