On Aug 7, 2012, at 9:32 AM, Fujii Masao wrote: > On Mon, Aug 6, 2012 at 3:29 AM, Ben Chobot <bench@xxxxxxxxxxxxxxx> wrote: >> >> Oh, I would have though that doing a clean shutdown of the old master (step 1) would have made sure that all the unstreamed wal records would be flushed to any connected slaves as part of the master shutting down. In retrospect, I don't remember reading that anywhere, so I must have made that up because I wanted it to be that way. Is it wishful thinking? > > When clean shutdown is requested, the master sends all WAL records to > the standby, > but it doesn't wait for the standby to receive them. So there is no > guarantee that all WAL > records have been flushed to the standby. Walreceiver process in the > standby might > detect the termination of replication connection and exit before > receiving all WAL records. > Unfortunately I've encountered that case some times. Oh, I see. Well, that's unfortunate. Thanks for the help though! It shouldn't be too hard to script up what you suggest. -- Sent via pgsql-general mailing list (pgsql-general@xxxxxxxxxxxxxx) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-general
