Hoi! I've successfully configured a postgresql highavibility system with pgcluster and heartbeat under debian.
It includes 2 nodes each one with a pg_ctl and a pg_replicate, on one of the nodes the loadbalancer is working. So far everything is working fine. I've noticed that if someone is logged in per psql on the loadbalancer and it is going to stop/start it fails. In this case the command "/usr/local/pgsql/bin/pglb -l -D /usr/local/pgsql/etc stop" stop only one process of defaultly workig three, the logged in user still can read/write, but no further login is possible. I think the loadbalancer isn't ready for a "fast shutdown" like pg_ctl or is there a (back)ground for this? pglb sends by a "stop" the SIGTERM, so it waits for connections to be closed. A "-m fast" parameter in pg_ctl cause a SIGINT... For first I'll expand the pglb command for my needings, isn't it an idea for further development on pgcluster? Nice greetings from the grey germany ;-) _______________________________________________ Pgcluster-general mailing list [email protected] http://pgfoundry.org/mailman/listinfo/pgcluster-general
