On 07/01/2011 08:07 AM, Jiri Denemark wrote:
It has two modes, normal and interactive. In normal mode, messages
are
passed from libvirt to qemu and back as they come and copyied to stdout
so that one can monitor the communication.
Cool! But post-0.9.3, so I haven't reviewed the patch itself yet.
To start monitor-proxy one needs to follow the following steps:
- start a domain
- run monitor-proxy [-i] <monitor socket path>
for example: monitor-proxy /var/lib/libvirt/qemu/vm1.monitor
- restart libvirtd (this makes libvirtd disconnect from qemu and
reconnect to the proxy)
Sounds like it would be useful to add a libvirt API (perhaps under the
qemu namespace, similar to our existing 'virsh qemu-monitor-command'
extensions, and thus not officially supported) to force libvirt to
reconnect to a monitor socket without having to restart all of libvirtd.
monitor-proxy supports libvirtd restarts, i.e., if libvirtd
disconnects,
the proxy just waits until it reconnects again. Stopping the proxy
while libvirtd is running kills the domain since libvirtd sees monitor
EOF. To stop using monitor-proxy without killing the domain, one needs
to stop libvirtd, stop monitor-proxy, and start libvirtd.
Again, a libvirt monitor-reconnect command would be handy.
--
Eric Blake eblake(a)redhat.com +1-801-349-2682
Libvirt virtualization library
http://libvirt.org