Hi Alessandra/All
Based on my observations I would not conclude that objections that have been raised have not been heard (how else would you explain the MUPJ policy suspension for a year). The problem is that there is no consensus about how best to resolve the problem of identity switching, an issue that we in the UK felt strongly about at the start of the discussions. I think one of the reasons we are still having this discussion is because both sides of the argument are being heard and the dominant view keeps shifting (which probably reduced the focus on developing a definitive and acceptable solution).
You make a good point about the future of glexec and also about what has (not) been done to improve it over the years of discussion. I believe there has been improvement brought about by various security reviews, but the end product is still not what many wanted and there isn't a relocatable version available. However, if it is deployed and being used then there will be more community commitment to keeping it supported.
Recent presentations from the experiments indicate backing for using (something like) glexec. Checking today we now have over half of GridPP sites publishing support for glexec. In the survey I did last year only one (possibly two) GridPP site(s) indicated that they would be unlikely to deploy glexec. I would see this as an indication that we should continue with our present strategy of glexec deployment as WLCG is requesting. You may later be proved correct in your fears, but we need to commit to one path now to reduce the risks we can manage.
Jeremy
On 1 Jun 2012, at 13:01, Alessandra Forti wrote:
> The argument is that no other objection was listened to for the past 6 years. The fact that EMI is now ending is another worry about wasting time on something that has no future.
>
> cheers
> alessandra
>
> On 31/05/2012 23:17, John Gordon wrote:
>> Alessandra and Wahid, in the security TEG there were representatives of all the experiments and some sites. The output was presented to several GDBs, MBs, a whole day pre-GDB, and the WLCG workshop. Slightly more than running it past some other groups. The only remaining argument I hear is 'we've been talking about this for so long we can't possibly do it now'.
>>
>> Regards,
>>
>> John
>>
>>> -----Original Message-----
>>> From: Alessandra Forti [mailto:[log in to unmask]]
>>> Sent: 31 May 2012 18:32
>>> To: Testbed Support for GridPP member institutes
>>> Cc: Gordon, John (STFC,RAL,ESC)
>>> Subject: Re: glexec tests for ATLAS
>>>
>>> > Similarly, WLCG is a collaboration between sites, experiments, and
>>> middleware providers. It is not separate from the experiments.
>>> > For six months the output from the TEGs has been that identity switching
>>> is crucial for identity tracing and separation of workload.
>>>
>>> John, the requirement and glexec have been discussed for about 6 years =
>>> 72 months = 3/4 funding rounds it has gone through reviews and management
>>> boards and still at each stage either sites or experiments refused to
>>> install it.
>>>
>>> On top of the past history we are now almost at the end of EMI without a
>>> clear plan on how each middleware component will be maintained. I know in
>>> the EMI talk it was officially said they will continue but since even DPM
>>> is in a fragile state for what concerns funding I have my doubts on how
>>> this will develop. And from where I stand DPM is a much more important
>>> component than glexec. But hey let's install glexec...
>>>
>>> cheers
>>> alessandra
>>>
>>>>> -----Original Message-----
>>>>> From: Testbed Support for GridPP member institutes [mailto:TB-
>>>>> [log in to unmask]] On Behalf Of Alessandra Forti
>>>>> Sent: 31 May 2012 15:21
>>>>> To:[log in to unmask]
>>>>> Subject: Re: glexec tests for ATLAS
>>>>>
>>>>> Hi Jeremy,
>>>>>
>>>>> Alessandro just got the tests working during CHEP, he knows that not
>>>>> all sites are up and running as we talked about it. The driver behind
>>>>> this is again WLCG and not the experiments so perhaps the GDB is a
>>>>> good place where to have a discussion about deadlines.
>>>>>
>>>>> cheers
>>>>> alessandra
>>>>>
>>>>>
>>>>> On 31/05/2012 11:52, Jeremy Coles wrote:
>>>>>> Hi Alastair
>>>>>>
>>>>>> It is not the case that glexec is installed and working at the
>>>>> majority of UK sites. From where did you get that impression?
