mail.apache.org is dead; long live mail.apache.org!

Over the weekend, daedalus finally kicked the bucket. It was spontaneously rebooting on us, and the ASF didn't have the capability to reboot it or access its serial console. So, instead of spending a lot of time to fix it (which we don't have anyway and reboots like that typically indicate a hardware failure of some sort!), we decided to promote hermes into production and send daedalus to its final resting place.

We were pushing over 6 million messages a day through a dual PIII/800 with a qmail/ezmlm setup. In addition, before minotaur arrived last year, daedalus was also our primary webserver servicing about an equal number of hits per day. One of daedalus's hard drives had already failed in the last month or so. We've been prepping hermes, a new IBM x345 Series to replace it. But, as it goes, we weren't going to promote it until something like this happened. (Yes, we would like to have been able to do actual load testing and had been discussing it; alas...)

hermes seems to be responding to the load fine. We've added clamav scanning, and may add in SA later. I've got to take hermes down later tonight to disable atime. There will be some growing pains (i.e. we had to kick gump off this morning); but hermes is probably about 4x as powerful as daedalus was - this grants us the necessary headroom to add services like clamav and SA to the mix. This also completes the Collab ISP transition - the ASF is now fully self-hosting via UL and responsible for its own bandwidth.

Kudos to Brian and Noel. Assists to Tris and Fitz for sending daedalus on its way.

RIP daedalus (Nov. 20, 2000-May 9, 2004). She served us mighty well.

[/apache] permanent link