[openstack-dev] instance migration strangeness in devstack

Vishvananda Ishaya vishvananda at gmail.com
Thu Jan 16 16:29:09 UTC 2014


In that case, this sounds like a bug to me related to lvm volumes. You should check the nova-compute.log from both hosts and the nova-conductor.log. If it isn’t obvious what the problem is, you should open a bug and attach as much info as possible.

Vish

On Jan 16, 2014, at 8:04 AM, Dan Genin <daniel.genin at jhuapl.edu> wrote:

> Thank you for replying, Vish. I did sync and verified that the file was written to the host disk by mounting the LVM volume on the host.
> 
> When I tried live migration I got a Horizon blurb "Error: Failed to live migrate instance to host" but there were no errors in syslog.
> 
> I have been able to successfully migrate a Qcow2 backed instance.
> Dan
> 
> On 01/16/2014 03:18 AM, Vishvananda Ishaya wrote:
>> This is probably more of a usage question, but I will go ahead and answer it.
>> 
>> If you are writing to the root drive you may need to run the sync command a few times to make sure that the data has been flushed to disk before you kick off the migration.
>> 
>> The confirm resize step should be deleting the old data, but there may be a bug in the lvm backend if this isn’t happening. Live(block) migration will probably be a bit more intuitive.
>> 
>> Vish
>> On Jan 15, 2014, at 2:40 PM, Dan Genin <daniel.genin at jhuapl.edu> wrote:
>> 
>>> I think this qualifies as a development question but please let me know if I am wrong.
>>> 
>>> I have been trying to test instance migration in devstack by setting up a multi-node devstack following directions at http://devstack.org/guides/multinode-lab.html. I tested that indeed there were multiple availability zones and that it was possible to create instances in each. The I tried migrating an instance from one compute node to another using the Horizon interface (I could not find a way to confirm migration, which is a necessary step, from the command line). I created a test file on the instance's ephemeral disk, before migrating it, to verify that the data was moved to the destination compute node. After migration, I observed an instance with the same id active on the destination node but the test file was not present.
>>> 
>>> Perhaps I misunderstand how migration is supposed to work but I expected that the data on the ephemeral disk would be migrated with the instance. I suppose it could take some time for the ephemeral disk to be copied but then I would not expect the instance to become active on the destination node before the copy operation was complete.
>>> 
>>> I also noticed that the ephemeral disk on the source compute node was not removed after the instance was migrated, although, the instance directory was. Nor was the disk removed after the instance was destroyed. I was using LVM backend for my tests. 
>>> 
>>> I can provide more information about my setup but I just wanted to check whether I was doing (or expecting) something obviously stupid.
>>> 
>>> Thank you,
>>> Dan
>>> _______________________________________________
>>> OpenStack-dev mailing list
>>> OpenStack-dev at lists.openstack.org
>>> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
>> 
> 
> _______________________________________________
> OpenStack-dev mailing list
> OpenStack-dev at lists.openstack.org
> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack-dev/attachments/20140116/e861a4b0/attachment.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 455 bytes
Desc: Message signed with OpenPGP using GPGMail
URL: <http://lists.openstack.org/pipermail/openstack-dev/attachments/20140116/e861a4b0/attachment.pgp>


More information about the OpenStack-dev mailing list