Обсуждение: [HACKERS] Segmentation fault when creating a BRIN, 10beta1

Hi,

I can reproduce a segmentation fault when creating a BRIN concurrently
with set pages_per_range and autosummarize.

# Reproduce
CREATE TABLE brin_test AS
    SELECT
        series AS id,
        MD5(series::TEXT) AS VALUE,
        '2015-10-31 13:37:00.313370+01'::TIMESTAMP + (series::TEXT || '
Minute')::INTERVAL AS TIME
    FROM
        GENERATE_SERIES(0,10000) AS series;

CREATE INDEX CONCURRENTLY brin_test_time_abrin_8 ON brin_test USING brin
(TIME) WITH (pages_per_range=8, autosummarize=true);

2017-05-30 16:08:59.451 CEST [20689] LOG:  server process (PID 20881)
was terminated by signal 11: Segmentation fault
2017-05-30 16:08:59.451 CEST [20689] DETAIL:  Failed process was
running: CREATE INDEX CONCURRENTLY brin_test_time_abrin_8 ON brin_test
USING brin (TIME) WITH (pages_per_range=8, autosummarize=true);
2017-05-30 16:08:59.451 CEST [20689] LOG:  terminating any other active
server processes
2017-05-30 16:08:59.451 CEST [20806] WARNING:  terminating connection
because of crash of another server process
2017-05-30 16:08:59.451 CEST [20806] DETAIL:  The postmaster has
commanded this server process to roll back the current transaction and
exit, because another server process exited abnormally and possibly
corrupted shared memory.
2017-05-30 16:08:59.451 CEST [20806] HINT:  In a moment you should be
able to reconnect to the database and repeat your command.
2017-05-30 16:08:59.452 CEST [21204] postgres@postgres FATAL:  the
database system is in recovery mode

The debian packages where used, version "PostgreSQL 10beta1 on
x86_64-pc-linux-gnu, compiled by gcc (Debian 6.3.0-18) 6.3.0 20170516,
64-bit".
I also tested the latest git commit
d5cb3bab564e0927ffac7c8729eacf181a12dd40 with the same result.
A more detailed log (debug5) can be found attached.


Kind regards,
Alexander Sosna

-- 
Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers

Alexander Sosna <alexander.sosna@credativ.de> writes:
> I can reproduce a segmentation fault when creating a BRIN concurrently
> with set pages_per_range and autosummarize.

I wonder if this isn't the same issue reported in
https://www.postgresql.org/message-id/flat/20170524063323.29941.46339%40wrigleys.postgresql.org

Could you try the patch suggested there?
        regards, tom lane

Tom Lane wrote:
> Alexander Sosna <alexander.sosna@credativ.de> writes:
> > I can reproduce a segmentation fault when creating a BRIN concurrently
> > with set pages_per_range and autosummarize.
> 
> I wonder if this isn't the same issue reported in
> https://www.postgresql.org/message-id/flat/20170524063323.29941.46339%40wrigleys.postgresql.org
> 
> Could you try the patch suggested there?

With the patch, it crashes immediately for me.  Looking into it now.

-- 
Álvaro Herrera                https://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services

Alexander Sosna wrote:
> Hi,
> 
> I can reproduce a segmentation fault when creating a BRIN concurrently
> with set pages_per_range and autosummarize.

Pushed fix just now.  Please give it a try.  Thanks for testing and
reporting,

-- 
Álvaro Herrera                https://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services

On 2017-05-30 18:21:10 -0400, Alvaro Herrera wrote:
> Alexander Sosna wrote:
> > Hi,
> > 
> > I can reproduce a segmentation fault when creating a BRIN concurrently
> > with set pages_per_range and autosummarize.
> 
> Pushed fix just now.  Please give it a try.  Thanks for testing and
> reporting,

Shouldn't this have been uncovered by a regression test? In other words,
do I understand correctly that the new summarization stuff is largely
not tested by regression tests?

- Andres

Alvaro Herrera writes:
>> I can reproduce a segmentation fault when creating a BRIN concurrently
>> with set pages_per_range and autosummarize.
> 
> Pushed fix just now.  Please give it a try.  Thanks for testing and
> reporting,

Thank you for fixing, works for me.


Regards,
Alexander Sosna

Andres Freund wrote:
> On 2017-05-30 18:21:10 -0400, Alvaro Herrera wrote:
> > Alexander Sosna wrote:
> > > Hi,
> > > 
> > > I can reproduce a segmentation fault when creating a BRIN concurrently
> > > with set pages_per_range and autosummarize.
> > 
> > Pushed fix just now.  Please give it a try.  Thanks for testing and
> > reporting,
> 
> Shouldn't this have been uncovered by a regression test? In other words,
> do I understand correctly that the new summarization stuff is largely
> not tested by regression tests?

You understand correctly and I had the same concern myself upon finding
about the bug.  I cannot promise to create one right now but I can do so
after June 12th.

-- 
Álvaro Herrera                https://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services