Your customers may want to take a snapshot of their VMs before performing kURL upgrades, but there are a few things to bear in mind while taking snapshots.
-
Always power off all VMs in the cluster within 5 minutes of eachother before taking shapshots.
- this is to ensure filesystem integrity during the snapshot.
-
Snapshotting isn’t always supported on all hypervisors.
- vmware / esxi snapshots are not a supported method of ensuring rollback points for VMs, VMware’s own documentation is clear on this, and it’s suggested instead to use a proper backup solution.
- snapshots in KVM based hypervisors such as xen, libvirt or openstack are fine as long as you’re using the “internal” snapshot for all disks.
-
all disks must be included in any snapshots or backups.
- admins of kURL installs with storage providers that use additional disks such as rook will have to take extra care to ensure every disk attached to every node is included in backups.
-
In the event of a restore, all nodes must be restored to their respective snapshots for that time.
- this should prevent state drift for the cluster and inconsistency in stored data.