I was successful in doing a cassandra-medusa backup of a docker-based (non k8ssandra) one-node cluster into Azure Blobstore. I also configured k8s (v1.3.0) for Azure blobstore and can successfully do a test backup and a restore as described in the docs.
But when I try to restore the real, “alien” backup, it creates the restore object and does not start the restore. In the logs of the …sts-0 pod he complains that the indicated backup is not there.
I am unsure how to specify the name of the backup.
The folder structure of the test backup/restore done in k8ssandra is:
<bucket_name>/prod-k8ssandra-dc1-rack1-sts-0/backup1
<bucket_name>/prod-k8ssandra-dc1-rack1-sts-0/data
I take “backup1” as the backup name and everything is fine.
The folder structure of the “alien” backup is:
<bucket_name>/customer/a2e71349da8f/202107051529
<bucket_name>/customer/a2e71349da8f/data
when trying
helm install restore-customer k8ssandra/restore --set name=restore-customer,backup.name=a2e71349da8f,cassandraDatacenter.name=dc1
it says that there is no backup a2e71349da8f.
when trying
helm install restore-customer k8ssandra/restore --set name=restore-customer,backup.name=202107051529,cassandraDatacenter.name=dc1
it says
“spec.backup: Invalid value: “integer”: spec.backup in body must be of type string: “integer””
So I have these questions:
- Can the restore of a backup, which was not made on the same cluster, work at all ?
- If yes, how do I correctly specify the backup name ?
- Is there a way to execute plain medusa CLI commands to e.g. list the existing backups ? Something like kubectl exec podname -c containername – medusa list
Regards
Peter