[460] in I/T Delivery
Pismere Project Status Report as of 3/30
daemon@ATHENA.MIT.EDU (Thomas L. Thornton)
Tue Apr 3 16:29:25 2001
Date: Tue, 3 Apr 2001 16:29:13 -0400 (EDT)
Message-Id: <200104032029.QAA04130@the-rim.mit.edu>
From: "Thomas L. Thornton" <tomt@MIT.EDU>
To: delivery@mit.edu, integration-ptl@mit.edu, dennis@mit.edu, njj@mit.edu,
jdb@mit.edu, rar@mit.edu, vkumar@mit.edu
CC: pismere-team@mit.edu
Project Name: Pismere Delivery, Integration
Team Members: Danilo Almeida, Joe Calzaretta, Qing Dong,
Paul Hill, Wael Hishmeh, Dave Tanner, Tom Thornton
Report Date: 30 March, 2001
Submitted by: Tom Thornton
URL: http://web.mit.edu/pismere
General
-------
Building the production domain requires weeks more than expected, so
cluster machines deployment to DUSP is still at least a couple weeks
away. Server machines and RIS installation servers are installed and
resolvable through DNS. We have one new subnet, W20, under DHCP
coverage, but need others.
The major features we need to deliver to DUSP are DHCP coverage
from W92 to N42, 9 and perhaps 39; and an RIS image with the AFS
loopback/redirector. Applications in lockers awaits AFS.
Accomplishments in March
------------------------
- Hardware: We have a spate of broken switches, a server and the DLT.
Wael returns one of the two 2x8 matrix switches and awaits a Dell
recommendation on a replacement switch; he put a quad switch on the
server running win.mit.edu, and it is still up. Wael sends email and
makes several calls in to Dell. Tom writes to Dell about several
support technical and support matters, to which Sue Green promptly
replies - Dell agrees to replace the server and send a switch quote.
Wael is repairing the tape drive, and diagnosing backups, which may
involve Retrospect bugs. Danilo speaks to Meeroh about how the Mac
people perform backups.
- On servers, Wael finds DNS settings were innacurate due to multiple
net interfaces, so he removes the superfluous NICs. The machine room
router is confused, so NetOps reboots it, and it works. Thanks to
Chris, DNS tables are up and we can see all new hosts. Danilo writes
a script to routinely compare the DC DNS in netlogon.dns files with
the current NetOps zone table in AFS. Although BDIN was stuck in
mixed mode, Wael reset it to the correct native mode.
Dave and Wael install the populator machine in the machine room; Wael
offers to get the IP address. He installs RIS servers, connects them
to the net, and joins them to the win domain. Dave installs and tests
LDAP code on the populator machine, requiring an account manually
entered by Garry.
Paul reinstalls the DIN and BDIN domain controller servers, and
documents the process. He extends the AD schema with Services for
UNIX, InetOrg, EduPerson and ExchangeServer. He establishes trust
relationships, admin accounts for the staff, sets up BDIN to begin
replicating. The third server waits until the bulk load. Joe
installs the syslogd and audit log policies. Wael takes on organizing
and keeping records of machine room installations.
Dave starts populator bulk and incremental loads, plus AD
synchronization. He debugs a three-second-per userid delay. He gets
a new IP address and test keytab for the machine, and submits a new
keytab request. He works with Paul on an LDAP problem. In any spare
cycles, he plans to start documenting the populator programs,
utilities, and attributes. He studies putting cluster information
from Moira into the AD.
Population is working in the win domain. Replication is not yet
happening in production, which Paul and Dave diagnose. He has users
and groups loaded, but the groups need more specification before he
can populate them. Paul, Danilo, Joe and Dave are working on this,
perhaps extending Moira. Dave will continue using and improving
bulkload utilities and get those changes into winad.
Wael further documents server installation. He troubleshoots
replication. He installs W2K server on the remote install servers RIN
and RIN2 and he made them join the domain. File replication services
to duplicate RIN onto RIN2 needs work.
- AFS: Davie works on AFS debugging, and contacting Transarc and
OpenAFS to figure out how we can merge our changes into those source
trees. Paul relates a discussion he had with Davie, that our AFS
still depends on netbios naming. He says that ultimately, we have to
get rid of the machine name.
- Paul and Tom keep pinging the MS consultant Thom to let him know we
still would like his help with an RIS solution.
- JJV tests standalone apps. He makes a pass through the Early Look
document. He looks at the Logo requirements to see if our directory
plans are compliant, and looking for other issues such as when we
should use "Common Files." We decide to use the following
installation directory for common files:
%ProgramFiles%\Mit\Shared Files
JJV also runs SWRT tests for KLP and WinZephyr. He works on Zephyr
documentation and has many edits. Heather will help him, and assist
or take over from Janet. JJV suggests using a label for documentation
to indicate the intended audience for a document. Colorado uses an
icon. Tech Pubs *may* adopt this, but for now intend to keep writing
documents aimed at the lowest-level user. This leaves to Pismere the
tasks of writing documentation for students, operators, and developers
and maintainers.
- Paul reports the ITS project is back. SAP rescheduled their Linux
product delivery, which will require Paul, Danilo and Joe to get back
to work on the NT version.
- Joe enables building installers for the separately-installable and
-removable parts of the pismere msi, which used to be called
components, but now are called "features." He puts a feature flag
into the installers, and finds that once a Pismere machine installs
the OS, other feature installers may take a very long time to
complete, at least several minutes. He suggests displaying a "Please
wait" message before initiating these. He is working on group
policies in the win domain, and finds he can crash afsd while fetching
scripts. Paul will seek afsd diagnostic tools.
