Argo CD Interview Questions for Experienced Engineers (2027)

42 most asked Argo CD questions for experienced engineers: sync waves, hooks, ApplicationSets, RBAC, sharding and Argo CD 3.x changes.

Sep 17, 2025 - 11:36
Nov 8, 2026 - 02:57
 0  69
Argo CD Interview Questions for Experienced Engineers (2027)

Experienced engineers are asked different Argo CD questions from freshers: how to structure projects and RBAC for many teams, how sync waves and hooks order a release, how to tune diffing and health checks, how to scale the controller, and what changed in Argo CD 3.x. This guide has 42 of those questions with direct answers, grouped by level, plus troubleshooting scenarios.

These are the questions that come up most often once you list three or more years of Argo CD on your CV. For incident-style practice, use our scenario-based Argo CD interview questions, the main Argo CD page on this site. For CLI and day-to-day operations, see real-time Argo CD questions, and for principles, GitOps and Argo CD questions. Facts below were checked against the Argo CD documentation in November 2026, when v3.6.1 was the latest stable release on GitHub.

Key takeaways

  • Design questions centre on AppProjects: which repos, clusters, namespaces and resource kinds each team may use.
  • Sync waves (argocd.argoproj.io/sync-wave) order resources; hooks (argocd.argoproj.io/hook) run Jobs at PreSync, Sync, PostSync, SyncFail, PreDelete or PostDelete.
  • Argo CD 3.0 changed several defaults: annotation-based resource tracking, fine-grained RBAC for update and delete, and enforced log access.
  • Most "always OutOfSync" problems are fixed with ignoreDifferences and RespectIgnoreDifferences=true, not by turning off auto-sync.
  • Scaling answers name the component: shard the application controller, scale and limit the repo server, and use webhooks instead of relying on polling.

What experienced candidates are tested on

AreaTypical questionQuestions here
Core behaviour refresherRefresh vs sync, health vs sync status1 to 8
Release designWaves, hooks, sync options, ApplicationSets9 to 22
Platform and governanceProjects, RBAC, SSO, multi-cluster, scaling, 3.x changes23 to 36
Troubleshooting scenariosStuck syncs, drift, timeouts37 to 42

Refresher questions (asked even at senior level)

1. What is the difference between refresh and sync?

Refresh compares the live state with the desired state from Git and updates the status. Sync applies the desired state to the cluster. A refresh never changes cluster resources.

2. What is a hard refresh?

A hard refresh also discards the cached manifests and regenerates them in the repo server. Use it when you suspect a stale cache or a cached manifest generation error.

3. Sync status versus health status?

Sync status says whether live resources match Git (Synced or OutOfSync). Health says whether those resources are working (Healthy, Progressing, Degraded, Suspended, Missing). An app can be Synced and Degraded at the same time.

4. How often does Argo CD check Git by default?

About every three minutes: 120 seconds plus up to 60 seconds of jitter, set by timeout.reconciliation and timeout.reconciliation.jitter in argocd-cm. Git webhooks trigger refresh immediately.

5. What do prune and selfHeal do in automated sync?

prune deletes resources that were removed from Git. selfHeal reverts manual changes in the cluster back to the Git state. Both are off by default when you enable automated sync.

6. What are the main Argo CD components?

The API server (UI, CLI and API), the repo server (renders manifests from Git, Helm, Kustomize or plugins), the application controller (compares and syncs), Redis (cache), and optionally the ApplicationSet controller, notifications controller and Dex for SSO.

7. What is the app-of-apps pattern?

One parent Application points to a directory of Application manifests, so creating the parent creates the children. It is used to bootstrap clusters. ApplicationSets are often a better fit when the children follow a pattern.

8. How does Argo CD know a resource belongs to an app?

Through resource tracking. Since Argo CD 3.0 the default method is annotation-based; label-based tracking can still be selected with application.resourceTrackingMethod.

2 to 5 years: release design

9. How do you make a database migration run before the new Deployment?

Annotate the migration Job with argocd.argoproj.io/hook: PreSync. Argo CD runs it before applying other manifests and stops the sync if it fails.

10. What hook delete policies exist?

HookSucceeded, HookFailed and BeforeHookCreation. BeforeHookCreation is the default when no policy is set; it deletes the previous hook object before creating a new one.

11. How do sync waves work?

Resources carry an integer in argocd.argoproj.io/sync-wave (default 0, negatives allowed). Argo CD applies lower waves first and waits for each wave to become healthy before the next. Waves also apply within hook phases.

