Thank you both for the feedback. Wolfgang, I was actually using an incarnation of your script the other day :) Thanks for that. I'll probably backup all my volumes the way Wolfgang does (because it's so simple), but also backup from within the vm on the vms where I have databases. To have a backup for the backup.
Dave Spano
From: "Wolfgang Hennerbichler" <wolfgang.hennerbichler@xxxxxxxxxxxxxxxx>
To: "Sebastien Han" <sebastien.han@xxxxxxxxxxxx>
Cc: "Dave Spano" <dspano@xxxxxxxxxxxxxx>, ceph-users@xxxxxxxxxxxxxx
Sent: Tuesday, April 9, 2013 5:00:41 AM
Subject: Re: [ceph-users] Question about Backing Up RBD Volumes in Openstack
On Tue, Apr 09, 2013 at 10:09:11AM +0200, Sebastien Han wrote:
> So the memory is _not_ saved, only the disk is. Note that it's always hard to make consistent snapshot. I assume that freezing the filesystem itself is the only solution to have a consistent snapshot, and still this doesn't mean that your application's data are consistent , in term of commits for instance with a database. I don't know which applications do you run but at the end, you might consider to stop some daemon as well, freeze the fs (to be sure), take your snapshot, unfreeze, start your daemon. You could reach 100% consistency with this method… I guess :)
I believe in journaling filesystems and their ability to recover from 'crashes' like these. So what I do is just take a snapshot of a running vm. If you re-mount that snapshot (e. g. in a rbd child) the journal replay assures a consistent filesystem. Database systems are designed to recover from power loss, this is nothing different than a power loss. And pg_dump or mysql_dumps (or whatever_dump) are a good habit anyway, if the database is small enough.
I've written up how I do ceph backups here:
http://www.wogri.at/Ceph-VM-Backup.339.0.html
Be aware that disabling rbd_cache in libvirt for now is more safe in terms of stability than enabling it (patches are out there but not in the OS packages yet).
> Cheers.
my 2c
Wolfgang
--
http://www.wogri.com
To: "Sebastien Han" <sebastien.han@xxxxxxxxxxxx>
Cc: "Dave Spano" <dspano@xxxxxxxxxxxxxx>, ceph-users@xxxxxxxxxxxxxx
Sent: Tuesday, April 9, 2013 5:00:41 AM
Subject: Re: [ceph-users] Question about Backing Up RBD Volumes in Openstack
On Tue, Apr 09, 2013 at 10:09:11AM +0200, Sebastien Han wrote:
> So the memory is _not_ saved, only the disk is. Note that it's always hard to make consistent snapshot. I assume that freezing the filesystem itself is the only solution to have a consistent snapshot, and still this doesn't mean that your application's data are consistent , in term of commits for instance with a database. I don't know which applications do you run but at the end, you might consider to stop some daemon as well, freeze the fs (to be sure), take your snapshot, unfreeze, start your daemon. You could reach 100% consistency with this method… I guess :)
I believe in journaling filesystems and their ability to recover from 'crashes' like these. So what I do is just take a snapshot of a running vm. If you re-mount that snapshot (e. g. in a rbd child) the journal replay assures a consistent filesystem. Database systems are designed to recover from power loss, this is nothing different than a power loss. And pg_dump or mysql_dumps (or whatever_dump) are a good habit anyway, if the database is small enough.
I've written up how I do ceph backups here:
http://www.wogri.at/Ceph-VM-Backup.339.0.html
Be aware that disabling rbd_cache in libvirt for now is more safe in terms of stability than enabling it (patches are out there but not in the OS packages yet).
> Cheers.
my 2c
Wolfgang
--
http://www.wogri.com
_______________________________________________ ceph-users mailing list ceph-users@xxxxxxxxxxxxxx http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
