Re: strange parallel query behavior after OOM crashes

On 04/05/2017 04:15 PM, Robert Haas wrote:
> On Wed, Apr 5, 2017 at 10:09 AM, Kuntal Ghosh
> <kuntalghosh(dot)2007(at)gmail(dot)com> wrote:
>>> Did you intend to attach that patch to this email?
>>>
>> Actually, I'm confused how we should ensure (register_count >
>> terminate_count) invariant. I think there can be a system crash what
>> Tomas has suggested up in the thread.
>>
>> Assert(parallel_register_count - parallel_terminate_count <=
>> max_parallel_workers);
>> Backend 1 > SET max_parallel_worker = 8;
>> Backend 1 > Execute a long running parallel query q1 with number of
>> parallel worker spawned is say, 4.
>
> At this point, parallel_register_count should be equal to
> parallel_terminate_count. 4 workers were started, and 4 have
> terminated.
>

No, the workers were started, but are still running, so

register_count = 4
terminate_count = 0

>> Backend 2> SET max_parallel_worker = 3;
>> Backend 2 > Try to execute any parallel query q2 with number of
>> parallel worker spawned > 0.
>
> Now here parallel_register_count should get bumped up to 4+(# of
> workers now launched) at the beginning and then
> parallel_terminate_count at the end. No problem.
>
> What's going wrong, here?
>

Well, assuming the other workers are still running, you get

register_count = 4
terminate_count = 0

and the difference is greater than max_parallel_workers = 3. Which
violates the assert.

regards

--
Tomas Vondra http://www.2ndQuadrant.com
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services