On 2013-08-31 04:04, Mark Chaney wrote:
If I am going to use a Ceph RBD cluster to store my kvm instances, how
should I be doing backups?
1) I would prefer them to be incremental so that a whole backup doesnt
have to happen every night.
2) I would also like the instances to obviously stay online during the
backup
3) Backups will be stored off the ceph cluster on slower sata drives
on another storage server.
I would rather imagine it depends on _why_ you're backing up; is it to
do a full restore, or do you also want to support retrieval of
individual files?
If all you require is full system restore, taking snapshots of the block
storage, and incrementally storing the differences seems to be the most
straight-forward way to proceed. If you take high-frequency backups (to
allow for rolling back only the last n minutes of processing), but only
permanently store daily backups for the last week, weekly for the last
month, etc. It would probably make sense to leave the high-frequency
snapshots on ceph, and only transfer backup data off the cluster once a day.
If you need to restore individual files (I would imagine that this is
quite likely), then you'd use the same backup facility as you would have
if the VMs were instead distinct physical machines being backed up
across the network, ie. bacula, backuppc, etc.
There could be a certain amount to gain from taking a snapshot, then
attaching that to your backup VM, so that doing a backup doesn't cause
extra load to be placed on the working VMs, but your ability to be able
to do that would depend on what backup engine you're using.
--
Martin Rudat
_______________________________________________
ceph-users mailing list
ceph-users@xxxxxxxxxxxxxx
http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
