How frequently is this occurring? How many labs is this occurring on? I haven't been able to reproduce yet (~10 compute lock/unlock attempts).
Around the time this is occurring in the latest logs, I see that there is a platform CPU alarm on compute-1, and the report_state RPCs timing out so neutron-server thinks some agents are dead. I expect this would not be seen if you increase the platform CPUs on the worker node to more than just 0,44.
How frequently is this occurring? How many labs is this occurring on? I haven't been able to reproduce yet (~10 compute lock/unlock attempts).
Around the time this is occurring in the latest logs, I see that there is a platform CPU alarm on compute-1, and the report_state RPCs timing out so neutron-server thinks some agents are dead. I expect this would not be seen if you increase the platform CPUs on the worker node to more than just 0,44.