Re: postgres_fdw : altering foreign table not invalidating prepare statement execution plan. - Mailing list pgsql-hackers

From Amit Langote
Subject Re: postgres_fdw : altering foreign table not invalidating prepare statement execution plan.
Date
Msg-id CA+HiwqFMLJHfi8AeH7gZvy8ZayqaptxNX+duQi6uf2+qwjS2oA@mail.gmail.com
Whole thread Raw
In response to Re: postgres_fdw : altering foreign table not invalidating prepare statement execution plan.  (Tom Lane <tgl@sss.pgh.pa.us>)
Responses Re: postgres_fdw : altering foreign table not invalidating prepare statement execution plan.
List pgsql-hackers
On Mon, Apr 4, 2016 at 11:24 PM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
> Amit Langote <Langote_Amit_f8@lab.ntt.co.jp> writes:
>> On 2016/04/04 15:17, Rajkumar Raghuwanshi wrote:
>>> * .Observation: *Prepare statement execution plan is not getting changed
>>> even after altering foreign table to point to new schema.
>
>> I wonder if performing relcache invalidation upon ATExecGenericOptions()
>> is the correct solution for this problem.  The attached patch fixes the
>> issue for me.
>
> A forced relcache inval will certainly fix it, but I'm not sure if that's
> the best (or only) place to put it.
>
> A related issue, now that I've seen this example, is that altering
> FDW-level or server-level options won't cause a replan either.  I'm
> not sure there's any very good fix for that.  Surely we don't want
> to try to identify all tables belonging to the FDW or server and
> issue relcache invals on all of them.

Hm, some kind of PlanInvalItem-based solution could work maybe? Or
some solution on lines of recent PlanCacheUserMappingCallback() added
by fbe5a3fb, wherein I see this comment which sounds like a similar
solution could be built for servers and FDWs:

+       /*
+        * If the plan has pushed down foreign joins, those join may become
+        * unsafe to push down because of user mapping changes. Invalidate only
+        * the generic plan, since changes to user mapping do not invalidate the
+        * parse tree.
+        */

Missing something am I?

Thanks,
Amit



pgsql-hackers by date:

Previous
From: Julien Rouhaud
Date:
Subject: Re: Choosing parallel_degree
Next
From: Artur Zakirov
Date:
Subject: Patch: fix typo, duplicated word in indexam.sgml