Horton

Re: buildfarm: could not read block 3 in file "base/16384/2662": read only 0 of 8192 bytes

Andres Freund <andres@anarazel.de>

View thread View on postgr.es
From: Andres Freund <andres@anarazel.de>
To: Tomas Vondra <tomas.vondra@2ndquadrant.com>
Cc: Tom Lane <tgl@sss.pgh.pa.us>, Peter Geoghegan <pg@bowt.ie>, PostgreSQL Hackers <pgsql-hackers@lists.postgresql.org>
Date: 2018-08-11T14:08:23Z
Lists: pgsql-hackers 
Hi,

On 2018-08-11 15:40:19 +0200, Tomas Vondra wrote:
> For the record, I can actually reproduce this on 9.6 (haven't tried
> older releases, but I suspect it's there too). Instead of using the
> failing subscription, I've used another pgbench script doing this:

>   SET statement_timeout = 5;
>   COPY t TO '/dev/null';
> 
> and doing something like:
> 
>    pgbench -n -c 20 -T 300 -f copy.sql test

Just to confirm: That's with the vacuum full and insert running
concurrently? And then just restarting the above copy.sql (as pgbench
errors out after the timeouts) until you get the error?

I'm a bit confused what the copy + timeout is doing here? It shouldn't
trigger any invalidations itself, and the backtrace appears to be from
the insert.sql you posted earlier? Unclear why a copy to /dev/null
should trigger anything like this?

Greetings,

Andres Freund

Commits

  1. Limit depth of forced recursion for CLOBBER_CACHE_RECURSIVELY.

  2. Fix longstanding recursion hazard in sinval message processing.

  3. Fix bugs in vacuum of shared rels, by keeping their relcache entries current.