On Thu, Nov 24, 2011 at 02:35:55PM +0100, Patrick Westenberg wrote:
- What happens if the MTA accepted a message but the backend-server fails and is unable to store it?
Your MTA will be notified about the failure trough SMTP/LMTP error codes, and retry if it was a temp-failure.
- What about load-balancing and/or failover? Maybe all users on backend-server1 are fetching their mails causing high load while backend-server2 is idle. Is it possible to specify more than one backend-server IPs as I can do for the database hosts for userdb queries?
We use a clusterfs (so it's not very important which backend server the user is directed to), and have configured the dovecot director with:
director_mail_servers = 192.168.42.7 192.168.42.8 192.168.42.9 192.168.42.10 192.168.42.11 192.168.42.28 192.168.42.29
director_servers = 192.168.42.15 192.168.42.17
passdb {
args = proxy=y nopassword=y
driver = static
}which gives us loadbalancing of new logins, and also failover should a backend server fail.
If you want to work with at local index and maildirs on NFS, you'll need to configure something smarter for selecting backend server. Maybe use proxy extrafield to specify a host for the user, and point this at a failover IP that will hang on the preferred server by default, but fail over to another working server if the preferred server fails..
http://wiki2.dovecot.org/PasswordDatabase/ExtraField-jf