Add VictoriaMetrics observability stack + sync catalog for monitoring test
- VM stack 10 charts: victoria-metrics-cluster/auth, victoria-logs-cluster, victoria-metrics-agent/alert, opentelemetry-collector, kube-state-metrics, prometheus-node-exporter, alertmanager, perses (JWT/OIDC, Infisical-ready) - ArgoCD ApplicationSet (syncWave) + per-chart dip-values overlays - doc/victoria-metrics-architecture.md, define-chart-resources updates - includes pending working-tree changes (mlflow, kubeflow, apisix, CLAUDE.md) Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
This commit is contained in:
@@ -0,0 +1,184 @@
|
||||
{{- $storageName := "server-volume" }}
|
||||
{{- $app := .Values.alertmanager }}
|
||||
{{- $pvc := $app.persistentVolume }}
|
||||
{{- $mode := $app.mode }}
|
||||
{{- if and $mode $app.enabled (hasKey $app $mode) -}}
|
||||
{{- $modeOpts := index $app $mode }}
|
||||
{{- $ctx := dict "helm" . "appKey" "alertmanager" }}
|
||||
{{- $fullname := include "vm.plain.fullname" $ctx }}
|
||||
{{- $ns := include "vm.namespace" $ctx }}
|
||||
{{- $sa := include "vm.fullname" . }}
|
||||
{{- if and (ne $mode "statefulSet") (gt (int $app.replicaCount) 1) }}
|
||||
{{- fail "Alertmanager HA mode is not supported for Deployment. Consider switching to statefulset instead using `alertmanager.mode: statefulSet`" -}}
|
||||
{{- end }}
|
||||
apiVersion: apps/v1
|
||||
kind: {{ title $mode }}
|
||||
metadata:
|
||||
name: {{ $fullname }}
|
||||
namespace: {{ $ns }}
|
||||
labels: {{ include "vm.labels" $ctx | nindent 4 }}
|
||||
spec:
|
||||
{{- with $modeOpts.spec }}
|
||||
{{- toYaml . | nindent 2 }}
|
||||
{{- end }}
|
||||
replicas: {{ $app.replicaCount | default 1 }}
|
||||
{{- if eq $mode "statefulSet" }}
|
||||
serviceName: {{ $fullname }}
|
||||
{{- end }}
|
||||
selector:
|
||||
matchLabels: {{ include "vm.selectorLabels" $ctx | nindent 6 }}
|
||||
template:
|
||||
metadata:
|
||||
{{- $_ := set $ctx "extraLabels" $app.podLabels }}
|
||||
labels: {{ include "vm.podLabels" $ctx | nindent 8 }}
|
||||
{{- $_ := unset $ctx "extraLabels" }}
|
||||
{{- $annotations := dict "checksum/config" (include (print .Template.BasePath "/alertmanager-configmap.yaml") . | sha256sum) }}
|
||||
{{- $annotations = merge $annotations (deepCopy $app.podAnnotations) }}
|
||||
annotations: {{ toYaml $annotations | nindent 8 }}
|
||||
spec:
|
||||
{{- if or (.Values.serviceAccount).name (.Values.serviceAccount).create }}
|
||||
serviceAccountName: {{ tpl ((.Values.serviceAccount).name | default $sa) $ctx }}
|
||||
automountServiceAccountToken: {{ .Values.serviceAccount.automountToken }}
|
||||
{{- end }}
|
||||
{{- if $app.podSecurityContext.enabled }}
|
||||
securityContext: {{ include "vm.securityContext" (dict "securityContext" $app.podSecurityContext "helm" .) | nindent 8 }}
|
||||
{{- end }}
|
||||
{{- with ($app.imagePullSecrets | default .Values.global.imagePullSecrets) }}
|
||||
imagePullSecrets: {{ toYaml . | nindent 8 }}
|
||||
{{- end }}
|
||||
{{- with $app.initContainers }}
|
||||
initContainers: {{ toYaml . | nindent 8 }}
|
||||
{{- end }}
|
||||
containers:
|
||||
- name: alertmanager
|
||||
{{- if $app.securityContext.enabled }}
|
||||
securityContext: {{ include "vm.securityContext" (dict "securityContext" $app.securityContext "helm" .) | nindent 12 }}
|
||||
{{- end }}
|
||||
image: {{ include "vm.image" $ctx }}
|
||||
args: {{ include "alertmanager.args" $ctx | nindent 12 }}
|
||||
ports:
|
||||
- name: web
|
||||
containerPort: {{ include "vm.port.from.flag" (dict "flag" $app.listenAddress "default" "9093") }}
|
||||
{{- if gt (int $app.replicaCount) 1 }}
|
||||
- name: cluster-tcp
|
||||
containerPort: {{ include "vm.port.from.flag" (dict "flag" $app.cluster.listenAddress "default" "9094") }}
|
||||
protocol: TCP
|
||||
- name: cluster-udp
|
||||
containerPort: {{ include "vm.port.from.flag" (dict "flag" $app.cluster.listenAddress "default" "9094") }}
|
||||
protocol: UDP
|
||||
{{- end }}
|
||||
