[ovirt-users] Hosted Engine Migration fails

Roy Golan rgolan at redhat.com
Wed Feb 18 11:12:08 UTC 2015


On 02/16/2015 04:55 AM, Soeren Malchow wrote:
>
> Dear all,
>
> we ahve a setup with several hosts running fedora 10 with the 
> virt-preview packages installed (for snapshot live merge) and a hosted 
> engine running centos 6.6-
>
> We are experiencing a problem with the Live Migration of the Hosted 
> Engine, in the case of setting the host for the Engine into 
> maintenance as well as a manual migration.
>
> I tried this on the “ovirtmgmt” network and when that failed I did 
> some research and tried to use another network interface (separate 
> from ovirtmgmt on layer 2), this also fails.
>
> It looks as if the migration is still going through the ovirtmgmt 
> interface, at least judging from the network traffic, and I think the 
> error that I found (RHBZ#919201) is actually the right one.
>
>  --
>
> vdsm.log:Thread-6745::WARNING::2015-02-16 
> 03:22:25,743::migration::458::vm.Vm::(monitor_migration) 
> vmId=`XXX`::Migration stalling: remaining (35MiB) > lowmark (15MiB). 
> Refer to RHBZ#919201.
>
> vdsm.log:Thread-6745::WARNING::2015-02-16 
> 03:22:35,745::migration::458::vm.Vm::(monitor_migration) 
> vmId=`XXX`::Migration stalling: remaining (129MiB) > lowmark (15MiB). 
> Refer to RHBZ#919201.
>
> vdsm.log:Thread-6745::WARNING::2015-02-16 
> 03:22:45,747::migration::458::vm.Vm::(monitor_migration) 
> vmId=`XXX`::Migration stalling: remaining (42MiB) > lowmark (15MiB). 
> Refer to RHBZ#919201.
>
> vdsm.log:Thread-6745::WARNING::2015-02-16 
> 03:22:55,749::migration::458::vm.Vm::(monitor_migration) 
> vmId=`XXX`::Migration stalling: remaining (88MiB) > lowmark (15MiB). 
> Refer to RHBZ#919201.
>
> --
>
> The ovirtmgmt interface is 2 x 1Gbit (LACP connected to Dell Switches 
> with MLAG) and by far not fully utilized.
>
> Can anyone help me where to go form here ?
>

still need the whole log. probably the guest (engine) is doing lots of 
memory i/o if you got a fair amount of running VMs and Hosts. that will 
stalls the migration because
the guest pages are getting dirty faster than qemu can copy.


you have 2 options:

1. try several more times.

2. shutdown the engine vm, it should start on another host

> Regards
>
> Soeren
>
>
>
> _______________________________________________
> Users mailing list
> Users at ovirt.org
> http://lists.ovirt.org/mailman/listinfo/users

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.ovirt.org/pipermail/users/attachments/20150218/eab1835e/attachment-0001.html>


More information about the Users mailing list