-
Environment DetailsBeen managing several Version info:
Expected BehaviorClusters shouldn't overwhelm filesystem by storing multiple progressively heavy snapshots. Observed BehaviorSome clusters tend to house multiple heavy snapshots in the following location:
The snapshots-directory contains To avoid overwhelmed/full disk we restart the cluster-agents. After agents rejoin the k3s cluster, the pods come up very slowly and the Question ?
|
Beta Was this translation helpful? Give feedback.
Replies: 1 comment 1 reply
-
This is really more of a containerd question, not k3s... This is where containerd stores filesystems for running containers when using the You can reference the upstream docs for more information: https://github.com/containerd/containerd/blob/main/docs/snapshotters/README.md |
Beta Was this translation helpful? Give feedback.
This is really more of a containerd question, not k3s...
This is where containerd stores filesystems for running containers when using the
native
snapshotter. This is an inefficient snapshotter; generally theoverlayfs
snapshotter is preferred, but that requires an underlying filesystem that supports overlayfs, and it sounds like you've configured containerd to use the native snapshotter instead.You can reference the upstream docs for more information: https://github.com/containerd/containerd/blob/main/docs/snapshotters/README.md