>>>>>> The current picture under the ops VO (which I would expect to be a
>>>>> little ahead of what ATLAS sees):http://tinyurl.com/cwef84h.
>>>>>> 7 sites have at least some glexec presence though it is not working
>>>>> for all their CEs.
>>>>>> 7 sites have not got it installed and working but some of these do
>>>>> have previous experience
>>>>>> 4 sites depend on a relocatable version that is not yet working.
>>>>>>
>>>>>> Glexec is up for discussion at the June GDB in a few weeks. We were
>>>>> anyway working towards wider rollout and so the answer to Alessandro
>>>>> is that yes we have plans to install glexec. However, I am sure
>>>>> sites will increase the local priority if a clear indication of
>>>>> experiment plans is given. Is there a date when ATLAS will set the
>>>>> glexec test as critical?
>>>>>> Jeremy
>>>>>>
>>>>>>
>>>>>> On 31 May 2012, at 11:33, Alastair Dewhurst wrote:
>>>>>>
>>>>>>> Hi all
>>>>>>>
>>>>>>> I just received the following email from Alessandro di Girolamo
>>>>> about new ATLAS glexec tests.
>>>>>>> I was under the impression that most UK sites had glexec installed
>>>>> so it confuses me slightly that the tests don't work at the majority
>>>>> of UK sites.
>>>>>>> My knowledge of glexec is very limited but if sites could check
>>>>> their test result and try and work out why they are failing that
>>>>> would be very helpful.
>>>>>>> Thank you.
>>>>>>>
>>>>>>> Alastair
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>> Begin forwarded message:
>>>>>>>
>>>>>>>> From: Alessandro Di Girolamo<[log in to unmask]>
>>>>>>>> Date: 31 May 2012 11:02:04 GMT+01:00
>>>>>>>> To: Alastair Dewhurst<[log in to unmask]>, Alessandra
>>>>> Forti<[log in to unmask]>
>>>>>>>> Subject: glexec tests for ATLAS
>>>>>>>>
>>>>>>>> Ciao Alastair and Alessandra,
>>>>>>>>
>>>>>>>> we just set up a first glexec basic test for all the ATLAS CEs
>>>>>>>>
>>>>>>>> There are now jobs sent with a Role=pilot which try to execute
>>>>> glexec
>>>>>>>> this is what the test mainly does:
>>>>>>>> https://tomtools.cern.ch/confluence/display/SAM/WN#WN-
>>>>> org.sam.glexec.WNgLExec
>>>>>>>> ( in details the code is
>>>>>>>>
>>>>> http://svnweb.cern.ch/world/wsvn/sam/trunk/probes/src/gridmetrics/wnm
>>>>> et
>>>>> rics.py )
>>>>>>>> all the tests are not critical.
>>>>>>>>
>>>>>>>> I'm now launching them on all the sites and I would like to
>>>>> understand the various situations, just to have a global idea on the
>>>>> issues/difficulties (if any), so that we can give to ATLAS an
>>>>> overview based on the real situation.
>>>>>>>> https://sam-atlas-prod.cern.ch/nagios/cgi-
>>>>> bin/status.cgi?servicegroup=%2Fatlas%2FRole%3Dpilot&style=overview
>>>>>>>> I see that most (if not all) the UK sites seem not have glexec
>>>>> installed.
>>>>>>>> Do you have any plan to install it?
>>>>>>>>
>>>>>>>> THanks
>>>>>>>> Ale
>>>>>>>>
>>>>>>>> --- --- --- --- --- --- --- --- --- --- --- Alessandro Di Girolamo
>>>>>>>> CERN IT-ES-VOS
>>>>>>>> +41764870532 (16-0532)
>>>>>>>> http://www.cern.ch/Diggi
>>>>>>>> --- --- --- --- --- --- --- --- --- --- ---
>>>>>>>>
>>>>> --
>>>>> Facts aren't facts if they come from the wrong people. (Paul Krugman)
>>> --
>>> Facts aren't facts if they come from the wrong people. (Paul Krugman)
>>>
>
>
> --
> Facts aren't facts if they come from the wrong people. (Paul Krugman)
|