Good point, we'd have to modify the destination xml, which is kind of hairy.
I actually cant reproduce this on ubuntu 12.04 and the 4.2 RC. I mean, 'virsh pool-list' is blank as the bug reports, but so what? Is there a separate bug we're discussing now? I created nfs primary storage, launched a vm on it, then restarted libvirt. I verified the pools were erased via 'virsh pool-list', and then attempted to launch a new vm with the nfs storage tag. It registered in Libvirt and the VM is now running. My intent was to try a "umount -l", expecting that the existing VM's open filehandles would stay working, and it would allow a fresh mount of the primary storage, but there was no issue observed. Maybe I'm missing the root of the issue reported. root@devcloud-kvm-u:~# virsh pool-list Name State Autostart ----------------------------------------- 2fe9a944-505e-38cb-bf87-72623634be4a active no <--- nfs primary 609d4339-e66a-4298-909d-74dca7205a7b active no vg0 active no <--- clvm storage root@devcloud-kvm-u:~# /etc/init.d/libvirt-bin restart libvirt-bin stop/waiting libvirt-bin start/running, process 7652 root@devcloud-kvm-u:~# virsh pool-list Name State Autostart ----------------------------------------- ... wait for new vm to start root@devcloud-kvm-u:~# virsh pool-list Name State Autostart ----------------------------------------- 2fe9a944-505e-38cb-bf87-72623634be4a active no <--- nfs primary is back ... launch clvm storage-based vm, wait for it to start root@devcloud-kvm-u:~# virsh pool-list Name State Autostart ----------------------------------------- 2fe9a944-505e-38cb-bf87-72623634be4a active no 609d4339-e66a-4298-909d-74dca7205a7b active no vg0 active no <--- clvm On Mon, Sep 16, 2013 at 2:15 PM, Wido den Hollander <[email protected]> wrote: > > > On 09/16/2013 07:16 PM, Marcus Sorensen wrote: >> >> I agree, it makes things more complicated. However, updating libvirt >> should be a maintenance mode thing; I've seen it lose track of VMs >> between major libvirt upgrades and have to kill/restart all vms to >> reregister them with libvirt, best to be in maintenance. >> >> If I remember right, the issues with persistent storage definitiions >> were primarily related to failure scenarios, things like hard power >> off of host, host would build up dozens of primary storage definitions >> that it didn't need, creating unnecessary dependencies and potential >> unnecessary outages (see other email about what happens when storage >> goes down). >> >> If edison has a patch then that's great. One idea I head was that if >> the pool fails to register due to mountpoint already existing, we >> could use an alternate mount point to register it. Old stuff will keep >> working, new stuff will continue. > > > No! That will break migrations to machines which don't have that mountpoint. > Backing devices with QCOW2 will also break on the longer run. > > Please do not do that. > > Wido > > >> >> On Mon, Sep 16, 2013 at 3:43 AM, Wei Zhou (JIRA) <[email protected]> wrote: >>> >>> >>> [ >>> https://issues.apache.org/jira/browse/CLOUDSTACK-3565?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13768197#comment-13768197 >>> ] >>> >>> Wei Zhou commented on CLOUDSTACK-3565: >>> -------------------------------------- >>> >>> Marcus, >>> >>> The change of the host to maintenance one by one makes upgrade/change >>> complicated. >>> Edison has changed the source to support re-creating storage pool in the >>> fix of CLOUDSTACK-2729. >>> However, the test failed in my environment. >>> >>> By the way, I did not see any other issue when I used cloudstack 4.0 >>> while the storage pools are persistent. >>> I do not know what issue will be leader by it. >>> >>>> Restarting libvirtd service leading to destroy storage pool >>>> ----------------------------------------------------------- >>>> >>>> Key: CLOUDSTACK-3565 >>>> URL: >>>> https://issues.apache.org/jira/browse/CLOUDSTACK-3565 >>>> Project: CloudStack >>>> Issue Type: Bug >>>> Security Level: Public(Anyone can view this level - this is the >>>> default.) >>>> Components: KVM >>>> Affects Versions: 4.2.0 >>>> Environment: KVM >>>> Branch 4.2 >>>> Reporter: Rayees Namathponnan >>>> Assignee: Marcus Sorensen >>>> Priority: Blocker >>>> Labels: documentation >>>> Fix For: 4.2.0 >>>> >>>> >>>> Steps to reproduce >>>> Step 1 : Create cloudstack step in kvm >>>> Step 2 : From kvm host check "virsh pool-list" >>>> Step 3: Stop and start libvirtd service >>>> Step 4 : Check "virsh pool-list" >>>> Actual Result >>>> "virsh pool-list" is blank after restart libvird service >>>> [root@Rack2Host12 agent]# virsh pool-list >>>> Name State Autostart >>>> ----------------------------------------- >>>> 41b632b5-40b3-3024-a38b-ea259c72579f active no >>>> 469da865-0712-4d4b-a4cf-a2d68f99f1b6 active no >>>> fff90cb5-06dd-33b3-8815-d78c08ca01d9 active no >>>> [root@Rack2Host12 agent]# service cloudstack-agent stop >>>> Stopping Cloud Agent: >>>> [root@Rack2Host12 agent]# virsh pool-list >>>> Name State Autostart >>>> ----------------------------------------- >>>> 41b632b5-40b3-3024-a38b-ea259c72579f active no >>>> 469da865-0712-4d4b-a4cf-a2d68f99f1b6 active no >>>> fff90cb5-06dd-33b3-8815-d78c08ca01d9 active no >>>> [root@Rack2Host12 agent]# virsh list >>>> Id Name State >>>> ---------------------------------------------------- >>>> [root@Rack2Host12 agent]# service libvirtd stop >>>> Stopping libvirtd daemon: [ OK ] >>>> [root@Rack2Host12 agent]# service libvirtd start >>>> Starting libvirtd daemon: [ OK ] >>>> [root@Rack2Host12 agent]# virsh pool-list >>>> Name State Autostart >>>> ----------------------------------------- >>> >>> >>> -- >>> This message is automatically generated by JIRA. >>> If you think it was sent incorrectly, please contact your JIRA >>> administrators >>> For more information on JIRA, see: http://www.atlassian.com/software/jira
