checkpoint_segments has been set at 12 for a while and was never set higher than that. (before that it was set to the PG default - 3 I think) Before the server crashed I was running an update that updates a boolean flag on two large tables (10 million rows each) for transactions older than today (roughly 80% of the rows) The transaction ran for a long time and I assume is what caused the pg_xlog to fill up. On Fri, 22 Dec 2006 17:36:39 +0000, "Simon Riggs" <simon@xxxxxxxxxxxxxxx> said: > On Fri, 2006-12-22 at 12:30 -0500, Jeremy Haile wrote: > > The archive_status directory is empty. I've never seen any files in > > there and I've never set archive_command. > > > > Well, the problem has since resolved, but here is what is in the > > directory now. Previously there were hundreds of files, but these > > disappeared after Postgres performed the automatic recovery. > > What were you doing before the server crashed? > > Did you previously have checkpoint_segments set higher? When/how was it > reduced? > > -- > Simon Riggs > EnterpriseDB http://www.enterprisedb.com > >
