Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.
On Tue, Jul 19, 2005 at 02:49:48PM -0400, Diego Julian Remolina wrote: > Is anybody already running sendmail on top of drbd? Not exactly, no. But I do have a mail setup using drbd, and would be interested in comparing our reasoning. Here is mine: I'm happy to assume that a harddisk failure could cause loss of mail, and drbd could be used to manage that risk, but sendmail spools can't be shared (?) so that means two different spools. I'm not worrying about any other scenario just now, but I'm always happy to hear horror stories :) In the failover case you would want to be able to run the queue in the failed-over spool. The worst case might be a split-brain leading to a double delivery, and that's allowed anyway. If you do some sort of VM failover then thati could take care of the queue running. Other than that I see no obvious reason not to make the setup as symetrical as possible. Run sendmail simlutaneously on the two servers. If you need drbd'ed local delivery, then deliver by nfs to maildir from the secondary. Make a decision about syslogging. Anything else ? What parameters do you want to make different? People talk about cyrus in the context of mail clusters as well, but I don't profess to know anything about it. I'm not doing this at the moment, but I may heading this way. I'd be very interested in knowing what setup you end up with. > I am currently running sendmail with the mimedefang milter, checking viruses > with clamd and tagging spam with spamassassin and then deliverying mail to > Cyrus imapd. All of these services are started through heartbeat on the > haresources file. I also have spamassassin, and a question on the back-burner is how to share information across bayes DBs. Regards, Paddy -- Perl 6 will give you the big knob. -- Larry Wall