Recently, I made changes to my Postgres 8.3 database - added a couple new users and added these roles to the pg_hba.conf file. However, after running pg_ctl reload -D [DB directory], the changes didn't appear to take affect. After about an hour, the changes finally appeared.
Checking the logs, the server was sent to SIGHUP signal to reload the config files immediately after running pg_ctl. All log output afterwards was errors from my failed connection attempts (i.e. trying to connect with the new usernames I added).
I've been trying to find some documentation suggesting that there is a delay when the pg_ctl tool is used... Or, something to explain why there was such a long delay. Could it have been from either 1) the addition of the roles within the database or 2) the updating of the pg_hba.conf file??
restartinstead ofreload, and then the DB waited for all existing clients to disconnect before shutting down and starting anew. – Milen A. Radev Oct 29 '11 at 14:54