Hi there, I’m working with LSST folks in Southampton who have code that is updated quite regularly. A while ago, I tested to see whether git could clone at time of execution (I cannot remember which site this was at), and it was a success. However, RAL does not have git installed, as similar code fails at that site. I’ve asked this before and didn’t really get a full answer. Is there any way of knowing which sites have which software installed? It would save new users a lot of pain when trying to get up and running, or even just testing. Cheers Darren ------------------------------------------------------------------------- Dr Darren J. White Applications Developer EPCC, University of Edinburgh Email: d.white@epcc.ed.ac.uk<mailto:d.white@epcc.ed.ac.uk> Telephone: +44 (0) 131 6513415 Room 2411, James Clerk Maxwell Building, Kings Buildings, Mayfield Road, Edinburgh EH9 3JZ, United Kingdom ------------------------------------------------------------------------- The University of Edinburgh is a charitable body, registered in Scotland, with registration number SC005336.
Hi Darren, As far as I know there is no commitment from sites to make available additional packages not in a default sl6/cc7 install (other than packages needed for supporting large VOs such as cvmfs/xrootd/...) Would it possible to use the direct tarball download of the git repo to access the repo via standard http? e.g. for the develop branch on the ganga project `wget https://github.com/ganga-devs/ganga/archive/develop.zip` If not then it may be a good idea to package up the data and upload it ahead of the job as an LFN which can be brought to the WN ahead of the job running, Regards, Rob On 04/09/2017 03:54 PM, WHITE Darren wrote:
Hi there,
I'm working with LSST folks in Southampton who have code that is updated quite regularly. A while ago, I tested to see whether git could clone at time of execution (I cannot remember which site this was at), and it was a success. However, RAL does not have git installed, as similar code fails at that site. I've asked this before and didn't really get a full answer. Is there any way of knowing which sites have which software installed? It would save new users a lot of pain when trying to get up and running, or even just testing.
Cheers
Darren
------------------------------------------------------------------------- Dr Darren J. White Applications Developer EPCC, University of Edinburgh
Email: d.white@epcc.ed.ac.uk Telephone: +44 (0) 131 6513415
Room 2411, James Clerk Maxwell Building, Kings Buildings, Mayfield Road, Edinburgh EH9 3JZ, United Kingdom ------------------------------------------------------------------------- The University of Edinburgh is a charitable body, registered in Scotland, with registration number SC005336.
Cheers Rob! You always have useful answers. I always assumed the git archive was for specific releases, but just managed to figure out the archive path for the specific branch, so running with that now. I will update when it finally runs. Cheers Darren ------------------------------------------------------------------------- Dr Darren J. White Applications Developer EPCC, University of Edinburgh Email: d.white@epcc.ed.ac.uk<mailto:d.white@epcc.ed.ac.uk> Telephone: +44 (0) 131 6513415 Room 2411, James Clerk Maxwell Building, Kings Buildings, Mayfield Road, Edinburgh EH9 3JZ, United Kingdom ------------------------------------------------------------------------- On 4 Sep 2017, at 16:11, Robert Currie <rcurrie@cern.ch<mailto:rcurrie@cern.ch>> wrote: Hi Darren, As far as I know there is no commitment from sites to make available additional packages not in a default sl6/cc7 install (other than packages needed for supporting large VOs such as cvmfs/xrootd/...) Would it possible to use the direct tarball download of the git repo to access the repo via standard http? e.g. for the develop branch on the ganga project `wget https://github.com/ganga-devs/ganga/archive/develop.zip` If not then it may be a good idea to package up the data and upload it ahead of the job as an LFN which can be brought to the WN ahead of the job running, Regards, Rob On 04/09/2017 03:54 PM, WHITE Darren wrote: Hi there, I'm working with LSST folks in Southampton who have code that is updated quite regularly. A while ago, I tested to see whether git could clone at time of execution (I cannot remember which site this was at), and it was a success. However, RAL does not have git installed, as similar code fails at that site. I've asked this before and didn't really get a full answer. Is there any way of knowing which sites have which software installed? It would save new users a lot of pain when trying to get up and running, or even just testing. Cheers Darren ------------------------------------------------------------------------- Dr Darren J. White Applications Developer EPCC, University of Edinburgh Email: d.white@epcc.ed.ac.uk<mailto:d.white@epcc.ed.ac.uk> Telephone: +44 (0) 131 6513415 Room 2411, James Clerk Maxwell Building, Kings Buildings, Mayfield Road, Edinburgh EH9 3JZ, United Kingdom ------------------------------------------------------------------------- The University of Edinburgh is a charitable body, registered in Scotland, with registration number SC005336.