12. Give an example of waves and hooks together.

apiVersion: batch/v1
kind: Job
metadata:
  name: db-migrate
  annotations:
    argocd.argoproj.io/hook: PreSync
    argocd.argoproj.io/hook-delete-policy: BeforeHookCreation
---
apiVersion: v1
kind: ConfigMap
metadata:
  name: app-config
  annotations:
    argocd.argoproj.io/sync-wave: "-1"

The migration runs in PreSync, the ConfigMap is applied in wave -1, and the Deployment follows in wave 0.

13. How do you send a notification only when a sync fails?

Use a SyncFail hook Job, or better, the notifications controller with a trigger on failed sync and a template for Slack or email.

14. When would you use ServerSideApply=true?

When resources are too large for the last-applied annotation (large CRDs are the usual case), or when you need field ownership to cooperate with other controllers that set fields on the same object.

15. What does Replace=true do, and when is it risky?

It uses replace (or create) instead of apply. It helps with immutable fields, but it can drop fields set by other controllers, so use it per resource, not per app.

16. Which sync options do you use for namespaces and pruning?

CreateNamespace=true creates the destination namespace. PruneLast=true prunes after all other resources are healthy. Prune=false on a resource protects it from pruning, and Prune=confirm requires manual confirmation.

17. What is ApplyOutOfSyncOnly=true for?

It applies only resources that are out of sync, which reduces API load for apps with thousands of objects. After the 3.0 upgrade to annotation tracking, the upgrade notes say such apps may need an explicit full sync to avoid orphaned resources.

18. What are ApplicationSet generators?

The documented generators are List, Cluster, Git, Matrix, Merge, SCM Provider, Pull Request, Cluster Decision Resource, Plugin and OCI. Matrix and Merge combine other generators.

19. How do you deploy every microservice folder to every cluster?

Use a Matrix generator that combines a Git directory generator (one entry per service folder) with a Cluster generator (one entry per registered cluster, filtered by labels such as env=prod).

20. How do you create preview environments per pull request?

Use the Pull Request generator with your Git provider. Each open pull request creates an Application in its own namespace; closing the pull request deletes it.

21. How do you stop an ApplicationSet from deleting apps when a generator returns nothing?

Use the ApplicationSet sync policy settings that prevent deletion (create-only or create-update instead of the default sync behaviour) and preserve resources on deletion where needed. Explain that a misconfigured generator returning empty results is a real outage risk.

22. How do you combine a Helm chart from one repo with values from another?

Use multiple sources (spec.sources) with a ref on the values repository, and refer to its files with that ref name in the Helm source, for example $values/charts/app/values.yaml when the ref is values.

Senior: platform, governance and scale

23. How do you isolate teams in one Argo CD instance?

One AppProject per team, limiting sourceRepos, destinations (cluster and namespace), and allowed cluster-scoped and namespaced resource kinds. Bind team groups to project roles through RBAC.

24. Where is RBAC configured?

In the argocd-rbac-cm ConfigMap, using policy.csv lines and policy.default, or in project roles inside an AppProject. Groups usually come from SSO claims.

25. What changed in RBAC in Argo CD 3.0?

update and delete on an Application no longer apply to its sub-resources by default; grant update/* or delete/* explicitly. Log viewing now needs an explicit logs, get permission. For Dex, subjects use the federated_claims.user_id claim instead of sub.

26. Which other 3.0 defaults should you know before upgrading?

Changes to .status are ignored by default, high-churn kinds such as Endpoints, EndpointSlice and Lease are excluded, resource health is no longer stored in the Application CR, and repositories defined in argocd-cm must move to Secrets.

27. How do you set up SSO?

Use the bundled Dex with a connector (for example GitHub or LDAP), or configure an OIDC provider directly in argocd-cm. Disable the local admin account after SSO works.

28. Hub-and-spoke or one Argo CD per cluster?

A central instance gives one view and simpler management but becomes a single point of failure and needs network access to every cluster. One instance per cluster isolates failures but multiplies upgrades. Many teams pick a hub per environment or region.

29. How do you scale the application controller?

Increase replicas of the argocd-application-controller StatefulSet and set ARGOCD_CONTROLLER_REPLICAS to match. Clusters are split across shards; the sharding methods are legacy (default), round-robin and consistent-hashing, the last two marked experimental. A cluster can be pinned with a shard field on its secret.

30. How do you stop the repo server running out of memory?

Scale its Deployment, set --parallelismlimit to cap concurrent manifest generations, and raise ARGOCD_EXEC_TIMEOUT (default 90 seconds) if large Helm or Kustomize renders time out.

