Power

The whole Crum building just faced a 15 minute power outtage. No fun! One of our UPSes (the Belkin one) couldn't take the strain, and died after 5-6 minutes. It died cleanly, though - and our servers went down gracefully. The other UPS made the full 15, but I shut the servers attached to it down as a precaution anyway. Amazingly enough, everything that should have restarted after the outage did, with two exceptions: authdaemond, part of SqWebmail really doesn't like the rc.d mechanism in FreeBSD, and Apache on Charizard. The latter is pretty odd, but it turns out that Apache is helpless without a good DNS server - and our DNS server hadn't restarted. "apachtctl start" fixed that one just fine.

So all is well on the power front, at least for now. Hopefully, brown outs won't be too common this summer; last year, they hit Re/Max a few times, as the City of Columbia power completely failed to keep up with demand for air conditioning in 100F weather.

Ironically enough, one of our problem children is called "ipowerweb" - apparently, it is bad power day! They run Exim as their mailer, and it is setup really stupidly: when TLS is denied (for whatever reason), it refuses to deliver at all, rather than failing gracefully. This is permissable but stupid (it breaches ALL the mail RFCs that mention TLS) for an individual trying to make a statement about the need to encrypt email at the envelope level. This is really, really stupid for a hosting company given that about 80% of the world doesn't speak TLS - so mail to 80% of clients will not deliver correctly.

To add insult to injury, not only do their servers require TLS - but they require TLS with very specific protocol requirements (failing to fall back when their protocol of choice is refused), and simply timing out the connection if any sort of DNS mismatch appears. Stupid!
Mood: accomplished
Music: None