Re: logical decoding / rewrite map vs. maxAllocatedDescs

On 08/11/2018 09:46 AM, Andres Freund wrote:
> On 2018-08-11 00:34:15 -0700, Andres Freund wrote:
>> I've run this numerous times now, and I haven't triggered it yet :/
> 
> Heh, I hit it literally seconds after hitting send:
> 
> 2018-08-11 00:35:52.804 PDT [2196][9/14864] LOG:  automatic analyze of table "postgres.pg_catalog.pg_depend" system usage: CPU: user: 0.02 s, system: 0.00 s, elapsed: 0.15 s
> 2018-08-11 00:36:00.391 PDT [2196][9/14866] LOG:  automatic analyze of table "postgres.public.t" system usage: CPU: user: 0.25 s, system: 0.02 s, elapsed: 7.56 s
> 2018-08-11 00:36:00.399 PDT [2171][4/200412] PANIC:  could not read block 3 in file "base/12397/167946": read only 0 of 8192 bytes
> 2018-08-11 00:36:00.399 PDT [2171][4/200412] STATEMENT:  insert into t (b,c) values (1,1);
> 2018-08-11 00:36:00.410 PDT [2196][9/14868] LOG:  skipping vacuum of "pg_class" --- lock not available
> 2018-08-11 00:36:00.448 PDT [389][] LOG:  server process (PID 2171) was terminated by signal 6: Aborted
> 2018-08-11 00:36:00.448 PDT [389][] DETAIL:  Failed process was running: insert into t (b,c) values (1,1);
> 2018-08-11 00:36:00.448 PDT [389][] LOG:  terminating any other active server processes
> 
> Below you can find the bt full showing a bunch of nested invalidations. Looking.
> 

Hmmm, it's difficult to compare "bt full" output, but my backtraces look
somewhat different (and all the backtraces I'm seeing are 100% exactly
the same). Attached for comparison.

regards

-- 
Tomas Vondra                  http://www.2ndQuadrant.com
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services