[ovirt-users] Info on live snapshot and agent interaction
Nir Soffer
nsoffer at redhat.com
Fri Jun 23 21:59:37 UTC 2017
On Thu, May 4, 2017 at 11:09 PM Michal Skrivanek <
michal.skrivanek at redhat.com> wrote:
>
> > On 4 May 2017, at 17:51, Gianluca Cecchi <gianluca.cecchi at gmail.com>
> wrote:
> >
> > Hello,
> > supposing to have a Linux VM with ovirt-guest-agent installed,
File system freezing is implemented using libvirt fsFreese/fsThaw, and is
implemented in the guest with qemu-guest-agent, not ovirt-guest-agent.
You can find the info on using it here:
https://access.redhat.com/documentation/en-US/Red_Hat_Enterprise_Linux/7/html/Virtualization_Deployment_and_Administration_Guide/sect-Using_the_QEMU_guest_virtual_machine_agent_protocol_CLI-libvirt_commands.html
> during a live snapshot operation it should be freeze of filesystems.
> > Where to find confirmation of correct/successful interaction?
>
> if it’s not successful there should be an event log message about that.
> And prior to taking the snapshot a warning in red at the bottom of the
> dialog (that check happens when you open the dialog, so it may not be 100%
> reliable)
>
> > /var/log/messages or agent log or other kind of files?
>
> if you want to doublecheck then this is noticable in vdsm.log. First we
> try to take the snapshot with fsfreeze, and only when it fails we take it
> again without it.
>
Since 3.6 we use fsFreeze/fsThaw explicitly, so we don't try snapshot twice.
In vdsm log you will find an INFO log before/after fsFreeze/fsThaw, or WARN
log if they failed.
> > Are there any limitations on filesystems that support freeze? Is it
> fsfreeze the command executed at VM OS level or any other low level command?
>
> It’s a matter of Linux and Windows implementation, they both have an API
> supporting that at kernel level. I’m not aware of filesystem limitations.
>
You should check qemu-guest-agent documentation.
Nir
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.ovirt.org/pipermail/users/attachments/20170623/933824cd/attachment.html>
More information about the Users
mailing list