Skip to main content
Severity: P1
Estimated time: 20-40 minutes This runbook operationalizes the upgrade story from issue #7 and points to the full design/procedure in the upgrade documentation.

​
Symptoms

  • Planned operator version upgrade.
  • Need to patch operator image without disrupting existing RedisCluster workloads.

​
Prerequisites

  • Helm access to the operator release.
  • Ability to inspect operator and RedisCluster resources.
  • Shell variables:
export OP_NS=<operator-namespace>
export RELEASE=<helm-release-name>
export CLUSTER_NS=<rediscluster-namespace>
export CLUSTER=<rediscluster-name>

​
Diagnosis

1

Capture pre-upgrade state

kubectl get redisclusters.redis.io -A
kubectl get pods -n "$CLUSTER_NS" -l redis.io/cluster="$CLUSTER" -o wide
kubectl get svc -n "$CLUSTER_NS" "$CLUSTER-leader" "$CLUSTER-replica" "$CLUSTER-any"
kubectl get events -n "$CLUSTER_NS" --sort-by=.lastTimestamp | tail -n 40
kubectl get rediscluster "$CLUSTER" -n "$CLUSTER_NS" -o jsonpath='{.status.currentPrimary}{"\n"}'
2

Confirm operator deployment identity

kubectl get deploy -n "$OP_NS" -l app.kubernetes.io/name=redis-operator
export OP_DEPLOY="$(kubectl get deploy -n "$OP_NS" -l app.kubernetes.io/name=redis-operator -o jsonpath='{.items[0].metadata.name}')"
echo "$OP_DEPLOY"

​
Recovery Steps

Follow the upgrade documentation as the source-of-truth procedure.
1

Run the Helm upgrade

helm upgrade "$RELEASE" charts/redis-operator \
  --namespace "$OP_NS" \
  --reuse-values \
  --set image.repository=<repo> \
  --set image.tag=<new-tag>
2

Watch operator rollout and leader lease

kubectl rollout status deployment/"$OP_DEPLOY" -n "$OP_NS"
kubectl get lease redis-operator-leader -n "$OP_NS"
kubectl logs -n "$OP_NS" deploy/"$OP_DEPLOY" --tail=200
3

Watch Redis cluster continuity during rollout

kubectl get rediscluster "$CLUSTER" -n "$CLUSTER_NS" -o yaml
kubectl get pods -n "$CLUSTER_NS" -l redis.io/cluster="$CLUSTER" -w

​
Verification

kubectl get rediscluster "$CLUSTER" -n "$CLUSTER_NS" -o jsonpath='{.status.phase}{"\n"}'
kubectl get rediscluster "$CLUSTER" -n "$CLUSTER_NS" -o jsonpath='{.status.currentPrimary}{"\n"}'
kubectl get pods -n "$CLUSTER_NS" -l redis.io/cluster="$CLUSTER"
Expected:
  • Operator rollout completes successfully with leader election continuity.
  • Existing Redis clusters remain available.
  • Data pods roll in controlled order only when spec hash is outdated (replicas first, primary last).
  • No bulk restart of all healthy data pods.
  • Cluster returns to Healthy.

​
Escalation

If availability degrades during upgrade, pause and assess rollback.
Rollback command: 
helm rollback "$RELEASE" <previous-revision> --namespace "$OP_NS"
Capture logs/events and compare against upgrade documentation expectations and issue #7 (.issues/007_operator_upgrade_story.md).

Build docs developers (and LLMs) love

Get started for freeTalk to us