[3362] in Release_7.7_team
Disconnected Linux-Athena status and open issues
daemon@ATHENA.MIT.EDU (Derek Atkins)
Tue Jun 25 15:22:01 2002
To: release-team@MIT.EDU
From: Derek Atkins <warlord@MIT.EDU>
Date: 25 Jun 2002 15:21:58 -0400
Message-ID: <sjmr8iv2mg9.fsf@kikki.mit.edu>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Hopefully we can discuss these at tomorrow's meeting...
-derek
CURRENT STATUS:
The status daemon and network event daemon are implemented and appear
to work fine. As network and apm events occur, various scripts are
executed within the system and user environments. Events are
accumulated during system boot and processed once the status daemon
starts. Afterwards, events are handled in real time.
A few scripts are already written, but more integration is required.
Currently implemented for the system:
- rename the machine at net-up events
- flush the email queue at net-up events
- force a time resync at net-up events if the time > 240 seconds off
- start zhm at first net-up event [1]
- run "fs checks -all" at net-up events [2]
And the following for the user:
- fix xauth cookies (if they are used)
- renew tickets when the ip address changes or the tickets are expired
- reload zephyr subscriptions (zctl load) at resume and net-up events
OPEN ISSUES FROM ABOVE:
[1] Zephyr (ZHM): the system scripts still try to start zhm at boot
time. Zhm, as currently implemented, will hang indefinitely at boot
time if the network does not exist. There are a number of options to
fix this, but we need to discuss how zephyr should behave in
disconnected mode.
[2] AFS: As currently configured AFS will not start without a network.
We need a hook to start AFS with dynroot (and also, optionally,
afsdb). I should note that this requires the OpenAFS in Athena-9.1
and will not work with 9.0, but that shouldn't be a showstopper.
TASKS TO DO AND MORE OPEN ISSUES:
- Email Delivery: We need to run a sendmail daemon locally that will
properly queue mail when off-net.
- Kerberized login: timeout quickly when network is down
- Time Sync: run ntp in some sane configuration
- Auto Update: Lots of tasks and issues here. We definitely need to
set AUTOUPDATE=false, but should we run the equivalent of
reactivate's check-for-update script during net-up events? Do we
need an automatic hot-fix update system?
- File Sync: How do users sync files (e.g. homedir)?
- User Event Hooks: do we need to provide the user with hooks into
the event system so a clueful user can (easily) create their own
functions to be executed at various events?
So, what should we do, and how do we approach these issues? In
particular, how do we support a single "distribution" that may or may
not have disconnected support?
--
Derek Atkins, SB '93 MIT EE, SM '95 MIT Media Laboratory
Member, MIT Student Information Processing Board (SIPB)
URL: http://web.mit.edu/warlord/ PP-ASEL-IA N1NWH
warlord@MIT.EDU PGP key available
