Hi folks, Re: Solution on mail server down time. A mail server may be out of work under 2 situations, 1) The server is broken down 2) ISP connection is out of service For 1) I can build a replicate server or just a box for mail queueing avoiding bouncing of mails For 2) It is out of my control I'm venturing to find a solution whether I can make use of the parking service on Registrar's server. Let the mails queneing there to be download back to the mail server later. OR are there any other suggestions? TIA Edit: P.S. the mail server only serves <15 users w/o heavy traffic B.R. satimis
2nd MX done by a different ISP is surely you're solution. The domains should also have dns done by difference ISPs so that you have a completely resilient fallback solution. I run a small ISP myself and that was pretty much my solution so I have 2 servers in 2 different datacentres works a treat - costly but works great. Scenario: Server1 goes down which runs MX10 and NS1 then server2 collects all the mail which runs MX20 and NS2. If I really need to get at the messages before server1 is back up and running I can push them to a popbox on the 2nd server and use sendmail to push them back to the first server once delivery has occurred. Otherwise they just sit in the postfix queue with a transport told to push the mail back to server1. I can run a manual "postfix flush" when the problem with server1 is resolved. This works really well as I would do a lot of MX backup for business that use MS Exchange on DSL connections. The only real bad thing is the extra hassle if I really need to get an email trapped on server2. Moreover: If point 1 is all you want to solve - you should try the heartbeat/drbd solution outlined in the HA NFS clustering howto. That way if server1 goes down server2 takes over as if nothing happened! Quite excellent! I've seen this in operation and it works great. Pretty much follow the guide except you add a few bits and bobs that allows Postfix to operate as well. Finally (bit overkill): If you want to be really smart you would do 2 servers in different datacentres run by the same ISP running BGP. This would mean the ISP is highly unlikely to go down as it's running BGP for multihoming and you could run DRBD/Heartbeat as they would have the same IP space in datacentre. Now this option is mega expensive I would imagine as you would want at least 100Mbit speed between the 2 servers to keep everything updated all the time even with just 15 users.
