Re: Segfault 11 on PG10 with max_parallel_workers_per_gather>3

On Wed, Oct 25, 2017 at 5:46 PM, Stefan Tzeggai
<tzeggai(at)empirica-systeme(dot)de> wrote:

> "Finalize Aggregate (cost=186411.37..186411.38 rows=1 width=8)"
> " -> Gather (cost=186410.95..186411.36 rows=4 width=8)"
> " Workers Planned: 4"
> " -> Partial Aggregate (cost=185410.95..185410.96 rows=1 width=8)"
> " -> Parallel Bitmap Heap Scan on
> as_20171025_20170930_ut78777 rt (cost=12058.69..185353.14 rows=23121
> width=0)"
> " Recheck Cond: (((oadr_gkz = ANY
> ('{2000000,5111000,5314000,5315000,5334002,5515000,6411000,6412000,7315000,8111000,8221000,9162000,9184119,11000000,14612000}'::integer[]))
> AND (objekttyp_grob = 1)) OR (oadr_gkz = 2000000))"
> " Filter: ((oart_zwangsversteigerung_janein IS NULL)
> AND (((oadr_gkz = ANY
> ('{2000000,5111000,5314000,5315000,5334002,5515000,6411000,6412000,7315000,8111000,8221000,9162000,9184119,11000000,14612000}'::integer[]))
> AND (objekttyp_grob = 1 (...)"
> " -> BitmapOr (cost=12058.69..12058.69 rows=94046
> width=0)"
> " -> BitmapAnd (cost=11726.20..11726.20
> rows=76321 width=0)"
> " -> Bitmap Index Scan on

By looking at the plan it seems like the issue what got fixed in below commit.

Author: Robert Haas <rhaas(at)postgresql(dot)org> 2017-10-14 00:23:28
Committer: Robert Haas <rhaas(at)postgresql(dot)org> 2017-10-14 00:35:14
Parent: d48bf6a94d295c3779c6af4df118d95a6606192f (Fix AggGetAggref()
so it won't lie to aggregate final functions.)
Child: cb591fcbfbba1df6fda1839ece53665e85e491e3 (Restore nodeAgg.c's
ability to check for improperly-nested aggregates.)
Branch: remotes/origin/REL_10_STABLE
Follows: REL_10_0
Precedes: REL_10_1

Fix possible crash with Parallel Bitmap Heap Scan.

If a Parallel Bitmap Heap scan's chain of leftmost descendents
includes a BitmapOr whose first child is a BitmapAnd, the prior coding
would mistakenly create a non-shared TIDBitmap and then try to perform
shared iteration.

Report by Tomas Vondra. Patch by Dilip Kumar.

--
Regards,
Dilip Kumar
EnterpriseDB: http://www.enterprisedb.com