…
Hello all.
I'm working on improving libvirt usage in VDSM. Rationale is that libvirt driver has global lock, that holds us in any operation on many VMs simultaneously. Several small patches are prepared, so I'm starting to commit them.
The first one [1] removes UUIDString check in VM startup. It was used to test local libvirt for the bug 603494 [2], which is closed so long time ago, so there is no need to support this case, fixed already all across the distributives, including RHEL 6.x.
I wouldn't care, but in mass VM startup (dozens and hundreds of VMs) each extra call costs. So thanks for review/merge.
[1] — http://gerrit.ovirt.org/#/c/11054/ [2] — https://bugzilla.redhat.com/show_bug.cgi?id=603494
On Tue, Jan 15, 2013 at 05:08:59PM +0100, Peter V. Saveliev wrote:
…
Hello all.
I'm working on improving libvirt usage in VDSM. Rationale is that libvirt driver has global lock, that holds us in any operation on many VMs simultaneously. Several small patches are prepared, so I'm starting to commit them.
FYI the global lock will be dead in either the next libvirt release, or the one after. So I wouldn't jump through too many hoops in VDSM for this IMHO.
The first one [1] removes UUIDString check in VM startup. It was used to test local libvirt for the bug 603494 [2], which is closed so long time ago, so there is no need to support this case, fixed already all across the distributives, including RHEL 6.x.
I wouldn't care, but in mass VM startup (dozens and hundreds of VMs) each extra call costs. So thanks for review/merge.
The UUIDString() method executes purely client side so does not involve any libvirt locks.
Daniel
----- Original Message -----
From: "Daniel P. Berrange" berrange@redhat.com To: "Peter V. Saveliev" peet@redhat.com Cc: "VDSM Project Development" vdsm-devel@lists.fedorahosted.org Sent: Tuesday, January 15, 2013 5:35:11 PM Subject: Re: [vdsm] libvirt interface usage improvements, #1
On Tue, Jan 15, 2013 at 05:08:59PM +0100, Peter V. Saveliev wrote:
…
Hello all.
I'm working on improving libvirt usage in VDSM. Rationale is that libvirt driver has global lock, that holds us in any operation on many VMs simultaneously. Several small patches are prepared, so I'm starting to commit them.
FYI the global lock will be dead in either the next libvirt release, or the one after. So I wouldn't jump through too many hoops in VDSM for this IMHO.
Good news! Keep up the good work ;-)
The first one [1] removes UUIDString check in VM startup. It was used to test local libvirt for the bug 603494 [2], which is closed so long time ago, so there is no need to support this case, fixed already all across the distributives, including RHEL 6.x.
I wouldn't care, but in mass VM startup (dozens and hundreds of VMs) each extra call costs. So thanks for review/merge.
The UUIDString() method executes purely client side so does not involve any libvirt locks.
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 :| _______________________________________________ vdsm-devel mailing list vdsm-devel@lists.fedorahosted.org https://lists.fedorahosted.org/mailman/listinfo/vdsm-devel
vdsm-devel@lists.stg.fedorahosted.org
-
asegurap＠redhat.com -
Daniel P. Berrange -
peet＠redhat.com