[17816] in Athena Bugs
SGIs unhappy when zip disk forcibly removed
daemon@ATHENA.MIT.EDU (Emil Sit)
Mon May 1 22:11:55 2000
Message-Id: <200005020211.WAA28739@department-of-alchemy.mit.edu>
To: bugs@MIT.EDU
cc: dtpaul@MIT.EDU
Date: Mon, 01 May 2000 22:11:44 -0400
From: Emil Sit <sit@MIT.EDU>
System name: w20-575-62.mit.edu
Type and version: SGI....
What were you trying to do?
Debug a user problem sending mail
What's wrong:
The machine wouldn't run it's mailq. On examination of the syslog, the load
was too high, verified by athinfo:
10:05pm up 62 days, 4:45, 3 users, load average: 62.00, 62.00, 62.00
Based on the output of ps (in ~dtpaul/Public/ps.out for the moment) and:
athena% grep zip /etc/mtab
/dev/rdsk/dks0d5vol /zip dos rw,partition=4,nosuid 0 0
I believe someone forcibly ejected a zip disk and the machine still
thinks its mounted. Then when the find cron job runs to
clean out dead.letter on local disks, it's getting stuck
in disk wait and driving up the load, thus causing sendmail to
decide the load is too high to warrant running the queue.
After asking on -c consult, we decided rebooting the machine
would be the simplest fix.
What should have happened:
The machine should've dealt properly.
One solution might be to specifically exclude /floppy and /zip from
the find jobs. Or just punt the "remove core and dead.letter" cron job.
This might be something worth considering on Suns as well?
--
Emil Sit / Bronx Science '95, MIT '99 -- SIPB, Athena Consulting, LCS/PDOS
