Description of problem: Use virsh command to migrate fail on xen i386 while can migrate success with virt-manager Version-Release number of selected component (if applicable): - RHEL5.4-Client-i386 - 2.6.18-158.el5xen - xen-3.0.3-92.el5 - libvirt-0.6.3-17.el5 How reproducible: Always Steps to Reproduce: On both machine 1.# virsh pool-define pool // (attach pool) 2.# virsh pool-start pool 3.# iptables -F 4.# setenforce 0 5.Config /etc/xen/xend-config.sxp On source machine 1.# virsh define guest // (attach guest) 2.# virsh start test1 3.# virsh migrate --live test1 xen+ssh://10.66.70.188 Actual results: [root@dhcp-66-70-82 Desktop]# virsh migrate --live test1 xen+ssh://10.66.70.188 root.70.188's password: error: Domain not found: xenUnifiedDomainLookupByName And in source machine, test1 is reboot in target machine, cannot find test1 Expected results: Migrate successfully Additional info: Can migrate success in virt-manager Attach pool xml, guest xml, capabilities info, system version info and debug info.
Created attachment 355796 [details] capabilities info
Created attachment 355797 [details] guest xml
Created attachment 355798 [details] pool xml
Created attachment 355799 [details] verison info
Created attachment 355800 [details] debug info
The debug log here has this line: 18:36:50.765: debug : xenDaemonDomainMigratePerform:4431 : hostname = localhost.localdomain, port = 0 18:36:57.355: debug : xenDaemonDomainMigratePerform:4445 : migration done which says that XenD told libvirt that the migration operation was complete. This error error: Domain not found: xenUnifiedDomainLookupByName says that the domain does not exist in the target though. So either migration failed and XenD didn't tell libvirt there was an error. Or migration succeeeded, but the guest then crashed. I think we need the XenD logs to diagnose this - it doesn't sound like a libvirt bug currently. Can you provide the /var/log/xen/xend.log & xend-error.log files, from both source & target hosts.
Attach source-xend.log and source-xend-debug.log (whose content is null)
Created attachment 355907 [details] source xend log
Attach target-xend.log. Cannot attach source-xend-debug.log and target-xend-debug.log while both of them are null.
Created attachment 355908 [details] target xend log
Can migrate successfully with xm command: # xm migrate --live test1 10.66.70.188 And I notice that, after run command # virsh migrate --live test1 xen+ssh://10.66.70.188 the file /etc/xen/test1 is deleted, so cause an error: error: Domain not found: xenUnifiedDomainLookupByName Hope this info can be useful
> Actual results: > [root@dhcp-66-70-82 Desktop]# virsh migrate --live test1 > xen+ssh://10.66.70.188 > root.70.188's password: > error: Domain not found: xenUnifiedDomainLookupByName > > And in source machine, test1 is reboot > in target machine, cannot find test1 > While verifying bug 449696 I got a different error: [root@localhost ~]# virsh migrate --live testvm-F10 xen+ssh://192.168.0.11 root.0.11's password: error: POST operation failed: xend_post: error from xen daemon: (xend.err "can't connect: Name or service not known") However xm migrate worked fine with the same setup.
As xen migration itself works, problem is probably in way libvirt handle it.
I don't think we can cover this for 5.5 at this point, retargetting it for 5.6 Daniel
Reproduced this bug via virsh and virt-manager with the same error, while can successfully migrate via xm. RHEL5.6-Server-i386 - xen-3.0.3-116.el5 - kernel-xen-2.6.18-219.el5 - libvirt-0.8.2-4.el5 [root@dhcp-93-192 etc]# virsh migrate --live 13 xen+ssh://10.66.93.12 root.93.12's password: error: Domain not found: xenUnifiedDomainLookupByName Still, after executing the command, the file /etc/xen/rhel-5.5-32-hvm is deleted.
Created attachment 447907 [details] rhel56-source-xend.log
Created attachment 447908 [details] rhel56-source-xend-debug.log
Created attachment 447909 [details] rhel56-target-xend.log
Created attachment 447910 [details] rhel56-target-xend-debug.log
Migration works just fine for me on xen on both i386 and x86_64. Any chance I could get to the machines where you are seeing the issue?
For comment 26: The problem was missing DNS entry for target hostname. After adding the target machine to /etc/hosts on source host, migration finished successfully.
We agreed with the reporter who tested this on some additional machines that this was a bug in 0.6.3-based libvirt package but reproducing it with 0.8.2 was most likely caused by some configuration/installation issues on the only two machines where this bug was reproduced. My suspicion is this was fixed by the patches from bug 593120. I'm closing this as a duplicate. Feel free to reopen if you hit this bug again. *** This bug has been marked as a duplicate of bug 593120 ***