[1450] in Release_7.7_team
notes, 8/11
daemon@ATHENA.MIT.EDU (Whatever you wish...do so to them)
Wed Aug 12 15:24:52 1998
Date: Wed, 12 Aug 1998 10:54:42 -0400
To: release-team@MIT.EDU
From: "Whatever you wish...do so to them" <mbarker@MIT.EDU>
attending: Greg Hudson, Bob Basch, Bill Cattey, Heather Anne, Oliver Thomas,
Abby Fox, Jonathon Weiss, Mike Barker
1. 8.2.9 is out now. 8.2.10 appears to be ready with the fixes for openGL,
but we are waiting for a test case.
2. Need to open a bug report with Transarc in regard to the apparent
problem where an 02 client crashed an AFS server.
3. We discussed support for private workstation owners. One point is to
make sure private workstation owners are on the cluster manglers mailing
list. Oliver volunteered to cross check Naomi's list with the cluster
managers membership.
4. We discussed the apparent slow update. A back of the envelope
calculation suggests that with approximately 1000 Suns in the release and
roughly 100 MB per machine, we had about 100 GB to transfer from four
replication sites. If each site is trying to ship 25 GB over 10 Mbit
ethernet, this means we need roughly 25,000 seconds if we can use the full
bandwidth. That's over six hours. The actual time of 10 to 12 hours seems
quite likely -- we probably did not have perfect balancing, all of the
bandwidth, and there were also the SGIs attempting to update.
For next year, or the next time we have an OS update, we recommend splitting
the update. There are two basic separations. First, we should do the
public machines, then the private machines. Second, we should do each
platform separately. This may result in spreading the update over four
nights, but it should make sure that each completes relatively rapidly.