Hi Darren, Just a question: why you do not stick with the cvmfs repository? There is ‘lsst.in2p3.fr’ repository (maintained at Lyon and replicated by Stratum1 at RAL) where I’d assume all software should be provisioned from. Is this code you’re talking about only interesting the UK sites? In this case you may want to have another (UK regional) ‘lsst’ CVMFS repository. Anyway we at RAL have the policy to install a minimal distribution on our WNs (and git is not part of it), assuming the VOs are provisioning their software via CVMFS. Regards, Catalin Condurache RAL Tier-1 Services From: gridpp-dirac-users-bounces@imperial.ac.uk [mailto:gridpp-dirac-users-bounces@imperial.ac.uk] On Behalf Of WHITE Darren Sent: 04 September 2017 15:54 To: gridpp-dirac-users@imperial.ac.uk Subject: [Gridpp-Dirac-Users] Sites with git Hi there, I’m working with LSST folks in Southampton who have code that is updated quite regularly. A while ago, I tested to see whether git could clone at time of execution (I cannot remember which site this was at), and it was a success. However, RAL does not have git installed, as similar code fails at that site. I’ve asked this before and didn’t really get a full answer. Is there any way of knowing which sites have which software installed? It would save new users a lot of pain when trying to get up and running, or even just testing. Cheers Darren ------------------------------------------------------------------------- Dr Darren J. White Applications Developer EPCC, University of Edinburgh Email: d.white@epcc.ed.ac.uk<mailto:d.white@epcc.ed.ac.uk> Telephone: +44 (0) 131 6513415 Room 2411, James Clerk Maxwell Building, Kings Buildings, Mayfield Road, Edinburgh EH9 3JZ, United Kingdom -------------------------------------------------------------------------
Hi Catalin, Even before this project got underway, we were warned to not do that for code that is in such active development. The time it takes for CVMFS updates to filter down can take far too much time for testing. I spent the last few days in Southampton last week and we were pushing changes to GitHub several times an hour that could instantly be available for testing. That’s simply not possible from what I understand/was told about CVMFS. Also, the Lyon repo is not available at all sites for LSST folks, as I found out the hard way. The specified repo I was told after many days of failure is apparently /cvmfs/lsst.opensciencegrid.org<http://lsst.opensciencegrid.org>, which we successfully use to set up EUPS, which we then, in turn, use to set up the many packages needed for this specific code (which are also not available on either the LYON or opensciencegrid repos and so packaged into a tarball on an SE). This is another issue for new folks trying to work on GridPP. There’s no real clarity of what CVMFS repos are available, what sites you can run at, what tools they include both at the site or on the repos, etc. Cheers Darren ------------------------------------------------------------------------- Dr Darren J. White Applications Developer EPCC, University of Edinburgh Email: d.white@epcc.ed.ac.uk<mailto:d.white@epcc.ed.ac.uk> Telephone: +44 (0) 131 6513415 Room 2411, James Clerk Maxwell Building, Kings Buildings, Mayfield Road, Edinburgh EH9 3JZ, United Kingdom ------------------------------------------------------------------------- On 4 Sep 2017, at 16:38, catalin.condurache@stfc.ac.uk<mailto:catalin.condurache@stfc.ac.uk> wrote: Hi Darren, Just a question: why you do not stick with the cvmfs repository? There is ‘lsst.in2p3.fr<http://lsst.in2p3.fr/>’ repository (maintained at Lyon and replicated by Stratum1 at RAL) where I’d assume all software should be provisioned from. Is this code you’re talking about only interesting the UK sites? In this case you may want to have another (UK regional) ‘lsst’ CVMFS repository. Anyway we at RAL have the policy to install a minimal distribution on our WNs (and git is not part of it), assuming the VOs are provisioning their software via CVMFS. Regards, Catalin Condurache RAL Tier-1 Services From: gridpp-dirac-users-bounces@imperial.ac.uk<mailto:gridpp-dirac-users-bounces@imperial.ac.uk> [mailto:gridpp-dirac-users-bounces@imperial.ac.uk] On Behalf Of WHITE Darren Sent: 04 September 2017 15:54 To: gridpp-dirac-users@imperial.ac.uk<mailto:gridpp-dirac-users@imperial.ac.uk> Subject: [Gridpp-Dirac-Users] Sites with git Hi there, I’m working with LSST folks in Southampton who have code that is updated quite regularly. A while ago, I tested to see whether git could clone at time of execution (I cannot remember which site this was at), and it was a success. However, RAL does not have git installed, as similar code fails at that site. I’ve asked this before and didn’t really get a full answer. Is there any way of knowing which sites have which software installed? It would save new users a lot of pain when trying to get up and running, or even just testing. Cheers Darren ------------------------------------------------------------------------- Dr Darren J. White Applications Developer EPCC, University of Edinburgh Email: d.white@epcc.ed.ac.uk<mailto:d.white@epcc.ed.ac.uk> Telephone: +44 (0) 131 6513415 Room 2411, James Clerk Maxwell Building, Kings Buildings, Mayfield Road, Edinburgh EH9 3JZ, United Kingdom ------------------------------------------------------------------------- -- _______________________________________________ Gridpp-Dirac-Users mailing list Gridpp-Dirac-Users@imperial.ac.uk<mailto:Gridpp-Dirac-Users@imperial.ac.uk> https://mailman.ic.ac.uk/mailman/listinfo/gridpp-dirac-users The University of Edinburgh is a charitable body, registered in Scotland, with registration number SC005336.
