We consider deploying a two server active-active setup with dsync replication for ~2500 users.
I was wondering how people handle sieve script synchronization in such setups.
We came up with a few options for syncing:
rsync/unison ~/sieve every x seconds
move sieve dirs to separatate partition and sync with gluster or drbd
patch roundcube managesieve sieve plugin to upload to both managesieve servers (only roundcube can access managesieve in our setup)
store sieve scripts in mysql and patch roundcube managesieve plugin to directly access the db instead of using managesieve (we tried this, but apparently sieve scripts read from db via dovecot dict can't have more than one line? show stopper due to roundcube sieve plugin needing comments to parse the scripts)
wait until Timo/Stephan add additional magic to dsync for sieve scripts as discussed in http://www.dovecot.org/list/dovecot/2011-October/061699.html ;-)
We are tending towards 1) as this seems simplest and most robust solution but before we re-invent the wheel we'd like to "hear" your thoughts...
Thanks Oli
-- This message does not contain horse meat