{{- with $app.envFrom }}
|
||||
envFrom: {{ toYaml . | nindent 12 }}
|
||||
{{- end }}
|
||||
env:
|
||||
- name: POD_IP
|
||||
valueFrom:
|
||||
fieldRef:
|
||||
apiVersion: v1
|
||||
fieldPath: status.podIP
|
||||
{{- with (fromYaml (include "vm.probe" (dict "app" $app "type" "readiness"))) }}
|
||||
readinessProbe: {{ toYaml . | nindent 12 }}
|
||||
{{- end }}
|
||||
{{- with (fromYaml (include "vm.probe" (dict "app" $app "type" "liveness"))) }}
|
||||
livenessProbe: {{ toYaml . | nindent 12 }}
|
||||
{{- end }}
|
||||
{{- with (fromYaml (include "vm.probe" (dict "app" $app "type" "startup"))) }}
|
||||
startupProbe: {{ toYaml . | nindent 12 }}
|
||||
{{- end }}
|
||||
volumeMounts:
|
||||
- name: {{ $storageName }}
|
||||
mountPath: {{ ternary $pvc.mountPath "/data" $pvc.enabled }}
|
||||
- name: config
|
||||
mountPath: /config
|
||||
readOnly: true
|
||||
{{- range $app.extraHostPathMounts }}
|
||||
- name: {{ .name }}
|
||||
mountPath: {{ .mountPath }}
|
||||
{{- with .subPath }}
|
||||
subPath: {{ . }}
|
||||
{{- end }}
|
||||
{{- with .readOnly }}
|
||||
readOnly: {{ . }}
|
||||
{{- end }}
|
||||
{{- end }}
|
||||
{{- with $app.extraVolumeMounts }}
|
||||
{{- toYaml . | nindent 12 }}
|
||||
{{- end }}
|
||||
{{- with $app.resources }}
|
||||
resources: {{ toYaml . | nindent 12 }}
|
||||
{{- end }}
|
||||
{{- with $app.extraContainers }}
|
||||
{{- toYaml . | nindent 8 }}
|
||||
{{- end }}
|
||||
{{- with $app.nodeSelector }}
|
||||
nodeSelector: {{ toYaml . | nindent 8 }}
|
||||
{{- end }}
|
||||
{{- if $app.priorityClassName }}
|
||||
priorityClassName: {{ $app.priorityClassName | quote }}
|
||||
{{- end }}
|
||||
{{- with $app.affinity }}
|
||||
affinity: {{ toYaml . | nindent 8 }}
|
||||
{{- end }}
|
||||
{{- with $app.topologySpreadConstraints }}
|
||||
topologySpreadConstraints:
|
||||
{{- range $constraint := . }}
|
||||
- {{ toYaml $constraint | nindent 10 | trim }}
|
||||
{{- if not $constraint.labelSelector }}
|
||||
labelSelector:
|
||||
matchLabels: {{ include "vm.selectorLabels" $ctx | nindent 14 }}
|
||||
{{- end }}
|
||||
{{- end }}
|
||||
{{- end }}
|
||||
{{- with $app.tolerations }}
|
||||
tolerations: {{ toYaml . | nindent 8 }}
|
||||
{{ end }}
|
||||
volumes:
|
||||
{{- $_ := set $ctx "appKey" "alertmanager" }}
|
||||
- name: config
|
||||
configMap:
|
||||
name: {{ include "alertmanager.config.name" $ctx }}
|
||||
{{- range $app.extraHostPathMounts }}
|
||||
- name: {{ .name }}
|
||||
hostPath:
|
||||
path: {{ .hostPath }}
|
||||
{{- end }}
|
||||
{{- with $app.extraVolumes }}
|
||||
{{- toYaml . | nindent 8 }}
|
||||
{{- end }}
|
||||
{{- if or (eq $mode "deployment") (not $pvc.enabled) $pvc.existingClaim }}
|
||||
- name: {{ $storageName }}
|
||||
{{- if or (and (eq $mode "deployment") $pvc.enabled) $pvc.existingClaim }}
|
||||
persistentVolumeClaim:
|
||||
claimName: {{ tpl ($pvc.existingClaim | default $pvc.name | default $fullname) $ctx }}
|
||||
{{- else }}
|
||||
emptyDir: {{ toYaml $app.emptyDir | nindent 12 }}
|
||||
{{- end }}
|
||||
{{- end }}
|
||||
{{- if and (eq $mode "statefulSet") $pvc.enabled (not $pvc.existingClaim) }}
|
||||
volumeClaimTemplates:
|
||||
- apiVersion: v1
|
||||
kind: PersistentVolumeClaim
|
||||
metadata:
|
||||
name: {{ tpl ($pvc.name | default $storageName) $ctx }}
|
||||
{{- with $pvc.annotations }}
|
||||
annotations: {{ toYaml . | nindent 10 }}
|
||||
{{- end }}
|
||||
{{- with $pvc.extraLabels }}
|
||||
labels: {{ toYaml . | nindent 10 }}
|
||||
{{- end }}
|
||||
spec:
|
||||
{{- with $pvc.accessModes }}
|
||||
accessModes: {{ toYaml . | nindent 10 }}
|
||||
{{- end }}
|
||||
resources:
|
||||
requests:
|
||||
storage: {{ $pvc.size }}
|
||||
{{- with $pvc.storageClassName }}
|
||||
storageClassName: {{ ternary "" . (eq "-" .) }}
|
||||
{{- end }}
|
||||
{{- with $pvc.matchLabels }}
|
||||
selector:
|
||||
matchLabels: {{ toYaml . | nindent 12 }}
|
||||
{{- end }}
|
||||
{{- end }}
|
||||
{{- end -}}
|
||||
Reference in New Issue
Block a user