Open Stack

I've proposed [1], which I think should solve the issue. Could you
test with that patch and let us know if the bug goes away?

Thank again for helping improve this!

[1] https://review.opendev.org/#/c/684409/

On Tue, Sep 24, 2019 at 5:04 AM wang.ya <wang.ya at 99cloud.net> wrote:
>
> I think the two issues should be similar. As I said, the first instance live migrate to host, but in resource tracker, the cache 'cn' not updated, at the moment, second instance live migrate to same host, then the vCPU pin policy broken.
> The issue is not reproducible every time, it need to go through multiple live migrate (I wrote a script to run live migrate automatic).
>
> I have checked the nova's config, the ' max_concurrent_live_migrations' option is default :)
>
> I've report the issue to launchpad, you can find the log in attachment: https://bugs.launchpad.net/nova/+bug/1845146
>
>
> On 2019/9/20, 11:52 PM, "Matt Riedemann" <openstack-discuss-bounces+wang.ya=99cloud.net at lists.openstack.org on behalf of mriedemos at gmail.com> wrote:
>
>     On 9/17/2019 7:44 AM, wang.ya wrote:
>     > But if add the property “hw:cpu_policy='dedicated'”, it will not correct
>     > after serval live migrations.
>     >
>     > Which means the live migrate can be success, but the vCPU pin are not
>     > correct(two instance have serval same vCPU pin on same host).
>     >
>
>     Is the race you're describing the same issue reported in this bug?
>
>     https://bugs.launchpad.net/nova/+bug/1829349
>
>     Also, what is the max_concurrent_live_migrations config option set to?
>     That defaults to 1 but I'm wondering if you've changed it at all.
>
>     --
>
>     Thanks,
>
>     Matt
>
>
>
>
>
>


-- 
Artom Lifshitz
Software Engineer, OpenStack Compute DFG