Automated Sync, Self-Heal, Pruning, and Sync Waves

1. 1

   Turn on automation

   retry handles transient failures (a CRD not yet ready). revisionHistoryLimit keeps the history list short.

   apiVersion: argoproj.io/v1alpha1
   kind: Application
   metadata:
     name: web-dev
     namespace: argocd
   spec:
     project: default
     source:
       repoURL: https://github.com/<you>/shoplite-gitops.git
       targetRevision: main
       path: apps/web
     destination:
       server: https://kubernetes.default.svc
       namespace: shoplite-dev
     revisionHistoryLimit: 10
     syncPolicy:
       automated:
         prune: true
         selfHeal: true
       retry:
         limit: 3
         backoff: { duration: 10s, factor: 2, maxDuration: 2m }
       syncOptions:
         - CreateNamespace=true
     ignoreDifferences:
       - group: apps
         kind: Deployment
         jsonPointers: [/spec/replicas]

2. 2

   Watch self-heal undo a manual change

   Scale by hand and watch. Within seconds Argo notices live ≠ Git and restores it. Here the replicas field is ignored, so try an image change instead, which is exactly the kind of 'quick fix' GitOps forbids.

   terminalcopy cmd

   $ kubectl apply -f argocd/web-dev.yaml

   kubectl -n shoplite-dev set image deploy/web web=nginxdemos/hello:latest

   sleep 10; kubectl -n shoplite-dev get deploy web -o jsonpath='{.spec.template.spec.containers[0].image}'

   argocd app history web-dev | tail -1

   ── expected output ──

   deployment.apps/web image updated

   nginxdemos/hello:0.3

   2 2026-09-27 10:31:02 +0530 IST main (d4e5f6a)

3. 3

   Let the HPA own replicas

   Remove replicas: 2 from the Deployment and add an HPA (Kubernetes course, autoscaling). With the ignoreDifferences entry, Argo stays Synced while the HPA scales.

   apiVersion: autoscaling/v2
   kind: HorizontalPodAutoscaler
   metadata:
     name: web
   spec:
     scaleTargetRef: { apiVersion: apps/v1, kind: Deployment, name: web }
     minReplicas: 2
     maxReplicas: 10
     metrics:
       - type: Resource
         resource: { name: cpu, target: { type: Utilization, averageUtilization: 70 } }

4. 4

   Run a migration before the new version starts

   A PreSync hook Job runs before anything else in the sync. hook-delete-policy: BeforeHookCreation deletes the previous run's Job so the name can be reused. Migrations must be backward compatible (expand, then contract), because old pods keep running during the rollout.

   apiVersion: batch/v1
   kind: Job
   metadata:
     name: db-migrate
     annotations:
       argocd.argoproj.io/hook: PreSync
       argocd.argoproj.io/hook-delete-policy: BeforeHookCreation
   spec:
     backoffLimit: 1
     template:
       spec:
         restartPolicy: Never
         containers:
           - name: migrate
             image: busybox:1.37
             command: ["sh", "-c", "echo running migrations for 0.3 && sleep 5"]

   terminalcopy cmd

   $ git add . && git commit -m 'web: hpa + presync migration' && git push

   argocd app wait web-dev --sync --health && argocd app get web-dev | grep -A6 HOOK

   ── expected output ──

   GROUP KIND NAMESPACE NAME STATUS HEALTH HOOK MESSAGE

   batch Job shoplite-dev db-migrate Succeeded PreSync job.batch/db-migrate created

   ...

5. 5

   Prune: remove from Git, removed from cluster

   Delete the HPA file, push, and it disappears from the cluster. Without prune: true it would stay forever as an orphan, marked OutOfSync with a 'requires pruning' flag.

   terminalcopy cmd

   $ git rm apps/web/hpa.yaml && git commit -m 'web: drop hpa' && git push

   argocd app wait web-dev --sync && kubectl -n shoplite-dev get hpa

   ── expected output ──

   No resources found in shoplite-dev namespace.

Make each change, run the command, and read the error before revealing the diagnosis. Recognising these messages on sight is what makes you fast on a real team. Undo the change afterwards.