-
Bug
-
Resolution: Done
-
Undefined
-
None
-
2.7.0.GA
-
None
-
None
-
False
-
-
False
-
-
-
- Diagnostic Tools(Report.sh) throws an error if there are two replicasets of ClusterOperator
Error from server (NotFound): replicasets.apps "strimzi-cluster-operator-xxx\nstrimzi-cluster-operator-yyy" not found
- Directly, strimzi-cluster-operator is managed as a Deployment, but Replicasets and Pods are created internally by OpenShift from that Deployment definition.
- The Replicaset of the strimzi-cluster-operator is normally only one, but when changing the strimzi-cluster-operator, such as version upgrades, two replicasets may be created temporarily.
- Below is the code around L247 in Diagnostic Tools(Report.sh). I think the error is probably caused by L257, where it is assumed that there is only one name in '$CO_RS', but there are two replicasets. Pods use a for loop to deal with multiple Pods, so a similar solution would be implemented for Replicaset.
echo "clusteroperator" CO_DEPLOY=$($KUBE_CLIENT get deploy strimzi-cluster-operator -o name -n "$NAMESPACE" --ignore-not-found) if [[ -n $CO_DEPLOY ]]; then echo " $CO_DEPLOY" CO_DEPLOY=$(echo "$CO_DEPLOY" | cut -d "/" -f 2) && readonly CO_DEPLOY $KUBE_CLIENT get deploy "$CO_DEPLOY" -o yaml -n "$NAMESPACE" > "$OUT_DIR"/reports/deployments/"$CO_DEPLOY".yaml CO_RS=$($KUBE_CLIENT get rs -l strimzi.io/kind=cluster-operator -o name -n "$NAMESPACE" --ignore-not-found) if [[ -n $CO_RS ]]; then echo " $CO_RS" CO_RS=$(echo "$CO_RS" | cut -d "/" -f 2) && readonly CO_RS $KUBE_CLIENT get rs "$CO_RS" -o yaml -n "$NAMESPACE" > "$OUT_DIR"/reports/replicasets/"$CO_RS".yaml fi mapfile -t CO_PODS < <($KUBE_CLIENT get po -l strimzi.io/kind=cluster-operator -o name -n "$NAMESPACE" --ignore-not-found) if [[ ${#CO_PODS[@]} -ne 0 ]]; then for pod in "${CO_PODS[@]}"; do echo " $pod" CO_POD=$(echo "$pod" | cut -d "/" -f 2) $KUBE_CLIENT get po "$CO_POD" -o yaml -n "$NAMESPACE" > "$OUT_DIR"/reports/pods/"$CO_POD".yaml get_pod_logs "$CO_POD" done fi
