ugh.book

240 System Administration

Where Did I Go Wrong?

Date: Thu, 20 Dec 90 18:45 CST

From: Chris Garrigues <[email protected]>

To: UNIX-HATERS

Subject: Support of Unix machines

I was thinking the other day about how my life has changed since

Lisp Machines were declared undesirable around here.

Until two years ago, I was single-handedly supporting about 30

LispMs. I was doing both hardware and software support. I had time

to hack for myself. I always got the daily paper read before I left in

the afternoon, and often before lunch. I took long lunches and rarely

stayed much after 5pm. I never stayed after 6pm. During that year

and a half, I worked one (1) weekend. When I arrived, I thought the

environment was a mess, so I put in that single weekend to fix the

namespace (which lost things mysteriously) and moved things

around. I reported bugs to Symbolics and when I wasn’t ignored, the

fixes eventually got merged into the system.

Then things changed. Now I’m one of four people supporting about

50 Suns. We get hardware support from Sun, so we’re only doing

software. I also take care of our few remaining LispMs and our Cisco

gateways, but they don’t require much care. We have an Auspex, but

that’s just a Sun which was designed to be a server. I work late all the

time. I work lots of weekends. I even sacrificed my entire Thanksgiv-

ing weekend. Two years later, we’re still cleaning up the mess in the

environment and it’s full of things that we don’t understand at all.

There are multiple copies of identical data which we’ve been unable

to merge (mostly lists of the hosts at our site). Buying the Auspex

brought us from multiple single points of failure to one huge single

point of failure. It’s better, but it seems that in my past, people fre-

quently didn’t know that a server was down until it came back up.

Even with this, when the mail server is down, “pwd” still fails and

nobody, including root, can log in. Running multiple version of any

software from the OS down is awkward at best, impossible at worst.

New OS versions cause things to break due to shared libraries. I

report bugs to Sun and when I’m not ignored, I’m told that that’s the

way it’s supposed to work.

Where did I go wrong?