On Wed, Jan 09, 2013 at 11:38:07PM +0200, Zeeshan Ali (Khattak) wrote:
Hi,
Too lazy to explain again so I'll just c&p conversation I had about
the issue with Eric on IRC:
<zeenix> is timeouts on session libvirt (or in general) something new?
<zeenix> libvirtd startup seems to now take a few seconds so Boxes UI
is all empty for that amount of time
<eblake> it's been around for a while; qemu.conf has settings to
control how long it should take
<zeenix> unless Boxes is launched before libvirtd timesout
<eblake> but there is a relatively new patch that makes the libvirtd
-t actually be useful for qemu:///session
<zeenix> don't think i ever changed that, at least not recently
<zeenix> that could be it
<zeenix> i only saw it happening after i update libvirtd (after a month)
<eblake> if you're seeing some delays in starting libvirtd, those are
probably worth fixing
<zeenix> s/update/recently updated/
<zeenix> yeah
<zeenix> i'll try to write a test app
<zeenix> no need, i can reproduce with virsh :)
[zeenix@z-laptop virt]$ time virsh list
Id Name State
----------------------------------------------------
real 0m6.349s
user 0m0.016s
sys 0m0.010s
[zeenix@z-laptop virt]$ time virsh list
Id Name State
----------------------------------------------------
real 0m0.024s
user 0m0.010s
sys 0m0.011s
If I had to guess I'd blame the QEMU capabilities extraction.
I have a feeling probing via QMP is slower than just parsing
-help output. Though we cache the data, so in the future it
is much faster.
If this is correct diagnosis, perhaps we need to switch over
to a parse-on-first-use approach rather than on startup
Daniel
--
|:
http://berrange.com -o-
http://www.flickr.com/photos/dberrange/ :|
|:
http://libvirt.org -o-
http://virt-manager.org :|
|:
http://autobuild.org -o-
http://search.cpan.org/~danberr/ :|
|:
http://entangle-photo.org -o-
http://live.gnome.org/gtk-vnc :|