Portainer: Concurrent Stack Updates on Docker Swarm and Kubernetes Endpoints

Answer to the Core Question

Portainer does not attempt to keep a single logical stack in sync across Docker Swarm and Kubernetes endpoints. Each endpoint stores its own stack definition, its own revision number, and applies labels, annotations, and resource limits only to that orchestrator. Consequently, concurrent updates on the same stack name in different endpoints are independent operations and can lead to configuration drift.

Confirmed Facts (Portainer 2.19 and Earlier)

• Stack definitions are kept per‑endpoint; the same stack name appears twice in the UI, once under each endpoint’s Stacks list.
• Revision IDs and version counters are maintained separately. Updating a Swarm stack does not affect the revision of the Kubernetes stack and vice‑versa.
• All orchestrator‑specific metadata (labels, annotations, resource limits) are applied only to the target platform. Changes on one platform do not propagate automatically to the other.
• Portainer provides no cross‑endpoint locking or conflict detection. Two users can simultaneously update the same stack name on different endpoints without interference.
• The Business Edition offers an optional “Stack Synchronization” feature that can be enabled per endpoint, but the default behavior remains endpoint‑specific and independent.

Practical Verification Steps

1. Create a stack named demo-stack on a Docker Swarm endpoint via the UI or API.
2. Create another stack with the same name on a Kubernetes endpoint.
3. Navigate to each stack’s details page and note the Revision number or Revision ID field.
4. Update the Swarm stack’s resource limits (e.g., increase CPU). Re‑inspect the Kubernetes stack; the limits should remain unchanged.
5. Open two browser tabs, each editing demo-stack on a different endpoint, and perform updates simultaneously. Observe that both updates succeed without cross‑endpoint errors.
6. Optionally, query the API: curl -s http://localhost:9000/api/stacks?endpointId=SWARM_ID and curl -s http://localhost:9000/api/stacks?endpointId=KUBE_ID to confirm distinct stack IDs and revision histories.

What to Check Next

If you are using Portainer Business Edition, verify whether the optional “Stack Synchronization” feature is enabled for your endpoints. If it is, the behavior described above may be overridden by the sync configuration.

Summary

Because Portainer treats Swarm and Kubernetes stacks as separate objects, concurrent updates on the same stack name across orchestrators do not interfere. However, this independence can lead to configuration drift if you rely on a single logical stack name. There is no built‑in locking or coordination between endpoints; any concurrency control is handled by the underlying orchestrator.