feat(scaling): enable horizontal scaling for API and worker pods

- Add unauthenticated /api/diagnostic/healthz/live and /healthz/ready
  probe endpoints for Kubernetes liveness/readiness checks
- Separate Celery Beat into dedicated beat service in docker-compose.yaml
- Remove container_name from api and worker services to allow scaling
- Create Helm beat-deployment.yaml for standalone Beat scheduler pod
- Remove -B flag from worker-deployment.yaml so workers can scale safely
- Add beat section and fix probe paths in Helm values.yaml
- Add tests for the new probe endpoints

Co-authored-by: christianlouis <361235+christianlouis@users.noreply.github.com>
This commit is contained in:
copilot-swe-agent[bot]
2026-03-16 23:52:56 +00:00
parent b3dce16838
commit f75b125992
6 changed files with 282 additions and 9 deletions
@@ -0,0 +1,83 @@
{{- /*
Celery Beat scheduler — publishes periodic tasks to the broker.
Exactly ONE replica must run; never scale this deployment.
*/ -}}
apiVersion: apps/v1
kind: Deployment
metadata:
name: {{ include "docuelevate.fullname" . }}-beat
namespace: {{ .Release.Namespace }}
labels:
{{- include "docuelevate.labels" . | nindent 4 }}
app.kubernetes.io/component: beat
spec:
replicas: 1
strategy:
type: Recreate # Prevent two Beat instances from running simultaneously
selector:
matchLabels:
{{- include "docuelevate.selectorLabels" . | nindent 6 }}
app.kubernetes.io/component: beat
template:
metadata:
labels:
{{- include "docuelevate.selectorLabels" . | nindent 8 }}
app.kubernetes.io/component: beat
{{- with .Values.beat.podAnnotations }}
annotations:
{{- toYaml . | nindent 8 }}
{{- end }}
spec:
serviceAccountName: {{ include "docuelevate.serviceAccountName" . }}
{{- with .Values.imagePullSecrets }}
imagePullSecrets:
{{- toYaml . | nindent 8 }}
{{- end }}
{{- with .Values.beat.podSecurityContext }}
securityContext:
{{- toYaml . | nindent 8 }}
{{- end }}
containers:
- name: beat
image: {{ include "docuelevate.image" . }}
imagePullPolicy: {{ .Values.image.pullPolicy }}
command:
- celery
- -A
- app.celery_worker
- beat
- --loglevel=info
envFrom:
- configMapRef:
name: {{ include "docuelevate.fullname" . }}-config
- secretRef:
name: {{ include "docuelevate.fullname" . }}-secret
{{- with .Values.beat.securityContext }}
securityContext:
{{- toYaml . | nindent 12 }}
{{- end }}
resources:
{{- toYaml .Values.beat.resources | nindent 12 }}
volumeMounts:
- name: workdir
mountPath: /workdir
volumes:
- name: workdir
{{- if .Values.workdir.persistence.enabled }}
persistentVolumeClaim:
claimName: {{ .Values.workdir.persistence.existingClaim | default (printf "%s-workdir" (include "docuelevate.fullname" .)) }}
{{- else }}
emptyDir: {}
{{- end }}
{{- with .Values.beat.nodeSelector }}
nodeSelector:
{{- toYaml . | nindent 8 }}
{{- end }}
{{- with .Values.beat.affinity }}
affinity:
{{- toYaml . | nindent 8 }}
{{- end }}
{{- with .Values.beat.tolerations }}
tolerations:
{{- toYaml . | nindent 8 }}
{{- end }}
@@ -42,7 +42,6 @@ spec:
- -A
- app.celery_worker
- worker
- -B
- --loglevel=info
- -Q
- document_processor,default,celery
+33 -4
View File
@@ -121,10 +121,10 @@ api:
type: ClusterIP
port: 8000
# Liveness / readiness probes
# Liveness / readiness probes (unauthenticated endpoints for kubelet)
livenessProbe:
httpGet:
path: /api/health
path: /api/diagnostic/healthz/live
port: 8000
initialDelaySeconds: 30
periodSeconds: 20
@@ -132,7 +132,7 @@ api:
readinessProbe:
httpGet:
path: /api/health
path: /api/diagnostic/healthz/ready
port: 8000
initialDelaySeconds: 15
periodSeconds: 10
@@ -191,7 +191,36 @@ worker:
drop: ["ALL"]
# ---------------------------------------------------------------------------
# Shared workdir volume (api + worker mount the same PVC)
# Celery Beat scheduler (singleton — always exactly 1 replica)
# Beat publishes periodic tasks; workers consume them from the broker.
# ---------------------------------------------------------------------------
beat:
resources:
requests:
cpu: 100m
memory: 256Mi
limits:
cpu: 500m
memory: 512Mi
podAnnotations: {}
nodeSelector: {}
tolerations: []
affinity: {}
podSecurityContext:
runAsNonRoot: true
runAsUser: 1000
fsGroup: 1000
securityContext:
allowPrivilegeEscalation: false
readOnlyRootFilesystem: false
capabilities:
drop: ["ALL"]
# ---------------------------------------------------------------------------
# Shared workdir volume (api + worker + beat mount the same PVC)
# ---------------------------------------------------------------------------
workdir:
persistence: