Cattle Cluster Agent flapping with runtime error: slice bounds out of range [:3] with capacity 0
Article Number: 000021006
Environment
This bug seems to affect all current Rancher versions as of March 2023.
The issue is now fixed since 2.7.5, please upgrade to the latest available version.
Situation
A downstream cluster flaps between Active and Unavailable. The cattle-cluster-agent logs show errors like the following:
0308 15:04:59.702627 55 runtime.go:78] Observed a panic: runtime.boundsError{x:3, y:0, signed:true, code:0x2} (runtime error: slice bounds out of range [:3] with capacity 0)
goroutine 3160 [running]:
k8s.io/apimachinery/pkg/util/runtime.logPanic({0x3cee760, 0xc00bef0660})
/go/pkg/mod/k8s.io/apimachinery@v0.23.3/pkg/util/runtime/runtime.go:74 +0x7d
k8s.io/apimachinery/pkg/util/runtime.HandleCrash({0x0, 0x0, 0xfffffffe})
/go/pkg/mod/k8s.io/apimachinery@v0.23.3/pkg/util/runtime/runtime.go:48 +0x75
panic({0x3cee760, 0xc00bef0660})
/usr/lib64/go/1.17/src/runtime/panic.go:1038 +0x215
github.com/rancher/rancher/pkg/catalogv2/helm.decodeHelm3({0x0, 0xc9d32d})
/go/src/github.com/rancher/rancher/pkg/catalogv2/helm/helm3.go:124 +0x1b1
github.com/rancher/rancher/pkg/catalogv2/helm.fromHelm3Data({0x0, 0xc004d60540}, 0x3fc4c34)
/go/src/github.com/rancher/rancher/pkg/catalogv2/helm/helm3.go:23 +0x25
github.com/rancher/rancher/pkg/catalogv2/helm.ToRelease({0x47b3a20, 0xc004d60540}, 0x6c696877206e6564)
/go/src/github.com/rancher/rancher/pkg/catalogv2/helm/release.go:74 +0x3eb
github.com/rancher/rancher/pkg/controllers/dashboard/helm.(*appHandler).OnSecretChange(0xc00baa1950, {0xc005b81080, 0x2d}, 0xc004d60540)
/go/src/github.com/rancher/rancher/pkg/controllers/dashboard/helm/apps.go:170 +0xa5
Cause
Occasionally, Helm release secrets are improperly stored, causing the release data to not be present. cattle-cluster-agent checks the Helm data everytime it starts, but will fail if there is no release data to check
Resolution
This error seems to be caused by bad Helm release data on the downstream cluster. The first check should be if any releases have no release data stored. The below command will list all Helm release secrets when run against the downstream cluster. If the data column shows 0, that release secret has no release data.
kubectl get secrets -A | grep helm.sh/release.v1
All secrets with no release data on the downstream cluster need to be deleted to allow cattle-cluster-agent to start properly.
kubectl delete secrets -n <NAMESPACE> <SECRET_NAME>
Once the bad Helm release secrets are removed, cattle-cluster-agent pods should successfully start. If desired, the current cattle-cluster-agent pods in a CrashLoopBackOff can be deleted to speed up this process.