``Immediate'' mode will abort all server processes without a clean shutdown. This will lead to a recovery run on restart.
Regards, Guido Barosio
On 3/14/06, Scott Marlowe < smarlowe@g2switchworks.com> wrote: > On Tue, 2006-03-14 at 14:05, Guido Barosio wrote: > > Hi Cory, > > > > Actually it seems that the database wasn't shut down properly, and > > while it was doing the recovery, it refused connections. Expected and > > what needs to happen, indeed. > > > > That's why if you are a client about to connect, you'll recieve the " > > FATAL ...database is starting up " message. > > > > Ideally, you (the dba) should let the server end this task (the > > recovery). AFAIK, the recovery started as a consecuence of dirty data > > somewere, probably a big transaction not comitted when the server went > > off. > > > > Are you aware of a shutdown prior to this messages? > > > > Nothing to do with your current issue, but you will receive comments > > related with your pgsql version if you are not using a pgsql 8 :) > > As mentioned, this is normal operation. > > Note that the command: > > pg_ctl -m immediate stop > > will cause a recovery. It's pretty much a functional equivalent of > senting kill -9 to all backends and the postmaster at once. Brutal but > effective. >
-- /"\ ASCII Ribbon Campaign . \ / - NO HTML/RTF in e-mail . X - NO Word docs in e-mail . / \ -----------------------------------------------------------------
Ok, thanks a lot guys.
It seems that I should have brought the database down to "local access" mode via pg_hba.conf and allow all transactions to fully recover before starting up replication or allowing other applications to access the db.