On 02/02/2011, at 4:23 AM, Daniel P. Berrange wrote:
On Tue, Feb 01, 2011 at 08:45:41AM -0800, Ed Swierk wrote:
> When troubleshooting a libvirt system the most obvious question I face
> is when a given VM started and stopped.
>
> The only way I've been able to obtain this information is by setting
> log_level = 1 in libvirtd.conf, which turns on the firehose,
> generating an enormous amount of output. I then have to sift through
> the logs for magic strings like "virExecWithHook" to see when a VM
> started, and "Shutting down VM" to see when a VM stopped. Is there no
> concise message that indicates a VM starting up? "virExecWithHook"
> doesn't exactly jump out.
>
> I'd consider starting up and shutting down a VM to be at least worth
> an info-level message in the log. I monkeyed with the log_filters
> setting but I couldn't figure out how to filter those two strings. Any
> hints?
In latest libvirt we write to the audit logs when a VM starts or
stop, and upon hotplug of nics & disks (see /var/log/audit/audit.log)
This sounds like the kind of thing that could be monitored pretty effectively
with the hook scripts too.
http://justinclift.fedorapeople.org/libvirt_website/hooks.html
(sorry for the non
libvirt.org URL, main site appears to be down atm)
I'm thinking "add a /etc/libvirt/hooks/qemu script" and make it output
updated
vm state somewhere useful to you?