2009/3/28 Jose Perez <jvoorhe...@gmail.com>: > So how can I make sure that a queue file (stored in one node of a > cluster) it will be managed by another postfix instance (running in a > different node of a cluster)? > > I thought that using DRBD (network mirrored data) over the postfix > queue directory would be the solution but apparently I'm wrong...
That'd be dependent on how you choose to use DRBD. In active-active configuration you need a cluster-aware filesystem on top of the block device. I don't know how Postfix will play in this situation, I suspect it would not be nice (since it'd be the same as two postfix instances running on the same system, using the same queue/data/config directories). If you use an active-passive setup (the one I'm somewhat familiar with) you'd be okay as long as you turn postfix into a shared HA resource. This would have a warm failover - stop postfix on the downed node, transfer control of the block device to the passive node, then start up its instance of postfix. I *think* that would work fine, but I'm not qualified to say for certain. Wietse wrote: > I don't know if DRBD propagates data before fsync() returns, or > whether its updates happen later. If the updates happen later, the > backup may never learn that mail was queued because the update > still sits in the sender's DRBD queue. This is dependent on your choice of replication protocol. http://www.drbd.org/users-guide-emb/s-replication-protocols.html The sanest option is protocol "C". Those docs only refer to "local write operations", but I assume that would include fsync(), in which case there's no problem.
