Re: [HACKERS] parallel.c oblivion of worker-startup failures

On Tue, Dec 12, 2017 at 9:00 AM, Amit Kapila <amit(dot)kapila16(at)gmail(dot)com> wrote:
> On Mon, Dec 11, 2017 at 11:27 PM, Robert Haas <robertmhaas(at)gmail(dot)com> wrote:
>> On Sun, Dec 10, 2017 at 11:07 PM, Amit Kapila <amit(dot)kapila16(at)gmail(dot)com> wrote:
>>> Okay, see the attached and let me know if that suffices the need?
>>
>> + * Check for unexpected worker death. This will ensure that if
>> + * the postmaster failed to start the worker, then we don't wait
>> + * for it indefinitely. For workers that are known to be
>> + * launched, we can rely on their error queue being freed once
>> + * they exit.
>>
>> Hmm. Is this really true? What if the worker starts up but then
>> crashes before attaching to the error queue?
>>
>
> If the worker errored out before attaching to the error queue, then we
> can't rely on error queue being freed. However, in that case, the
> worker status will be BGWH_STOPPED. I have adjusted the comment
> accordingly.
>

In particular, if the worker crashed (say somebody forcefully killed
the worker), then I think it will lead to a restart of the server. So
not sure, adding anything about that in the comment will make much
sense.

--
With Regards,
Amit Kapila.
EnterpriseDB: http://www.enterprisedb.com