TMB Minutes 28/10/09
Andrei, Xavier, Jeff Maite, Steven (chair), John, Massimo, Oliver
(minutes)
* Minutes of last meeting
JT - there was an action to poll sites on upgrade path from 3.2 -> 4.0.
once the minutes are circulated, we can get Daniele to make the broadcast.
JT - from steve traylen, I understood that UMD requires that packages
are OS conformant, and thus relocated to /usr.
OK - UMD has not stated that his will be the *only* way packages will be
made available. Current packaging should still be OK too.
* Actions
SN - where are we with enabling MPI SAM tests?
MB - will check if this is done and if we are opening tickets.
- will write an update
* Product Teams
OK briefly describes the doc here:
https://edms.cern.ch/document/1041827/1
SN - We don't need to discuss the detail as this is not a new document.
Any questions or comments?
[None]
SN - The TMB endorses the document
OK - Now that we have endorsement, SA3 can finish the tooling required
to allow PTs to manage parts of the release process. I hope we can announce
something at the next TMB.
SN - Should add a note about the interpretation of fields in jra1
taskplan
OK - we'll add this, and the document under discussion today, to
MSA3.4.2, updated release process milestone, expected PM20.
SN - Grnet will do software repo within EGI - they will need this info.
Will have to do handover next year.
* AOB
SN- staged rollout, was there a problem with site involvement?
MB - we move on with sites, we have a list of sites assigned to cover
each service. First release is being prepared.
SN - what about that rollback? problem?
OK - no problem with rollback, preparing the susequent release was more
complicated though, and there was a delay.
SN - a request went to NGIs to ask if they wanted to be involved in
activities like staged rollout, we'll see what comes back. If none are
interested, we'll have to make a decision.
* Gilda security
SN - Gilda are looking how they can be fully integrated with production
monitoring, this seems to be going ahead.
MB - at SA1 coord meeting, discussion was so complex we asked what the
orignin of this Gilda decision was?
SN - there are two goals. i) make sure gilda resources are monitored at
the same level as prod infr, so are suitably stable. ii) remove
partition between prod infr. and training infr. - no guarantee of
funding for separate training inf. in EGI, and NGIs will want to use
their existing resources to support training.
SN - stage 1, monitoring, is going ok, but has raised issues about the
gilda CA.
MB - main concern was the anonymity of jobs from Gilda.
Gilda assumes either site admin or trainers are responsible.
this is not clearly legal
these issues must be understood before we can move
ML - with present situation, we have logically distinct infr. If they
run in this anonymous mode, and the jobs are running on the prod infr,
it's a problem.
MB - at the moment the training infr. is quite isolated
ML - even before, training might be running on the same box as eg biomed...
SN - depends on the country. Italy has close integration, others are
different.
JW - gilda CAs should undergo eugridpma treatment.
SN - we have a portal security policy, defines what info needs to be
retained by the portal. If that's where the concern is, that's what we
should analyse first. The second point quality of identity mapping, but
now with things like SLCS overhead of getting a certificate is much
lower.
JW - Christoph can inform on how that went with the Swiss CA.
JT - security was so hard before, gilda were obliged to rip it out.
MB - sites will not accept things as they are
JT - aren't we pushing monitoring out the the regions in the EGI world?
Good time for italy to implement monitoring themselves
SN - t-infr. is pan european
MB - We discussed if gilda part of the infr. or just a VO. They say they
are more an infr. than a VO.
SN - Their current ad-hoc monitoring has resulted in a low quality infr.
JT - Are you sure?
SN - I looked at their monitoring (via QR), they were using an italian
monitoring solution, 1/3 sites weren't reporting, 1/3 had errors.
ML - was the problem with monitoring, or the infr?
SN - What people were training on was different from production,
reducing its effectiveness.
MB - ROCs are worried that training sites will bring down their
reliability figures. Gilda said that they have prod sites.
SN - this is a challenge that gilda either have to accept or not (in
which case they stay in italy and have no international funding)
SN - this is ongoing, but good to have a status
MB - will forward a summary from Dave Kelsey and Romain Wartel.
There are minutes attached to this event.
Show them.