Re: A new function to wait for the backend exit after termination

On Fri, Dec 4, 2020 at 2:02 PM Hou, Zhijie <houzj(dot)fnst(at)cn(dot)fujitsu(dot)com> wrote:
>
> Hi,
>
> When test pg_terminate_backend_and_wait with parallel query.
> I noticed that the function is not defined as parallel safe.
>
> I am not very familiar with the standard about whether a function should be parallel safe.
> But I found the following function are all defined as parallel safe:
>
> pg_promote
> pg_terminate_backend(integer)
> pg_sleep*
>
> Is there a reason why pg_terminate_backend_and_wait are not parallel safe ?
> (I'm sorry if I miss something in previous mails.)
>

I'm not quite sure of a use case where existing pg_terminate_backend()
or for that matter the new pg_terminate_backend_and_wait() and
pg_wait_backend() will ever get used from parallel workers. Having
said that, I marked the new functions as parallel safe to keep it the
way it is with existing pg_terminate_backend().

postgres=# select proparallel, proname, prosrc from pg_proc where
proname IN ('pg_wait_backend', 'pg_terminate_backend');
proparallel | proname | prosrc
-------------+----------------------+-------------------------------
s | pg_terminate_backend | pg_terminate_backend
s | pg_wait_backend | pg_wait_backend
s | pg_terminate_backend | pg_terminate_backend_and_wait
(3 rows)

Attaching v6 patch. Please have a look.

With Regards,
Bharath Rupireddy.
EnterpriseDB: http://www.enterprisedb.com