[6068] in Release_7.7_team
CVS auto-updater
daemon@ATHENA.MIT.EDU (ghudson@MIT.EDU)
Mon Sep 29 13:10:22 2008
Date: Mon, 29 Sep 2008 13:10:08 -0400 (EDT)
From: ghudson@MIT.EDU
Message-Id: <200809291710.m8THA8dY009167@outgoing-legacy.mit.edu>
To: release-team@mit.edu
X-Spam-Score: -0.245
X-Spam-Flag: NO
People on the source-wash list just got spammed with a whole bunch of
very large mail messages from equal-rites. My apologies. I hope it
didn't interfere with the MIT's email infrastructure.
The problem is this: AFS seems to occasionally malfunction for the CVS
locks directory, giving a timeout error and leaving behind a stale
lock. (This might be as simple as the update conflicting with the
weekly server restart; I'm not sure.) When this happens, the "cvs
update" processes run out of cron start piling up, waiting forever for
the stale lock to disappear.
Today I killed a large number of piled-up cvs processes, resulting in
a bunch of large logfiles (filled with "waiting for lock" messages)
being sent to source-wash.
I have disabled the wash update on equal-rites for now. This means
the CVS checkout of the Athena 9.4 tree will need to be updated by
hand during the preparation of 9.4 patch releases. I probably won't
be fixing this problem any time soon, but the impact of not having a
source-update cron job is pretty minor.
The scripts at issue are in /afs/dev.mit.edu/service/wash, for the
record.