I'm running some tests on a cluster with 4kB blocks, and it seems there's a bug in xlogreader.c, causing FATAL errors for example when running pg_xlogdump:

pg_xlogdump: FATAL: error in WAL record at 48/63970258: BKPIMAGE_HAS_HOLE not set, but hole offset 0 length 4096 at 48/63970258

This particular failure comes from "pg_xlogdump --stats", but my guess is this means recovery is broken with 4kB blocks (and possibly with some other non-standard block sizes).

The standard 8kB blocks seem to be unaffected, as I've done the same test on 8kB blocks many times and it never triggered this error.

FWIW the tests were done on bfcd07b4, so fairly recent code.


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

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

Reply via email to