31. How do you fix "context deadline exceeded" during reconciliation?

Increase --repo-server-timeout-seconds on the controller and scale the repo server, then find out why rendering is slow (huge charts, slow Git, plugins).

32. How do you handle secrets with Argo CD?

Keep secrets out of Git in plain text. Common choices are an external secret controller that syncs from a secret manager, Sealed Secrets, or SOPS through a plugin. Argo CD then syncs only references or encrypted data.

33. How do you write a custom health check?

Add a Lua script under resource.customizations.health.<group>_<kind> in argocd-cm that reads the resource's status and returns a health status and message. This is common for CRDs that Argo CD does not know.

34. How do you protect production from accidental syncs?

Sync windows on the AppProject (deny windows during business events), RBAC that allows sync only to certain groups, and manual sync for production apps while lower environments auto-sync.

35. How do you monitor Argo CD?

Scrape its Prometheus metrics. Since 3.0, app sync and health status are labels on argocd_app_info; the older separate status metrics were removed. Alert on apps that stay OutOfSync or Degraded and on controller queue depth.

36. How do you do progressive delivery with Argo CD?

Argo CD syncs the manifests; Argo Rollouts (or a service mesh) handles canary and blue-green steps and analysis. Argo CD understands Rollout health, so the app shows Progressing until the rollout completes.

Troubleshooting scenarios

37. An app is always OutOfSync because an HPA changes replicas.

Remove replicas from the Deployment manifest, or add an ignoreDifferences entry for /spec/replicas and set RespectIgnoreDifferences=true so syncs do not reset it.

38. A sync hangs at a PreSync hook.

Check the hook Job's Pods and logs. A Job that never completes blocks the sync; give it activeDeadlineSeconds and a backoff limit, then terminate the operation and resync.

39. After upgrading to 3.x, a team can no longer delete Pods from the UI.

Fine-grained RBAC no longer inherits delete to sub-resources. Add delete/* permission for that project role, or deliberately restore the old inheritance setting.

40. Resources were left behind after a sync with ApplyOutOfSyncOnly.

The tracking method changed to annotations in 3.0. Run a full sync once so Argo CD adds tracking annotations, then prune the orphaned resources.

41. A CRD sync fails because the annotation is too long.

Enable ServerSideApply=true for that resource or app so the last-applied annotation is not used.

42. Someone ran kubectl edit in production and the change vanished.

selfHeal reverted it to the Git state, which is the expected GitOps behaviour. Make the change in Git through a pull request.

How to prepare

  • Install Argo CD in a kind cluster, register a second cluster, and build one ApplicationSet with a Matrix generator.
  • Write a PreSync migration Job and a sync-wave sequence, then break the Job on purpose.
  • Read the 2.14 to 3.0 upgrade notes once; upgrade questions are common for senior roles.
  • Prepare one story about drift or a failed sync that you resolved.

Argo CD sits on top of Kubernetes skills. The CKA training and the DevOps training (which includes GitOps) run in Kharadi, Pune and live online across India. For the full list of GitOps tools, see top GitOps tools for Kubernetes.

FAQ

What Argo CD questions do experienced engineers get?

Project and RBAC design, sync waves and hooks, ApplicationSets, diffing and health customisation, scaling and the 3.x upgrade changes.

What is the latest Argo CD version?

GitHub listed v3.6.1 as the latest stable release in November 2026. Check the releases page before your interview.

What is the default Argo CD polling interval?

About three minutes: 120 seconds plus up to 60 seconds of jitter. Webhooks make changes appear immediately.

What are the Argo CD hook phases?

PreSync, Sync, PostSync, SyncFail, PreDelete and PostDelete, plus Skip to leave a manifest unapplied.

How do I fix an app that is always OutOfSync?

Find the field that differs with argocd app diff, then ignore it with ignoreDifferences or stop setting it in Git.

How is this different from the scenario-based Argo CD guide?

This page covers design and platform knowledge expected at three or more years. The scenario guide focuses on incident-style questions.

References

What's Your Reaction?

Like Like 0
Dislike Dislike 0
Love Love 0
Funny Funny 0
Angry Angry 0
Sad Sad 0
Wow Wow 0
Mridul I am a passionate technology enthusiast with a strong focus on DevOps, Cloud Computing, and Cybersecurity. Through my blogs at DevOps Training Institute, I aim to simplify complex concepts and share practical insights for learners and professionals. My goal is to empower readers with knowledge, hands-on tips, and industry best practices to stay ahead in the ever-evolving world of DevOps.