Fix for cluster-reset backup from s3 when etcd snapshots are disabled

[vitorsavian]

Proposed Changes

If cluster-reset was called with etcd-disable-snapshots , k3s will verify if the user have etcd-s3 and then restore the snapshot from S3.

Types of Changes

• Fix

Verification

Create a config.yaml in the /etc/rancher/k3s with this flags

cluster-init: true
etcd-disable-snapshots: true
etcd-s3: true
etcd-s3-access-key: 'secretKey'
etcd-s3-secret-key: 'secretKey'
etcd-s3-bucket: 'bucketName'
etcd-s3-region: 'awsRegion'
etcd-s3-timeout: '90s'

Now you can create the server

k3s server 

Then you need to save the snapshot, that will backup to s3

k3s etcd-snapshot save

Turn off the k3s and then

k3s server --cluster-reset --cluster-reset-restore-path="s3SnapshotName"

Testing

Linked Issues

• --etcd-disable-snapshots flag usage leads to unexpected behavior #8031

User-Facing Change

Further Comments

With this change, users can now manage and backup their own snapshots on demand from S3.

[dereknola]

This could likely be solved even easier by just changing the original if statement to:

if !cfg.EtcdDisableSnapshots  || cfg.ClusterReset {
}

So we always fill the etcd snapshot values if we see a clusterReset Flag