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

Add triggered_total metric to alertmanager controller

Browse Source 
Update client_golang for wrappable registerer
This commit is contained in:
Matthias Loibl
2018-10-08 15:29:18 +02:00
parent f855d9cac9
commit acd8924d57
4 changed files with 10 additions and 31 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
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.',
},
expr: |||
count_values("config_hash", alertmanager_config_hash{%(alertmanagerSelector)s}) BY (service) / ON(service) GROUP_LEFT() label_replace(prometheus_operator_alertmanager_spec_replicas{%(prometheusOperatorSelector)s}, "service", "alertmanager-$1", "alertmanager", "(.*)") != 1
count_values("config_hash", alertmanager_config_hash{%(alertmanagerSelector)s}) BY (service) / ON(service) GROUP_LEFT() label_replace(prometheus_operator_spec_replicas{%(prometheusOperatorSelector)s}, "service", "alertmanager-$1", "alertmanager", "(.*)") != 1
||| % $._config,
'for': '5m',
labels: {

19
jsonnet/kube-prometheus/alerts/prometheus-operator.libsonnet
View File

@@ -5,28 +5,15 @@
name: 'prometheus-operator',
rules: [
{
alert: 'PrometheusOperatorAlertmanagerReconcileErrors',
alert: 'PrometheusOperatorReconcileErrors',
expr: |||
rate(prometheus_operator_alertmanager_reconcile_errors_total{%(prometheusOperatorSelector)s}[5m]) > 0.1
rate(prometheus_operator_reconcile_errors_total{%(prometheusOperatorSelector)s}[5m]) > 0.1
||| % $._config,
labels: {
severity: 'warning',
},
annotations: {
message: 'Errors while reconciling Alertmanager in {{ $labels.namespace }} Namespace.',
},
'for': '10m',
},
{
alert: 'PrometheusOperatorPrometheusReconcileErrors',
expr: |||
rate(prometheus_operator_prometheus_reconcile_errors_total{%(prometheusOperatorSelector)s}[5m]) > 0.1
||| % $._config,
labels: {
severity: 'warning',
},
annotations: {
message: 'Errors while reconciling Prometheus in {{ $labels.namespace }} Namespace.',
message: 'Errors while reconciling {{ $labels.controller }} in {{ $labels.namespace }} Namespace.',
},
'for': '10m',
},

2
jsonnetfile.lock.json
View File

@@ -8,7 +8,7 @@
"subdir": "contrib/kube-prometheus/jsonnet/kube-prometheus"
}
},
"version": "049c48c931bfb3cd72efd313b7a47d2244456db0"
"version": "d874b5bc21649dd9d07ab42dd3bdea515038953e"
},
{
"name": "ksonnet",

18
manifests/prometheus-rules.yaml
View File

@@ -947,7 +947,7 @@ spec:
message: The configuration of the instances of the Alertmanager cluster `{{$labels.service}}`
are out of sync.
expr: |
count_values("config_hash", alertmanager_config_hash{job="alertmanager-main"}) BY (service) / ON(service) GROUP_LEFT() label_replace(prometheus_operator_alertmanager_spec_replicas{job="prometheus-operator"}, "service", "alertmanager-$1", "alertmanager", "(.*)") != 1
count_values("config_hash", alertmanager_config_hash{job="alertmanager-main"}) BY (service) / ON(service) GROUP_LEFT() label_replace(prometheus_operator_spec_replicas{job="prometheus-operator"}, "service", "alertmanager-$1", "alertmanager", "(.*)") != 1
for: 5m
labels:
severity: critical
@@ -1099,20 +1099,12 @@ spec:
severity: warning
- name: prometheus-operator
rules:
- alert: PrometheusOperatorAlertmanagerReconcileErrors
- alert: PrometheusOperatorReconcileErrors
annotations:
message: Errors while reconciling Alertmanager in {{ $labels.namespace }}
Namespace.
message: Errors while reconciling {{ $labels.controller }} in {{ $labels.namespace
}} Namespace.
expr: |
rate(prometheus_operator_alertmanager_reconcile_errors_total{job="prometheus-operator"}[5m]) > 0.1
for: 10m
labels:
severity: warning
- alert: PrometheusOperatorPrometheusReconcileErrors
annotations:
message: Errors while reconciling Prometheus in {{ $labels.namespace }} Namespace.
expr: |
rate(prometheus_operator_prometheus_reconcile_errors_total{job="prometheus-operator"}[5m]) > 0.1
rate(prometheus_operator_reconcile_errors_total{job="prometheus-operator"}[5m]) > 0.1
for: 10m
labels:
severity: warning