21/01/2009 @06:57:10 ^08:45:59
hardware changes
unexpected freeze
So, not-so-funny thing. Last week I woke up to find baron had completely frozen.
- didn't respond to pings
- no monitor output (monitor stayed in poweroff mode when I plugged its cable in, so obviously no signal)
- cable modem lights were hardly flashing (when usually the server is constantly getting hit by overzealous web spiders, brute force ssh attacks, and rtorrent is merrily chugging away)
- most tellingly the hard disk light was apparently permanently lit (it was obvious something was wrong here because I had noticed the light had been on for over a minute and it was nowhere near the time that cron.daily is run, which is the usual reason for prolonged disk activity. This was not a nice thing to slowly realise as you are waking up)
baron is the name of my server, an old machine that sits in the corner of the room being a router, firewall and host for this website. It also has most of the hard disks in it which it shares via NFS. Therefore it needs to be reliable. If NFS goes down everything else freezes as well.
It came back up when I pressed the reset button and acted as if nothing had happened, so I have no idea what is going on. The only thing I can think is that maybe there is a bad bit of memory somewhere (I remember about six months ago waking up to find irssi had inexplicably segfaulted during the night, which irssi never does)
dissatifaction with NFS
That's one thing. Another is NFS itself. Ever since I upgraded baron to Debian Etch and a 2.6 kernel, NFS had become rather unreliable. While I appreciate that this was as likely to be a dubious network card as much as a software issue, the fact is that every so often it would freeze, apparently uninterruptibly, for a period of time that can last anything up to an hour. You can imagine how annoying this is!
(Indeed it's more likely to be hardware as stressing it by copying hundred-megabyte-plus files around would vastly increase the chances of freezing - although, why had this only begun to happen after a software upgrade? Who knows?)
NFS is also comparatively slow, what with all your disk accesses going over a network. Running autoconf-generated configure scripts was especially irritating as was anything else that created or deleted a lot of files.
power consumption
A third issue and possibly the most important - the rising cost of energy. A few months ago we got an electricity bill that was far too high. It was around that time I decided having a machine sat there switched on 24/7 but, to be quite frank, doing pretty much nothing at all, wasn't the best use of resources. The time spent outside of the idle loop was less than a fifth of a percent.
conclusion
Okay this is getting pretty long so let's finish it up. I spent last week installing and copying over all the configuration files for the various bits of software that were on baron but missing from vile, my desktop machine that actually gets used. A couple of days ago I switched everything off and moved the disks from one machine to the other.
I believe the merge has been successful. As vile has twice as much memory and a more powerful CPU I figured it could handle the increased load, and I am hoping having only one machine drawing power will make a difference (although that remains to be seen since the one machine has three hard disks in it now)
Nevertheless it is a sad day on which you take out of usage a system that's been there and running mostly smoothly for well over seven years. However, as I keep telling myself, given the reasons described above, it's probably for the best.
