[libvirt-users] libvirtd loses all of its data after restart
Hi, I'm having a problem with libvirtd (backend being kvm) losing its if I restart it with /etc/init.d/libvirtd restart. What I did was: Start libvirtd Connect to it using virsh Create a new storage pool with pool-create-as Create some volumes with vol-create-as Create some virtual machines with create Restart libvirtd using /etc/init.d/libvirtd restart Reconnect with virsh After this, all the previously defined pools, volumes and virtual machines would no longer be listed with pool-list or list, respectively. (list --all doesn't show anything either) The virtual machines I had started earlier are still running and I can still connect to them using vnc, but obviously, I can no longer manage them using libvirt. Why is this happening? Is libvirtd's data not supposed to be persistent? Is this maybe just an old bug? How can I get the still running virtual machines back under control? The system I'm using for this is a recently installed CentOS 5.4 with libvirt-0.6.3-20.1.el5_4 installed directly from its package repository. The only changes I made from the vanilla install was to install and configure some SSL certs in /etc/libvirt/libvirtd.conf, as described on the libvirt website and to enable LIBVIRTD_ARGS="--listen" in /etc/sysconfig/libvirtd. Guido -- Too much multitasking isn't good for you.
On Wed, May 05, 2010 at 07:27:23PM +0200, guido wrote:
Hi,
I'm having a problem with libvirtd (backend being kvm) losing its if I restart it with /etc/init.d/libvirtd restart.
What I did was: Start libvirtd Connect to it using virsh Create a new storage pool with pool-create-as Create some volumes with vol-create-as Create some virtual machines with create Restart libvirtd using /etc/init.d/libvirtd restart Reconnect with virsh
After this, all the previously defined pools, volumes and virtual machines would no longer be listed with pool-list or list, respectively. (list --all doesn't show anything either) The virtual machines I had started earlier are still running and I can still connect to them using vnc, but obviously, I can no longer manage them using libvirt.
Why is this happening? Is libvirtd's data not supposed to be persistent? Is this maybe just an old bug? How can I get the still running virtual machines back under control?
The system I'm using for this is a recently installed CentOS 5.4 with libvirt-0.6.3-20.1.el5_4 installed directly from its package repository. The only changes I made from the vanilla install was to install and configure some SSL certs in /etc/libvirt/libvirtd.conf, as described on the libvirt website and to enable LIBVIRTD_ARGS="--listen" in /etc/sysconfig/libvirtd.
The version of libvirtd you have does not support seemless restarts of the daemon process while resources are active. You should shutdown all guests / storage pools before restarting the libvirtd daemon, or better yet, don't restart it at all. libvirt 0.7.x was the first series supporting restarts with stuff running Daniel -- |: Red Hat, Engineering, London -o- http://people.redhat.com/berrange/ :| |: http://libvirt.org -o- http://virt-manager.org -o- http://deltacloud.org :| |: http://autobuild.org -o- http://search.cpan.org/~danberr/ :| |: GnuPG: 7D3B9505 -o- F3C9 553F A1DA 4AC2 5648 23C1 B3DF F742 7D3B 9505 :|
Am Mittwoch, 5. Mai 2010 schrieben Sie:
The version of libvirtd you have does not support seemless restarts of the daemon process while resources are active. You should shutdown all guests / storage pools before restarting the libvirtd daemon, or better yet, don't restart it at all.
What if it crashes?
libvirt 0.7.x was the first series supporting restarts with stuff running
Okay, I've upgraded to libvirt 0.7, (this time on Fedora 12), but there are still problems. Running and defined virtual machines now persist across restarts, but storage pools and networks still disappear. What's worse, trying to define the networks that disappeared again (so that I can continue using them to create new vms) fails because the bridge by that and the dnsmasq process on that ip/port combination obviously already exist. Guido PS: Sorry, I originally sent the reply directly to you instead of the mailing list without realizing it. -- Too much multitasking isn't good for you.
participants (2)
-
Daniel P. Berrange -
guido