There is actually documentation supporting my view:
https://www.freeipa.org/page/Backup_and_Restore Look particularly at the section "Why snapshot and not backup and restore scripts?"
The difference is that they suggest stopping a replica before making a snapshot, while we snapshot a running system. I’ve done this with a variety of databases and other applications. My claim is that a point in time snapshot should be safe for any software that is designed to survive a crash, because a point in time snapshot is no harder to recover from than a crash. We have multiple snapshots, in case we can’t use one of them. But I’ve never seen that happen.
We always run complex software systems such a ipa in a VM.
Thanks Charles. Out IPA is a VM too on Openstack but for some reasons they said it's not good to take snapshots and rely on that for backups... I ll investigate further tho... cause my idea was exactly that. Snapshots!!!!
Thanks for sharing.
Our IPA servers are VMs. We do backups of snapshots, either through VMware or when the image is on a Netapp, through a Netapp snapshot. That guarantees that you have all the pieces in a consistent state. I’ve never had to restore a production server, but I have started copies of one of the backups to do experiments that I didn’t want to do on a production system. I’ve never had an issue starting from a backup, though I need to do some changes so the system thinks it has the same hostname as the original one.
--