On 2011-10-19 20:31, Dimitri Maziuk wrote: > On 10/19/2011 11:00 AM, Paul O'Rorke wrote: >> after thinking this through and doing a little reading it occurs to me that >> this will only help cache incomming mail until the primary MX mail server is >> back up. What I want is servers mirrored in realtime, with failover. I >> would have expected HA-Linux to be a good candidate for that. > > No, you can run the 2nd MX with the same priority and same config as the > 1st one. Its intended use is load balancing, but as a side effect, if > one MX goes down the other gets all the mail. > > The problem is shared resources: mbox'es in /var/spool/mail or MLM software. > > If you are delivering to ~/Maildir's and don't have mailing lists, it > should work. > > (If you're delivering to mbox'es in /var/spool/mail, don't. Consider > switching to maildirs as part of the upgrade.) > > If you insist on mbox'es you'll need drbd for mail spool and list > archives (if any) and probably also for config files to make your life > easier -- pretty much what Florian said. What he didn't say is it'll > take you a week of frustration to get a working configuration.
Eh, what's this? FUD dial stuck on "full blast"? I'll not touch upon the mbox vs Maildir point here as I think it's of limited relevance (and I've administered highly available, DRBD-backed MTA/MDA configurations delivering into Maildirs -- but I do fully concur with the idea that moving to Maildirs from mboxes is a good idea), but there are a couple other points that need correction. Config files: you _can_ put those on DRBD, but you don't have to. I tend to prefer csync2 or Puppet based configurations where the config does not live on DRBD. A week of frustration: perhaps, if doing it wrong. But if one does something wrong then they'll probably fail to do that in a week or whatever timeframe. For someone who does this on a daily basis, setting this up in a day is more realistic. If you're new to this but can _ask_ someone who does this on a daily basis, you'll get done in 2-3 days, perhaps, and learn something along the way. > The other thing he didn't say is that when you put those files/dirs on > drbd filesystem, you typically make symlinks from their original > location: e.g. /var/spool/mail->/drbd/mailspool. When drbd filesystem is > not mounted (on the passive node), all of them are broken symlinks. If one favors a symlink based approach, and has never heard of the symlink RA courtesy of our repeat contributor Dominik Klein, or chooses to not use it, then dangling symlinks will be a result, yes. Which gets me back to my original stance: if one does it wrong, then it's going to take $INTERVAL. If "one week of frustration" is someone's $INTERVAL, then OK, I'll take that person's word for it. Mind you, I'm not saying it's easy to get right. But that's just what high availability is like. Hope this is useful. Cheers, Florian -- Need help with High Availability? http://www.hastexo.com/now _______________________________________________ Linux-HA mailing list [email protected] http://lists.linux-ha.org/mailman/listinfo/linux-ha See also: http://linux-ha.org/ReportingProblems
