[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
[cobalt-users] Qube 2 hang solution found
- Subject: [cobalt-users] Qube 2 hang solution found
- From: Jim Frost <jimf@xxxxxxxxxxxxxx>
- Date: Wed Nov 14 00:22:19 2001
- List-id: Mailing list for users to share thoughts on Cobalt products. <cobalt-users.list.cobalt.com>
Awhile back I was poking around trying to see if anyone knew why my Qube2
was hanging every couple of days after doing an OS upgrade and patch
install. I got no bites, so I paid for a support incident.
They checked the logs and (whattaya know) there wasn't anything in them.
(If there had been, I would have had a clue what the problem is. :-) They
suggested running top in a telnet session so we could get a snapshot of the
failures. I did this for a month or so and caught several failures.
In all cases an imapd was running, always with one particular user, an
Outlook Express user. Now, there are only two users on the system and the
other user (me) uses a JavaMail based client which is far easier on the
system than OE. OE makes a new connection every time it does a "new mail"
scan, which is wasteful and particularly expensive since the Qube boxes use
the (horribly inefficient) UW imapd server. This user's mbox was
moderately large, several megabytes in size, but nowhere near as large as
the other user's mbox (21MB or so).
In a couple of cases glimpseindex was running.
They had two recommendations. First was to disable glimpseindex.
Apparently glimpse cannot deal well with "files with certain characters,
like spaces, in their names" and will eat up a lot of memory when it runs
across them. It can eat enough up to exhaust the system and hang it.
Seeing as I have no use whatsoever for glimpse, I disabled it.
The second was to have the user prune her mbox. We also did this.
The box has not gone down since then, now up for 20 days. I cannot
positively identify the culprit but if you run into this problem then I'd
start with disabling glimpse and, if that doesn't do it, run top and if
imapd shows up in the snapshot then have the user(s) whose mboxes are
indicated do some pruning.
Whatever the culprit, this did not happen prior to the OS upgrade despite
previously running glimpse and the same mbox.
Hopefully this will help someone else out. I must say that I'm pretty
irritated at having to pay to resolve an incident caused by the failure of
the standard software and configuration. I very specifically left the box
in stock condition rather than replacing things like imap support with more
efficient/reliable counterparts. Grr. So much for that approach.
jim