Run a remediation plan

Remediation turns AI-proposed fixes into executed actions, gated on your approval. See the concept for the model.

1. Enable remediation

spec:
  ai:
    enabled: true
    credentialsSecret: mirops-ai
    remediation:
      enabled: true
      maxRiskLevel: low          # ceiling on proposed actions
      autoApprove: false         # keep manual approval

When the analysis runs, the operator creates a RemediationPlan in pending-approval.

2. Review the proposed actions

kubectl get remediationplans
kubectl describe remediationplan <name>

Each action has a type, namespace/name, a reason, a risk, and a skip flag:

spec:
  upgradeAnalysisRef: to-1-34
  approved: false
  actions:
    - id: action-1
      type: restart-pod
      namespace: default
      name: crash-test-...
      reason: "CrashLoopBackOff; a restart may clear a transient failure"
      risk: low
      skip: false

3. Approve (or skip individual actions)

# approve the whole plan
kubectl patch remediationplan <name> --type=merge -p '{"spec":{"approved":true}}'

# ...or skip action-2, then approve
kubectl patch remediationplan <name> --type=json \
  -p '[{"op":"replace","path":"/spec/actions/1/skip","value":true}]'

You can also review and approve from the Web UI's Remediation plans view.

4. Watch it run

kubectl get remediationplan <name> -o jsonpath='{.status.phase}{"\n"}'

The phase moves pending-approval → running → completed | failed, and each action records a result (success / failed / skipped) on status.results.

Two gates, recap

Gate	Controls
maxRiskLevel	which actions enter the plan (low / medium / high)
approved	whether the plan runs (default false)
autoApprove	skips approval entirely — use with care

Scope

Actions are pod- and node-level (restart-pod, delete-pod, scale-deployment, cordon-node). Structural fixes — upgrading an add-on, editing a PDB, restoring a volume — stay in your hands; the analysis tells you exactly what to change.