On Fri, Apr 15, 2005 at 09:45:27AM +0100 or thereabouts, Steve Traylen wrote:
> On Thu, Apr 14, 2005 at 03:55:02PM +0100 or thereabouts, Coles, J (Jeremy) wrote:
> > Dear All
> >
> > Just a reminder that Steve Traylen will be chairing the monthly meeting
> > tomorrow which is focused on the 2.4.0 upgrades. It will be via VRVS (a
> > link to the agenda is:
> > http://agenda.cern.ch/fullAgenda.php?ida=a052215). I apologise if this
> > time is not convenient for everyone but it was one of the few slots
> > available.
>
> Hi,
>
> The VRVS room is the snow room.
Here are the minutes from Fridays meeting. I'm welcome corrections.
Discussions of problems with 2.4 installs.
+ RAL T2
Problems with R-GMA upgrade , solution exists but better one is
being worked on. Other than R-GMA upgrade complete.
+ Durham
Same R-GMA problem, MON box reinstalled as a solution.
Problem with some libraries being files and not symlinks.
ldconfig whinges as a result. Will produce some more notes later.
The install date is missing from the site-def (but a second comment that
it is not needed as it is calculated.)
+ Pete at Oxford.
Upgrade not started.
+ Ben Waugh at UCL.
Part way through the HEP farm , not stuck at the moment.
CCC some problems still but it has started.
The globus port range also missing from the site-def file.
A few things not clear in the site-def file. e.g in the BDII_ZONES.
+ Simon George @ Royal Holloway
Starting a fresh install , configuring site-def file.
+ Yves at Birmingham
The upgrade will start next week.
+ Henry at Brunel
Upgraded to LCG2_3_0 last week on RH7.3
Upgrading this week to SL3 2_4_0 , info provider on CE not working
Some questions about software directory, is this done for you
or must you do it your self? You have to do it yourself.
+ Matt at Lancastor
Upgrade to start next week.
+ Owen at Imperial
CE and WN now upgraded.
Some problems with RHEL over SL3 which slowed things down
but things are resolved.
+ Fraser at Glasgow
CE and torque separation are now resolved and finished.
Same problem with R-GMA.
General agreement that site-def file is not well documented, e.g
BDII (site or top one).
+ Upgrade Procedures
Remember to put in a downtime.
+ Some sites with very long downtime periods?
This was done some London sites and glasgow due to lack
of people. Putting site in downtime is a sensible thing
though.
+ Testing before going back online.
Check gstat and SFT.
While in downtime this still happens.
+ Questions on upgrade procedure,
Is there recipe for running tests?
There is a document discusses the tests. You can
also run the SFTs as well but this is a little tricky.
The link will be sent around as it is less than obvious to find.
No obvious definition of what services should and should not be there.
What services should be running?
A concise list of port changes between releases would be useful
The document tells you which service requires which port but still
there is no list of services per node so this is difficult to
actually implement.
+ Has the release had it had as much testing as it could have?
People say no it has not. (Personally I don't think it was
that bad other than the R-GMA which has caused quite a few problems.)
+ Question on Cern linux , is that what testing is done on?
Yes but if there is difference it is a bug with cern linux.
+ Comment from Lydia.
Possibly some fun getting the java upgrade done. Maybe a
apt-get -f is required to
+ VRVS.
People were happy, we could do presentations but people not sure
what the advantage is. Of course we don't know who could not make
it.
+ Upcoming meeting.
People are happy with the proposed time for the next meeting.
+ AOB.
Brunel hardware post now advertised.
Interested in having two CEs.
Still interest in SGE from Imperial and Durham.
Imperial hope to release it soon.
Lydia: SGE 5 or 6? It will work for either.
>
> Steve
> >
> > It is good to see a number of sites have already upgraded and many more
> > are in the process. Thank you for your support.
> >
> > Kind regards,
> > Jeremy
> >
> >
> >
> > -----Original Message-----
> > From: Testbed Support for GridPP member institutes
> > [mailto:[log in to unmask]] On Behalf Of Coles, J (Jeremy)
> > Sent: 11 April 2005 14:52
> > To: [log in to unmask]
> > Subject: 3-week target for LCG-2_4_0 upgrade
> >
> > Dear All
> >
> > As mentioned in my mail earlier today (regarding the meeting on Friday
> > http://agenda.cern.ch/fullAgenda.php?ida=a052215) 2.4.0 has now been
> > released. EGEE has agreed that it is reasonable to expect sites to
> > upgrade within three weeks of a published release date. In this case the
> > 1st day is 7th April (as 2.4.0 was only tagged on 1st April not
> > released). This means ALL sites should have upgraded to 2.4.0 by
> > Thursday 28th April. Sites that have not upgraded (for whatever reason)
> > will be flagged as failing the critical Site Functional Tests. These
> > tests are used in populating BDII files which means that sites running
> > the old LCG version are likely to be excluded from running jobs.
> >
> > I have to report weekly on how likely the UKI region is to meet the 3
> > week target. Unless I hear otherwise I assume that ALL sites will be
> > upgrading in this period. The GridPP Project Management Board and LCG
> > Grid Deployment Board are very concerned about sites failing to take
> > seriously the need to upgrade in a timely manner, which is why the
> > upgrade times for sites are now being recorded. I am happy to raise
> > reasonable concerns about being able to meet the 3 week window but if I
> > do not hear your views I have few arguments to present! As mentioned,
> > there is now a general consensus across EGEE that 3 weeks is a
> > reasonable period and I am assured that future release dates (the next
> > is 1st July) will be met given experiences with this latest release. To
> > allow some contingency it may be good practice to schedule manpower for
> > the 2nd or 3rd week of the 3 week upgrade period as updates will not
> > take more than a few days - but again this is my perception and there
> > may be problems I can not see. If so please tell me as it is within
> > everyone's interest for Tier-2s to meet resource levels committed to
> > GridPP and LCG.
> >
> > Finally I believe most sites have moved away from RH7.3. For those that
> > have not may I remind you that there are strong security reasons for not
> > remaining with RH7.3!
> >
> > Kind regards,
> > Jeremy
> >
> >
> > -----Original Message-----
> > From: LHC Computer Grid - Rollout [mailto:[log in to unmask]]
> > On Behalf Of Markus Schulz
> > Sent: 06 April 2005 13:07
> > To: [log in to unmask]
> > Subject: [LCG-ROLLOUT] Release of LCG-2_4_0
> >
> > Dear Site-managers,
> > after tests installations by Gergely and Eygene we now announce the
> > release which was scheduled for the 1st of April.
> >
> > You find the release material as always here:
> >
> > http://grid-deployment.web.cern.ch/grid-deployment/cgi-bin/index.cgi?
> > var=releases
> >
> > LCG-2_4_0 is a major release and we hope that you can find time to
> > upgrade within the next 3 weeks.
> > Of course the 3 weeks (3*5 working days) start from the 7th.
> >
> > Please contact us and your ROCs for problems that you encounter during
> > testing.
> >
> > markus
> >
> > p.s. If you see problems with R-GMA after the upgrade, it might be
> > related to the registry at RAL being moved to the new version as well
> >
> >
> > ************************************************************************
> > *******
> > Markus Schulz
> > CERN IT
>
> --
> Steve Traylen
> [log in to unmask]
> http://www.gridpp.ac.uk/
--
Steve Traylen
[log in to unmask]
http://www.gridpp.ac.uk/
|