The following bug has been logged on the website:
Bug reference: 10329
Logged by: Olivier Macchioni
Email address: olivier.macchioni@wingo.ch
PostgreSQL version: 9.1.13
Operating system: Debian 7.4 and 6.0.9
Description:
Hello all,
- Streaming Replication setup, 2 DB servers
- 9.1.13 on both sides (from apt.postgresql.org)
- Debian 6.0.9 on master server, 7.4 on standby server
When doing the following query on the standby server:
select * from ir_translation where src = 'a';
We have the following error in the logfiles:
ERROR: could not read block 0 in file "base/56100265/57047884": read only 0
of 8192 bytes
When checking the file usage, the following is reported:
postgres@db-stdby:~/9.1/main/base/56100265$
/usr/lib/postgresql/9.1/bin/oid2name -d xx -f 57047884
>From database "xx":
Filenode Table Name
---------------------------------------
57047884 ir_translation_src_hash_idx
ls -la shows that the file has a size of 0.
What's interesting as well is that this file has a last modification date of
April, 29th - when the DB has been created on the Master server much more
recently, by doing a "drop xx; rename xx_new xx;"
root@db-stdby:/var/lib/postgresql/9.1/main/base/56100265# stat 57047884
File: `57047884'
Size: 0 Blocks: 0 IO Block: 4096 regular empty
file
Device: 801h/2049d Inode: 1051439 Links: 1
Access: (0600/-rw-------) Uid: ( 106/postgres) Gid: ( 111/postgres)
Access: 2014-05-15 11:15:50.218038536 +0200
Modify: 2014-04-29 16:20:36.645605377 +0200
Change: 2014-04-29 16:20:36.645605377 +0200
Birth: -
root@db:/var/lib/postgresql/9.1/main/base/56100265# stat 57047884
File: `57047884'
Size: 2129920 Blocks: 4168 IO Block: 4096 regular file
Device: 801h/2049d Inode: 315480 Links: 1
Access: (0600/-rw-------) Uid: ( 104/postgres) Gid: ( 107/postgres)
Access: 2014-05-15 11:24:29.000000000 +0200
Modify: 2014-05-03 23:08:11.000000000 +0200
Change: 2014-05-03 23:08:11.000000000 +0200
The master server doesn't show this behavior.
The symptoms are similar to some previous (but much older) posts on this
list, for instance
http://www.postgresql.org/message-id/E1TEjPD-0002Yq-7L@wrigleys.postgresql.org
So I think this bug is not fixed yet...