Failed to restart nydus-snapshoter daemonset when modifying it's configuration

[Wu-Jiafeng]

Bug report:

I am deploying nydus by Use helm to install Nydus snapshotter for POC, but find that nydus-snapshotter daemonset will be stuck in ContainerCreating state after modifying it's configuration. Here is the k8s event error log:

LAST SEEN TYPE REASON OBJECT MESSAGE 32S Warning FailedCreatePodSandBox pod/nydus-snapshotter-gxqn7 Failed to create pod sandbox: rpc error: code = Unavailable desc = failed to create containerd container: error unpacking image: failed to prepare extraction snapshot "extract-981644212-oUA6 sha256:xxxxxxxxx" connection error: desc = "transport: Error while dialing: dail unix /run/containerd-nydus/containerd-nydus-grpc.sock: connect: connection refused": unavailable

It seems that the bug reported in dragonflyoss/helm-charts#206 does still exist. Any progress here? If not, is there any best practice for deploying nydus by helm chart? Also, I notice that the latest version of nydus-snapshotter is v0.15.2, maybe the nydus helm chart here needs upgrading as is under discussion in containerd/nydus-snapshotter#612 ...😢

Expected behavior:

nydus-snapshoter daemonset will be reconstructed easily after modification for its configuration or spec.

How to reproduce it:

1. deploy nydus step by step according to Use helm to install Nydus snapshotter
2. modify nydus-snapshotter daemonset's spec, such as adding container env and so on
3. the new nydus-snapshotter instance will be stuck in ContainerCreating state.

Environment:

• nydus helm chart version: v0.0.10
• OS: openeuler 22.03 LTS
• Kernel (e.g. uname -a): Linux $HOSTNAME 5.10.0-60.77.0.101.oe2203.x86_64 #1 #SMP Thu Jan 12 01:32:02 UTC 2023 x86_64 x86_64 x86_64 GNU/Linux
• K8s: v1.32.3
• containerd: 1.6.32

[imeoer]

Hi @Wu-Jiafeng , a new nydus-snapshotter helm chart is in progress here, but it still needs some test CI, will keep moving it forward.