Stefan G. Weichinger wrote:
Because we are fearing problems?
We do a vm snapshot to a backup server. Amanda backs up the snapshot
from the backup server.
And it is reliable?
We've never had a problem restoring machines. We often take a snapshot
of a production machine and turn it into a development machine.
By taking a snapshot everything is frozen as is. So what can you lose?
Any program running on the machine will continue to run once the machine
is back on. So the only thing you could lose is if data was coming from
a remote host. But if you're shutting down, you'll lose that
information anyways.
I guess if you had a program which was timing sensitive, it might not
like being suspended. I don't know of any like that, and it should be
easy to test.
VMware's entire backup offerings center around snapshots. Rarely a
snapshot will fail, but it doesn't effect the running vm. Even more
rarely the running vm will get stuck and needs to be power cycled just
like a normal server locking up.
I agree, it's more elegant and also faster, but the VMs there are pretty
important, so the first draft is to stop and start them.
How does your routine look like?
The command is different on different versions of VMware. On esx 2.5 its:
mount /mnt/vmimages #In fstab its a remote share - some people save to a
local partition but we don't have the space.
/usr/sbin/vmsnap.pl -f rl -c /home/vmware/cyrus/cyrus.vmx -d
/mnt/vmimages -l #This is a script from vmware, uses the perl api. One
esx 3 this line is different.
umount /mnt/vmimages
Then on that remote server the mount point is /files/vmimages/ The
snapshot script creates a directory for the image being backed up. So
in the amanda server I create a DLE for /files/vmimages/cyrus Amanda
is scheduled to run a couple hours after the snapshots run.
I don't automatically add things under /files/vmimages to the disklist,
but I do have a script which greps the disklist to and tells me if
something isn't being backed up.
