failed to start the kubelet with nydus-snapshotter deamonset after reboot #420
Description
When installing nydus-snapshotter using a helm chart, the helm chart modifies containerd's configuration file by adding nydus proxy plugin section with /run/containerd-nydus/containerd-nydus-grpc.sock. This .sock file is listened to by nydus-snapshotter Pod in the daemonset.
However, after a machine restart, the nydus-snapshotter Pod in the daemonset hasn't started yet, so the .sock file does not exist. Meanwhile, the startup of the nydus-snapshotter Pod also depends on this .sock file. This looks like a circular dependency.
Is this considered as a known issue? Or, is there any suggested work around for this?
Thanks.
This is my error log from kubelet:
Sep 28 15:51:07 vm-10-249-100-37 kubelet[847]: E0928 15:51:07.387708 847 pod_workers.go:1301] "Error syncing pod, skipping" err="failed to "CreatePodSandbox" for "nydus-snapshotter-g69fg_nydus-system(215290b7-3a65-4e50-b30b-7f31f67ff7bc)" with CreatePodSandboxError: "Failed to create sandbox for pod \"nydus-snapshotter-g69fg_nydus-system(215290b7-3a65-4e50-b30b-7f31f67ff7bc)\": rpc error: code = Unavailable desc = failed to start sandbox \"611cd65c8c312e7ef26a2ab3be636cc62dbd29023f47f5ac715e86cef2c355bc\": failed to get sandbox image \"registry.aliyuncs.com/google_containers/pause:3.10\": failed to pull image \"registry.aliyuncs.com/google_containers/pause:3.10\": failed to pull and unpack image \"registry.aliyuncs.com/google_containers/pause:3.10\": failed to stat snapshot sha256:d8bdedd33a4e22f5cb6ca641a7ada5f9e6a92ad001589480818a136184c7ce34: connection error: desc = \"transport: Error while dialing: dial unix:///run/containerd-nydus/containerd-nydus-grpc.sock: timeout\""" pod="nydus-system/nydus-snapshotter-g69fg" podUID="215290b7-3a65-4e50-b30b-7f31f67ff7bc"