Rolling Restart Backend Deployment

Perform a Kubernetes rolling restart of the backend deployment. This restarts all pods with the existing image - useful for config changes or resetting pod state.

Note: This does NOT rebuild or redeploy new code. For new code, use the deploy skill.

Workflow

Step 1: Get Current Repo

git config --get remote.origin.url
git branch --show-current

Parse repo as owner/repo.

Step 2: Find Existing Pipeline

Use list_pipelines to find the pipeline for this repo+branch.

If not found: Tell user "No pipeline found. Use the deploy skill first to create one."

Step 3: Trigger Rolling Restart

Use restart_deployment with the pipelineId.

Tell user: "Rolling restart triggered. Waiting for pods to restart..."

Step 4: Poll Until Complete (REQUIRED)

You MUST poll for status - do not skip this step.

Call get_deployment_status with the pipelineId in a loop:

LOOP (max 24 iterations = 2 minutes):
  1. Call get_deployment_status(pipelineId)
  2. Check response:
     - If ready=true → DONE, go to Step 5 (success)
     - If failed=true → DONE, go to Step 5 (failure)
     - If rolloutInProgress=true OR ready=false → Continue polling
  3. Report progress: "Pods: {readyReplicas}/{replicas} ready, rollout in progress..."
  4. Wait 5 seconds
  5. Repeat

Response fields to check:

• ready: boolean - true when rollout is fully complete
• rolloutInProgress: boolean - true while pods are being updated
• readyReplicas: number of ready pods
• replicas: total desired pods
• failed: boolean - true if pod failure detected
• failureReason: string - error details if failed

Step 5: Report Final Status

On success (ready=true):

✓ Rolling restart complete for {appName}
All pods healthy: {readyReplicas}/{replicas} ready
URLs: {deployedUrls}

On failure (failed=true):

✗ Restart failed: {failureReason}
Pod details: {pods}

On timeout (24 polls without ready=true):

⚠ Restart timeout - pods may still be cycling
Current: {readyReplicas}/{replicas} ready
Check pod status manually or try again

When to Use

• Config/env changes picked up from secrets
• Reset application state
• Clear memory leaks
• Quick pod refresh

When NOT to Use

• Deploying new code changes → use the deploy skill instead
• First-time deployment → use the deploy skill instead

Troubleshooting

If restart fails or pods have issues:

# Get logs from all pods
get_deployment_logs(pipelineId)

# Check env vars (secrets masked)
get_pod_env_vars(pipelineId)

# Get detailed pod status
get_deployment_status(pipelineId)