On Thu, Dec 11, 2008 at 9:59 AM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
> Greg Smith <gsmith@gregsmith.com> writes:
>> On Thu, 11 Dec 2008, Phillip Berry wrote:
>>> I'm not running PITR and checkpoint_segments is set to 100 as this is
>>> home to a very write intensive app.
>
>> That's weird then. It shouldn't ever keep around more than 201 WAL
>> segments. I've heard one report of a similarly mysterious excess of them,
>> from Robert Treat, but that was probably caused by a hardware failure.
>
> AFAIK the only non-PITR reason for WAL files to not get recycled is if
> checkpoints were failing. Do you still have the postmaster log from
> before the original crash, and if so is there anything in there about
> checkpoint failures?
Don't forget that the OP mentioned earlier that he had very long help
open connections with possible long help open transactions.
