Re: BUG #14189: Corrupted WAL through replication protocol

root(at)simply(dot)name wrote:

> root(at)xdb2018f /tmp # /usr/pgsql-9.5/bin/pg_xlogdump /tmp/normal/00000001000024DB00000072 /tmp/normal/00000001000024DB00000072 | fgrep -C1 24DB/72095CC8

So the problem record is:

> rmgr: Gin len (rec/tot): 2/ 412, tx: 193820712, lsn: 24DB/72095E50, prev 24DB/72095CC8, desc: INSERT isdata: F isleaf: T isdelete: T, blkref #0: rel 1663/16400/18030 blk 368350

for which you got this message:

> root(at)xdb2018f /tmp # /usr/pgsql-9.5/bin/pg_xlogdump /tmp/problem/00000001000024DB00000072 /tmp/problem/00000001000024DB00000072 | fgrep -C1 24DB/72095CC8
> pg_xlogdump: FATAL: error in WAL record at 24DB/72095E50: unexpected pageaddr 24D9/23096000 in log segment 00000000000024DB00000072, offset 614400

This is very odd. 24D9/23096000 looks like a valid address, perhaps
something that was left uninitialized from a previous write rather than
random corruption.

I don't suppose you can reproduce this at will, can you?

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