Обсуждение: BUG #17655: crash postgres by query

The following bug has been logged on the website:

Bug reference:      17655
Logged by:          Kos Dav
Email address:      kdavydenka@gmail.ru
PostgreSQL version: 10.19
Operating system:   Red Hat Enterprise Linux Server 7.8 (Maipo)
Description:

log from journalctl
Oct 20 14:52:21 server1 kernel: show_signal_msg: 537 callbacks suppressed
Oct 20 14:52:21 server1 kernel: postgres[9328]: segfault at 1e88ff3 ip
00007f55ad8414ff sp 00007fff0031e7f8 error 4 in
libc-2.17.so[7f55ad6eb000+1c3000]
Oct 20 14:52:21 server1 abrt-hook-ccpp[9329]: Process 35274(postgres) of
user 26 killed by SIGSEGV - dumping core


log from postgres
server process (PID 35274) was terminated by signal 11: Segmentation fault
 Failed process was running:
    select customer_id, birth_place from last_golden_record lgr where
birth_place like U&'%\000A%'

Hi,

On Thu, Oct 20, 2022 at 02:16:55PM +0000, PG Bug reporting form wrote:
> The following bug has been logged on the website:
>
> Bug reference:      17655
> Logged by:          Kos Dav
> Email address:      kdavydenka@gmail.ru
> PostgreSQL version: 10.19
> Operating system:   Red Hat Enterprise Linux Server 7.8 (Maipo)
> Description:
>
> log from journalctl
> Oct 20 14:52:21 server1 kernel: show_signal_msg: 537 callbacks suppressed
> Oct 20 14:52:21 server1 kernel: postgres[9328]: segfault at 1e88ff3 ip
> 00007f55ad8414ff sp 00007fff0031e7f8 error 4 in
> libc-2.17.so[7f55ad6eb000+1c3000]
> Oct 20 14:52:21 server1 abrt-hook-ccpp[9329]: Process 35274(postgres) of
> user 26 killed by SIGSEGV - dumping core
>
>
> log from postgres
> server process (PID 35274) was terminated by signal 11: Segmentation fault
>  Failed process was running:
>     select customer_id, birth_place from last_golden_record lgr where
> birth_place like U&'%\000A%'

Unfortunately this isn't enough information to understand the problem.  However
a common cause for a SIGSEGV on a query like that is some corruption on the
table.

Could you try to get a stack trace?  See

https://wiki.postgresql.org/wiki/Getting_a_stack_trace_of_a_running_PostgreSQL_backend_on_Linux/BSD#Getting_a_trace_from_a_randomly_crashing_backend.

PG Bug reporting form <noreply@postgresql.org> writes:
> Oct 20 14:52:21 server1 kernel: postgres[9328]: segfault at 1e88ff3 ip
> 00007f55ad8414ff sp 00007fff0031e7f8 error 4 in
> libc-2.17.so[7f55ad6eb000+1c3000]
> Oct 20 14:52:21 server1 abrt-hook-ccpp[9329]: Process 35274(postgres) of
> user 26 killed by SIGSEGV - dumping core

> log from postgres
> server process (PID 35274) was terminated by signal 11: Segmentation fault
>  Failed process was running:
>     select customer_id, birth_place from last_golden_record lgr where
> birth_place like U&'%\000A%'

With neither a stack trace nor a reproducible test case, there is
little anyone can do with this report.

There's some info about extracting stack traces from crashes at

https://wiki.postgresql.org/wiki/Generating_a_stack_trace_of_a_PostgreSQL_backend

but if you could make a self-contained test case that'd be
even better.

Also: it's possible that this is less about a software bug than
about corrupt data.  Of course, we'd prefer the server not crash
on corrupt data, but sometimes that happens.  A simple thing to do
that might help in such a case is to REINDEX whatever index(es)
the query is using.  But first see

https://wiki.postgresql.org/wiki/Corruption

Also: the 10.x series will reach EOL next month.  You should be
seriously working on updating to some later release branch.
Whether doing so will provide a fix for this problem is hard
to say, but it's not out of the question.

            regards, tom lane