Incidently, the RPM for the new UK e-Science CA is one of those mentioned
in the minutes without being named, that should appear in the EDG
repository this week.
Cheers,
Andrew
Minutes for ITeam Meeting
Thursday, February 6, 2003
Attendees:
WP1: Eli
WP2: Paul
WP3: Laurence
WP4: Maite
WP5: Owen
WP6: Markus, Emanuele, Cal, Steve T., Alessandro C., Fabio, Nadia, Anders
WP7: Robert
WP8:
WP9:
WP10:
LCG: Louis
Chair: Cal
Minutes: Cal
I. Results of Review
The review went very well. There were no serious concerns from the
reviewers. The demos themselves went exceedingly well; the only
problem being two failed job submissions during the WP9. These made it
to lxshare0227 but then failed in the processing when trying to access
a file on the SE which should have been available.
II. Status of Testbed (application and development testbed)
CERN: Reopened testbed this morning for all users otherwise no changes to
the application or development testbeds.
RAL: Opened up the testbed to everyone yesterday evening. There is now a
new CE at RAL which is a gateway onto the production batch system at RAL.
There are around 270 CPUs there. The new CE currently has a very limited
mkgridmap.conf file and no associated SE.
Lyon: Reopened testbed this morning no other changes. On the
development testbed a new BQS information provider will be tested.
There is a problem when the BQS master cannot be contacted. Currently
the queue information is not invalidated in this case.
CNAF: Still has limited mkgridmap.conf file from the demo. This will be
opened up immediately after the meeting.
ACTION: Site administrators are reminded to open up the mkgridmap
configuration files as soon as possible. Cal will send out a mail to remind
the site administrators to do this.
III. Discussion of Issues for Globus/RedHat Upgrades
Autobuild with RH7.3 and Globus 2.2.3+.
------------------------------------------
WP1: Already tried WP1 software with old version of globus and RH7.3 there
are no problems here. For the new version of Globus there are many changes
in the ssl_utils. This has been looked at and there will be no problems
having the software use the new release.
WP2: The WP2 software all builds OK under RH7.3 There are some dependency
problems which are being worked out.
WP3: The software builds and runs on RH7.3, but currently isn't building
with autobuild system. These are minor problems that Laurence will work out
will Yannick.
WP4: Everything works; some problems with autobuild. There were some
problems with LCAS due to linking; the code is being updated to fix this.
Martijn has built a new EDG gatekeeper with the new Globus. The testing ran
into a problem with the configuration. He is in contact with Anders; Anders
has promised to provide the new rpms by the beginning of next week. The
configuration for the gatekeeper/jobmanager will be in line with the "new"
format for the MDS and GridFTP.
[After meeting: The error is fixed now. Yannick P. changed something
in the globus configuration.]
WP5: All of the RH7.3 stuff builds under RH7.3; doesn't currently work
completely with autobuild. Owen will follow this up with Yannick to get
this working.
WP6: Need to check that edg-gridftp commands work correctly with new Globus.
WP7: Everything OK.
EDG 1.4.3 and LCFGng:
--------------------------
Bugs reported to LCFGng configuration are all fixed except there is still
problem with what is the correct thing to do with the '\"' in the various
source files. There are a few changes in the rpms which need to be
incorporated. These will be checked after the meeting. The 1.4.3 LCFGng
configuration will be retagged to include these changes.
Markus and Emanuele noted that some of the preprocessor macros are no longer
working with LCFGng. The questions are were there major changes in the
preprocessor, and is the preprocessor configuration documented somewhere?
Maite says that the documentation exists and will send out the pointer to
it. She also said that if macros have been broken between versions to
report these problems via bugzilla.
[After the meeting: The preprocessor syntax is not documented in LCFG
(it is supposed to be the ANSI C standard one). The differences were
due to post-processing of the file by LCFG.]
ACTION: In anticipation of the upcoming integration week, site managers (of
the development testbed) should try to get an LCFGng server running on RH7.3
and try to install the current EDG 1.4.3 with LCFGng. Having this available
will make the integration of the new Globus go faster.
[Warnings: Don't do the server upgrade if you still have LCFGng
client nodes running 62 on your testbed. With a RH6 server you can
manage both clients (RH6 and RH7).]
Security rpms for new CAs:
-----------------------------
There were about four new CAs approved. Anders will try to make the new
rpms for tomorrow and these will be included in the new release. The
production of these rpms will be moved into CVS to allow this to be
automated. Anders will get in contact with Yannick to work this out.
EDG Gatekeeper with new Globus
------------------------------------
See above for the current status. The gatekeeper will only work with the
new version of LCAS, so this will be incorporated at the same time the new
gatekeeper rpms are provided.
Need for new EDG release before Globus upgrade?
----------------------------------------------------
The overall concensus was that this is not neccessary, but that the
outstanding rpms should be included into the upgrade release. The
outstanding rpms are: new PBS info. provider that properly takes into
account offline nodes and a new mkgridmap rpm. For the mkgridmap the
configuration is slightly different, but shouldn't be difficult to
incorporate into the release.
Anticipating Changes for Globus and RH7.3 Upgrade
------------------------------------------------------
The latest version of MySQL 4 is needed for several pieces of software for
EDG 2.x. It would be better that we upgrade to this version of MySQL before
we continue with the EDG 2.x integration, otherwise we will have to change
software twice.
Eli has checked with the L&B folks about this problem and has been told that
there is no problem upgrading the L&B to this version.
Laurence said that the new version is no problem with the WP3 software.
Default Python Version
-------------------------
Need to determine if the WP1 UI software will work with the default version
of python 2 on RH7.3. Eli will inquire as to whether this does actually
work with the UI software.
There was also a question about the WP1 bash dependency (bash < 2.0). Eli
will test whether the scripts work with bash v2 and remove the dependency if
it is not necessary.
Clean up of Release for Integration Week
------------------------------------------
Removal of unused software: Removing old R-GMA and of FTree. Laurence said
that removing the old R-GMA is fine. Will look into what requires FTree and
possibly remake the rpm to get rid of this dependency.
There has been some desire to clean up the LCFG source files for the
release. As creating these files was very labor intensive, I was generally
felt it would be better to clean up easy things as we go along, but not do a
massive cleanup. This was seen to likely delay progress on incorporating
new features.
IV. AOB
Owen had a question about the interface between GDMP and the storage
element; namely, will it remain the same as we currently have? Paul
believes that the interface will remain the same but will check to ensure
that this is the case.
Markus suggested that we start a new "operations" mailing list. This is
intended to provide important information to the end users without
innundating them with the full volume of email on the ITeam mailing list.
The proposal was that the following information would be posted to this
list:
* ITeam agenda
* ITeam minutes
* Major operational changes or problems
It would be assumed that ALL EDG users would subscribe to this list and read
the mail there. The assumption would be that all users would subscribe to
this list. It was also suggested that a web page also be given. However,
the concern here is that this is yet another thing to keep updated and that
the manpower to do so is not readily available. Also having such a page
would reduce the likelihood that users actually read the mailing list.
To ease the reading of the minutes, we should have a standard section which
discusses the operational issues.
On a similar point, Owen suggested that we try to centralize the user
support in another place than the ITeam mailing list. I think everyone
actually agreed that this was a good idea; it was pointed out that this
mailing list already exists but is little used. Realistically it was felt
that for the near term that moving the user support off of the ITeam mailing
list is unlikely as everyone has become used to getting support there. Cal
will bring up at the WP6-FR meeting next week.
Meeting next week is canceled because of the tutorial next week in Lyon.
The next meeting will be Monday February 17 at 9:00 at CERN.
|
