Date: Fri, 29 Apr 2005 08:29:54 -0400 From: Chris Shenton <chris@shenton.org> To: Oliver Brandmueller <ob@e-Gitt.NET> Cc: freebsd-isp@freebsd.org Subject: Re: load balancing - email server Message-ID: <86y8b16a0t.fsf@PECTOPAH.shenton.org> In-Reply-To: <20050428080402.GP95908@e-Gitt.NET> (Oliver Brandmueller's message of "Thu, 28 Apr 2005 10:04:02 %2B0200") References: <005201c54b92$0cf63e60$0701a8c0@CIRIUM> <20050428072517.GO95908@e-Gitt.NET> <200504280944.12838.etienne@unix.za.org> <20050428080402.GP95908@e-Gitt.NET>
next in thread | previous in thread | raw e-mail | index | archive | help
Oliver Brandmueller <ob@e-Gitt.NET> writes: > Never looked at it; is it able to do weighting and failover? How does it > detect, if a service is down on one of the machines, so that you don't > have every third connection failing? Detecting and avoiding failure is the most critical thing. I've found true balancing isn't that important. For a client, I set up qmail-ldap on a handful of machines; each also runs courier-imap. They're behind an F5 load balancer, but you could use the nifty "pen" load balancer (/usr/ports/net/pen). They all use NFS-attached NetApp storage which is pretty high availability; the Maildir storage is the key. To get fault tolerance, you'd need to have two "pen" boxes (or pen running on two mail boxes) and configure pen to fail over if one dies; I haven't tried that because I have the F5. (The one thing that's not bullet proof in this set up is outgoing mail queues, which -- in qmail -- must be on the local box. If that box catches fire, undelivered mail is lost. That hasn't been much of an issue for us, however). The system's been up for the past year, with zero downtime, even when we've taken down individual boxes for OS upgrades.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?86y8b16a0t.fsf>