[1957] in Release_7.7_team
minutes, 10/20
daemon@ATHENA.MIT.EDU (Dan Winship)
Wed Oct 20 15:08:21 1999
Date: Wed, 20 Oct 1999 15:07:04 -0400 (EDT)
Message-Id: <199910201907.PAA39618@antharia.mit.edu>
From: Dan Winship <danw@MIT.EDU>
To: release-team@mit.edu
Attending: othomas, ajfox, rbasch, ghudson, danw, kcr, aurora, miki,
wdc, tb | jweiss
1) Outage Tuesday
1. DoS (packet flood) against dialups
2. Dialups stop acking zephyrs
3. Zephyr server unacked queues get very long
4. Zephyr servers start swapping, appear to hang
5a. bug in zhm kills zhm on some machines
5b. olcd hangs trying to zwrite/zlocate
5c. POP servers have problems because of zpopnotify hanging
5d. attach loses trying to get subs, so logins hang
Greg submitted a patch to fix 5a (and Garry Zacheiss sent Cluster a
list of all public workstations whose zhms needed to be restarted).
Ops may be thinking about 5b.
5c is in network-land.
Dan will submit a patch for 5d. ("Oh, that's what that mystery code
in the old attach source was there for.")
Some discussion of trying to fix the zephyr servers to not lose in
this case (maybe by not accepting zephyrs for users on hosts with
lots of queued zephyrs already), but Greg says the zephyr code is
big and gross, and it won't be easy.
2) Feedback forum
Ideas:
Do an lpq at logout time and offer to cancel queued jobs.
(Part of Dash logout interface?)
To deal with the abandoned print jobs problem. But what about
jobs that have already finished printing?
Zip drives tend to not work.
athinfo query to check zip drive status on a machine?
have larvnet monitor zip drives?
TeX uses bitmap fonts that look bad when you run the output
through Acrobat distiller.
There's a stock answer about how to use PS fonts to make it
work better, but we can't make that the default, because it
looks worse in the normal printing case.
Maybe just make the PS font workaround easier?
Low turnout to the forum... maybe just bad timing?
(If we do an Athena Desktop forum, maybe have it over IAP.)
3) Bug tracking update
We've started flagging transactions in the bugs discuss meeting.
Hard to say if it's working, because Greg has been paying extra
attention to the bugs meeting since we started.
Greg has looked at the Debian bug tracking system more, and it
actually does support merging bug reports (which we thought it
didn't)
4) Private Workstations update
Bill put Miki on this with Dan because "they have overlapping areas
of knowledge" or something like that. (Miki has talked with other
unsatisfied private workstation customers, and people who don't use
Athena at all even though they'd like to, because our current model
can't be made to meet their needs.) The scope of the Discovery
project is no longer as closely tied to the current Athena
install/update system. We might decide that to support our
customers we need to totally rewrite it. It is possible that we
won't be able to finish it by 8.4, in which case we'd probably want
to throw some un-Discovered hacks into the release to fix the
problems we personally know/care about.
5) Desktop update
No progress in the last two weeks (Bill on vacation / Greg sick).
Greg Anderson wants a meeting. Karl will deal.
Doc deadlines:
Crash and burn is fine for Heather (will give her time to
redo screenshots and rewrite a lot of the intro stuff).
Gary Dryfoos will need longer than that for minicourse changes.
Maybe we'll need to get him a student to help deal?
Bill will give Gary (and Jean) a heads up.
6) Sun report
Tuesday, four Sun engineers came to learn about Athena. (No causal
link to the dialup/zephyr outage has been established. :-)
The meeting went well. They realized they want to collaborate with
us on things like netbooting, filesystems, name service (they like
LDAP now), and support contracts. (We don't want to pay them tens
of thousands of dollars a year for support we aren't using just so
we can get source access.)
7) SGI report
Bill has asked SGI to help debug the Irix crashing problem. We
don't expect them to be useful.
They are being useful in getting X source (so we can debug the
GNOME crashing problem), but that's because that's the Cray people.
(They aren't sending us all of the GL bits because that's Super
Seekrit, but they'll send us .o files.)
CEASE update: Bill thinks we're going to decide to not buy any more
SGI hardware and keep the Indys for an extra year.
8) Linux update
The machines arrived... with 2-button mice. We want 3-button mice
for the cluster ones.
Thomas will be trying a full automated RPM build today. We're still
a few steps away from having a testable release.
Need some tweaks to deal with replaced vendor binaries (eg Linux
/bin/login vs krb5 /bin/login). Thomas is also going to investigate
whether we can use PAM to get the Linux login to do everything we
need. Another possibility is to not overwrite /bin/login and just
change the getty (or whatever) config file to run
/bin/athena/login.
Package release numbers... Greg and Thomas took this offline
Updates will be done by having a directory full of packages. Can
use the machine-independent part of the Athena update stuff, just
needs different OS-specific stuff.
Would be nice to still have os and srvd images in AFS for grepping
through, grabbing replacement binaries, etc. Not necessarily the
port maintainer's responsibility though.
9) Private Workstation Owner's Guide
There's a draft combined Sun/SGI PWOG. Heather will be putting it
on line soon and sending mail so people can proof it.