On Wed, Jun 16, 2010 at 05:19:25PM +0300, Elver Loho wrote: > Replication looks to be the solution. Surprisingly I hadn't discovered > it on my own. Many thanks to everyone for pointing it out! > > However, while setting it up, I noticed a really weird thing. When > authentication fails on the master-initiated master->slave connection, > master will block connections from all clients! That's a seriously > weird design choice. Does the same thing happen when the slave goes > down? For example, if I reboot the slave, is the master mail server > down for everyone while the slave reboots? I have not dared to > experiment as when authentication failed and the Cyrus master server > stopped letting clients connect, everyone in the office started > shouting within seconds. What version of cyrus are you using? I recommend adding the -o flag to sync_client in cyrus.conf (try ONCE then drop out) - otherwise it keeps trying for roughly 1000 seconds before giving up. And yes, that's a crazy design decision, which is why I added the -o flag! You also need to monitor it a bit to make sure there's still a sync_client running - because bail outs can cause it to go away. We run a cron job that checks on the sync clients every 10 minutes and reports if any are missing - runs sync_client -r -f on the files left behind by any that have previous died, and generally keeps things tidy :) Bron. ---- Cyrus Home Page: http://cyrusimap.web.cmu.edu/ Cyrus Wiki/FAQ: http://cyrusimap.web.cmu.edu/twiki List Archives/Info: http://asg.web.cmu.edu/cyrus/mailing-list.html
