[libvirt] libvirt 3.9.0 locks (was libvirt 3.5.0 locks)

[Nikolay Shirokovskiy]

On 04.12.2017 18:46, Vasiliy Tolstov wrote:
> 2017-12-01 16:14 GMT+03:00 Nikolay Shirokovskiy <nshirokovskiy at virtuozzo.com>:
>> On 01.12.2017 15:26, Vasiliy Tolstov wrote:
>>> Hi again. I'm still try to determine why my libvirtd locks, this is
>>> another portion of gdb stuff:
>>> https://gist.github.com/vtolstov/ae8c4a67e15b2fbd14bbb95c226fb427
>>>
>>> error looks in logs like:
>>> Dec 01 12:35:49 cn04 libvirtd[1171]: 2017-12-01 09:35:49.637+0000:
>>> 28063: error : qemuDomainObjBeginJobInternal:4403 : Timed out during
>>> operation: cannot acquire state change lock (held by
>>> remoteDispatchConnectGetAllDomainStats)
>>> Dec 01 12:35:49 cn04 libvirtd[1171]: 2017-12-01 09:35:49.637+0000:
>>> 28063: warning : qemuDomainObjBeginJobInternal:4391 : Cannot start job
>>> (destroy, none) for domain 225560-20001; current job is (query, none)
>>> owned by (1180 remoteDispatchConnectGetAllDomainStats, 0 <null>) for
>>> (30s, 0s)
>>>
>>> and virsh list says that domain in shutdown state (but qemu process
>>> already exited).
>>>
>>> Please, help me to solve this issue.
>>>
>>
>> Hi, Vasiliy.
>>
>> From your stacktrace I guess this could be issue introduced in libvirt version 3.4.0
>> by aeda1b8c. I tried to address it in [1] but later decide to take a different approach
>> but don't undertake any since then. One can just revert the mentioned patch.
>>
>> [1] https://www.redhat.com/archives/libvir-list/2017-September/msg00172.html
>>
>> Nikolay
> 
> So what next that can i do to fix this? (And i'm happy if this fixed
> before 3.10, but may be is too late...)
> 

You can revert aeda1b8c in your downstream build until it is fixed upstream.