winscape 2341

Amen.

This is to bring you a view from this side of things.

I have had a harrowing week of hardware problems. It seems there is no end to it, 4 disks have turned bad the last week, as well as a smoked motherboard; and one of the disks toasted the power as well. This followed a set of smoked linecards and a dead router.

This was the second disk in a month that decided to make a short circuit of itself when it died. I must have breathed some bad hardware karma lately. Probably need to sacrifice a small animal or junior salesperson to the power supply god or something.

Seven boxes were involved, both at home and at work.

The machines that didn't need installs run raids. It was a simple issue of obtaining new disk (system still running, but sulking) shutting down, opening cabinet , replacing disk, re-configure raid, and off it went. No serious downtime. I am getting convinced raids are to stay in my surroundings.

Reconstructing and installing systems has become fairly routine around here; but I'm considering making my own version of the installer so it can run 100% unattended. The software I install from source (because I track development, mostly) is safely stored on a raid, with the source tarballs safely on another raid on another machine. It is a small script job of unpack-configure-make-make install to rebuild this.

Yes, Amelia, I use network exports extensively to do this.

By now I have had it with bad disks. I am putting mirrored drives and-or software raids in all machines that can fit them, and I am setting up automated PXE boot (netboot, for you oldtimers) for the various boxes. Disk quality has gone down the drain the last two years after decades of steady improvement.

PXE boot is easy to make really fancy. It can boot, load and install *n*x all unattended with little tinkering from the box.

I simply cannot see how this week would have been survivable without tools the Open Source environment take for granted, but that is not used in other, "commercial" shops.

-- mrr