Log in

No account? Create an account
Chaotic outage plan - Alierak
August 22nd, 2003
09:42 pm


Previous Entry Share Next Entry
Chaotic outage plan
Last night I went to the trouble to figure out approximately what else needs to be done to finish upgrading chaoticmux.org (aka mux), and how long it should probably take. Here's what I came up with:

2m update header_checks
2m temporarily disable user startups
5m set to boot as mux (make sure all needed etc.mux files)
2m shutdown
2m take disk + ram to machine room
3m loosen mux in rack
2m shutdown
15sMudNet: save config
15sMudNet: shutdown
1mrestart mysqld to counter fd leakage
2mdump mysql dbs
30sstop mysqld
5mChaoticMUX: @backup
1mChaoticMUX: @shutdown
5mdisassemble machine, blow out dust
5mswap disks, add ram
5mreassemble and power up
1mtake old disk to workroom
30s put old disk in G
3m power up G
1m go back to office
3mcopy recent ChaoticMUX files
10mrestore from most recent @backup
5mcopy recent MudNet files
2mstart mudnet
5mcopy recent Webster files
1mstart webot
15mcopy mailman lists/archives and make sure they work
20mcopy mysql dumps & reload them
30mcopy chaoticmux.org webpages and debug
5mcopy kareila cvs repository
15mcopy recent kareila files
10mcopy recent rmg files
5mcopy crontabs (kareila, rmg, chaotic)
30mcopy old logfiles (not overwriting)
1mre-enable user startups
5mexamine postfix queue on old disk
??do something vague about iss-grads autoreply which was majordomo-based

H is a temporary identity for the new mux boot disk; I've been working on it for a couple weeks and have it mostly set up. At this point it just needs some last-minute data copied over from the old mux disk. G is a handy machine with extra drive bays where I can leave the old mux disk connected for a while as a non-boot disk.

The schedule from @backup to the end of Startmush adds up to 42 minutes, whee.

Current Mood: geeky

(4 comments | Leave a comment)

Date:August 22nd, 2003 08:42 pm (UTC)
Well it would have to be 42, wouldn't it? Just seems appropriate. :)
[User Picture]
Date:August 22nd, 2003 11:53 pm (UTC)
Seems to me that the iss-grads autoreply could be handled with some combination of a Perl script, formail and/or procmail. I doubt anyone will notice if it's MIA for a few days.
[User Picture]
Date:August 23rd, 2003 06:35 am (UTC)
True, but if I'm going to install mail handling software just for it, I might as well keep majordomo. The first thing I'm going to try is Mailman's autoresponder, and hopefully I can turn off all the other features that would make it look like an actual mailing list. I'm referring to the thing that sends back "the list isn't here anymore" whenever people send mail to iss-grads@mit.edu.
[User Picture]
Date:August 23rd, 2003 11:09 am (UTC)
rcorder: requirement `postfix' in file `/etc/rc.d/mailman' has no providers.
Powered by LiveJournal.com