On 16 March 2013 19:10, Tom Lane <tgl@xxxxxxxxxxxxx> wrote:
Alban Hertroys <haramrae@xxxxxxxxx> writes:More generally: since we're not hearing this type of complaint from
> If there's actually something wrong with the database; it looks a bit like your tables and your indexes get out of sync somehow, which normally wouldn't be possible. I'm mostly guessing, but perhaps one of the below has something to do with it:
> Maybe you turned fsync off?
> What type of index is that? A standard btree or one of the newer types?
> Are those tables and indexes perhaps on some kind of virtual storage or on a file-system that might be rolling back file-system transactions? It this database perhaps a replicated node?
other people, there must be something pretty unusual about your
installation. You've provided no information that would suggest what,
though. Aside from Alban's questions, some other things come to mind:
* is that a plain text column, or some other data type?
* what collation/ctype is your database using?
* what nondefault parameter settings are you using?
* where did you get the Postgres executables from? Some distro (whose)?
If they're self-built, what compiler and configuration settings did
you use?
* what platform is this? I would not rule out kernel bugs or flaky
hardware.
regards, tom lane
* it is varchar columns, 256 and 32 symbols length
* encoding, collation and ctype: UTF8, en_US.utf8, en_US.utf8
* autovacuum, fsync off, full_page_writes = on, wal_writer_delay = 500ms, commit_delay = 100, commit_siblings = 10, checkpoint_timeout = 20min, checkpoint_completion_target = 0.7
* postgres 9.2.3 installed via yum repository for version 9.2
* 64 bit Centos 6, installed and updated from yum repository