- Joe has ntsyslog working, but finds it is difficult to set logging
at a granularity that flags a directory for specific access modes.
Danilo wants to rename this service to eventsyslogger. Joe has
self-maintenance working on the AFS3.6 installer, to add and delete
registry values. He renames all references in it to eventsyslog and
looks at the syslog service as a separate product, including installer
and documentation. He talks to Jon Hunt about how to release and
distribute it, and puts it into a subdirectory of the pismere locker.
- Joe describes ntmaintenance as a service that runs a cleanup script in
the system context at logout, like reactivate. It also shadows the AD
object structure for a daily one-time script to run, desynched, when
no one is logged on. Joe agrees that this should work with a higher
than daily frequency. Danilo suggests a broader scope of running
scripts, perhaps to include a Pismere version number, so we might move
toward a mechanism supporting update. This demands a requirements
document, which Joe writes, and he makes the suggested changes.
- Qing completes the whodat service control panel applet. She gives
it to Jon Hunt and checks it into pismere-test.
- Jamie does a KLP cleanup, and sends a Beta 4 version to Bryant of
the SWRT; he writes a KLP design document and implements mk in the
build tree. He works on Kerberos library initializations, AFS, and
MSI validation, which Joe and Danilo will consider for the build
script. Over the past year he has sent five bugs to Wise without a
response.
- We invite ITLT to take a Pismere RIS install. Volunteers are:
Mike Barker
Dennis Baron
Bob Ferrara
Theresa Regan
Greg Aanderson
- Tom and Paul guide Phil Thompson through a W92 tour and explain the
difficulties of W2K domain servers
- New problem: Joe is trying to add an SPN to each installed machine.
and requires changing the DNS suffix, which can be changed only
manually. This relates to work we have requested from Thom Robbins.
It will continue to be coordinated by Paul, who is also looking for
the RIS Project Manager. Joe and Danilo will continue to refine RIS
specifications.
- Wael asks Brad for a status on the CSG report. There are are
initial drafts, but no results, so he writes other members requesting
a preliminary report, and they members agree. The whitepapers are
revised and approaching publication.
- Danilo and Qing work with Meeroh and Marshall to use their server
for bugzilla. Qing looks into it, works with Luke's code to interface
with an MySQL database, tests it and examines the Mac team's fields to
see which to add.
- Qing and Ian fix Zephyr known bugs, very rapid work. We still need
to rename initialization files to winzephyr_foo instead of zephyr_foo.
- Ian works on the local IMAP proxy and puts it into the build tree.
He styles it as a system tray, called KMP.
- Jweiss attends for a cross-team visit.
- Tom follows up Support contacts by touching base with Oliver.
- Tom is trying to gauge the number of UROP staff to look for this
summer. Some watchmakers have responded to him, and he works with
Cana in the recruitment effort. An application page is up and we see
at least a start to emailed submissions.
- Tom has been in touch with Phil Long who asks that we perform a
machine install for Alex as soon as possible. Tom explains that we
do not yet have the production domain up, and that Alex is high on the
list of first installs outside of W92.
- Danilo creates a new set of directories to contain specs and
descriptive documentation for the Pismere machine installation. It is
in the repository at /pismere/pismere/doc/site, and he writes a script
to check it out to the pismere locker under the dev directory. Paul
suggests documenting what will be installed via group policy
assignment right after the 1st reboot.
- NetOps notifies us that DHCP service supporting RIS is now available
qin W20, so the SIPB machine may take an installation when we get the
win domain up.
- Wael speaks to TomC about switches, who is checking on its
compatibility with channel extenders.
- Danilo digests beta discussions on MS XP licensing. This may affect
license cost projections.
- Paul says the plans for a Microsoft workshop in June must accomodate
attendees to a MS Sales conference the 14th, so he and Brad will
contact MS Sales.
- Garry comes in to discuss LPRng and Moira. We wonder how Athena
filters PJL files, so he will check on it and get back to Jamie.
Goals for April
---------------
- DHCP: Serve N42 and 9. Discuss with TomC whether we can support
multiple RIS servers.
- Production AFS installer.
- Install ITLT machines.
- Deliver pilot to DUSP.
- A new task we should consider would be a port of renew.
- Kevin is looking at our internally-developed ftp, and should compare
it to kftp.
- Continue MS consulting.
- Kent Post of LCS still needs a meeting with Paul and Tom to explain
the difficulties of W2K domain servers.
- Dave may help with a new task we should consider - a port of renew.
Longer Term Goals
-----------------
- Get help from Garry and TomC to put non-null Kerberos instances into
Moira.
- Generic ftp requirements.
- Release Zephyr, Plop and KLPR.
- Generic logging library proposal - package NT syslogger for other
schools, and get tech pubs help from here or a requesting school.
- We need to do further work to populate AD attributes, such as white
pages info, from the MIT data warehouses.
- Continue planning post-pilot work.
Issues
------
- Will ITS work again transfer resources from Pismere a few weeks? This
may slip pilot debugging.
- When shall we get MS SP and Hotfix source?
- How do we sync our AFS code with Transarc?
- When does the AUI Dash & gnome require Pismere revision?
Additional comments
-------------------
Barring unforseen AFS or DNS obstacles, we should be able to deliver
in a few weeks.