Hi Darren, We have a page (http://pprc.qmul.ac.uk/~lloyd/gridpp/votable.html) that shows which sites support which VOs that might be useful to new people. Cheers +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ Steve Lloyd Emeritus Professor, School of Physics and Astronomy Queen Mary University of London, Mile End Road, London E1 4NS, UK E-mail: s.l.lloyd@qmul.ac.uk +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
On 4 Sep 2017, at 16:52, WHITE Darren <d.white@epcc.ed.ac.uk> wrote:
Hi Catalin,
Even before this project got underway, we were warned to not do that for code that is in such active development. The time it takes for CVMFS updates to filter down can take far too much time for testing. I spent the last few days in Southampton last week and we were pushing changes to GitHub several times an hour that could instantly be available for testing. That’s simply not possible from what I understand/was told about CVMFS.
Also, the Lyon repo is not available at all sites for LSST folks, as I found out the hard way. The specified repo I was told after many days of failure is apparently /cvmfs/lsst.opensciencegrid.org, which we successfully use to set up EUPS, which we then, in turn, use to set up the many packages needed for this specific code (which are also not available on either the LYON or opensciencegrid repos and so packaged into a tarball on an SE). This is another issue for new folks trying to work on GridPP. There’s no real clarity of what CVMFS repos are available, what sites you can run at, what tools they include both at the site or on the repos, etc.
Cheers
Darren ------------------------------------------------------------------------- Dr Darren J. White Applications Developer EPCC, University of Edinburgh
Email: d.white@epcc.ed.ac.uk Telephone: +44 (0) 131 6513415
Room 2411, James Clerk Maxwell Building, Kings Buildings, Mayfield Road, Edinburgh EH9 3JZ, United Kingdom -------------------------------------------------------------------------
On 4 Sep 2017, at 16:38, catalin.condurache@stfc.ac.uk wrote:
Hi Darren,
Just a question: why you do not stick with the cvmfs repository? There is ‘lsst.in2p3.fr’ repository (maintained at Lyon and replicated by Stratum1 at RAL) where I’d assume all software should be provisioned from.
Is this code you’re talking about only interesting the UK sites? In this case you may want to have another (UK regional) ‘lsst’ CVMFS repository.
Anyway we at RAL have the policy to install a minimal distribution on our WNs (and git is not part of it), assuming the VOs are provisioning their software via CVMFS.
Regards, Catalin Condurache RAL Tier-1 Services
From: gridpp-dirac-users-bounces@imperial.ac.uk [mailto:gridpp-dirac-users-bounces@imperial.ac.uk] On Behalf Of WHITE Darren Sent: 04 September 2017 15:54 To: gridpp-dirac-users@imperial.ac.uk Subject: [Gridpp-Dirac-Users] Sites with git
Hi there,
I’m working with LSST folks in Southampton who have code that is updated quite regularly. A while ago, I tested to see whether git could clone at time of execution (I cannot remember which site this was at), and it was a success. However, RAL does not have git installed, as similar code fails at that site. I’ve asked this before and didn’t really get a full answer. Is there any way of knowing which sites have which software installed? It would save new users a lot of pain when trying to get up and running, or even just testing.
Cheers
Darren ------------------------------------------------------------------------- Dr Darren J. White Applications Developer EPCC, University of Edinburgh
Email: d.white@epcc.ed.ac.uk Telephone: +44 (0) 131 6513415
Room 2411, James Clerk Maxwell Building, Kings Buildings, Mayfield Road, Edinburgh EH9 3JZ, United Kingdom -------------------------------------------------------------------------
-- _______________________________________________ Gridpp-Dirac-Users mailing list Gridpp-Dirac-Users@imperial.ac.uk https://mailman.ic.ac.uk/mailman/listinfo/gridpp-dirac-users
The University of Edinburgh is a charitable body, registered in Scotland, with registration number SC005336. -- _______________________________________________ Gridpp-Dirac-Users mailing list Gridpp-Dirac-Users@imperial.ac.uk https://mailman.ic.ac.uk/mailman/listinfo/gridpp-dirac-users
Good morning Darren, 1) At WLCG level there used to be some time ago few repositories replicated every 10 mins (so-called fast replicas), so it can be done for LSST up to an extent (or at least one could give a try with a local repository). 2) Regarding LSST community, it is quite obvious that there is no consistency there with two repositories (opensciencegrid and in2p3). However these two repos are replicated by the Stratum-1 at RAL, so they should be accessible by any client with correct settings. If there is interest, we could work together and make lsst.in2p3.fr available via the CVMFS_CONFIG_REPOSITORY feature. 3) Another option I see is to add git tool to one of the existing lsst repos or to the grid.cern.ch repository (which should be available everywhere), but there might be other implications… Anyway at GridPP level we definitely could work to provide a consistent solution. I just didn’t know anyone involved in LSST from any UK site. But this could be the start. BTW are you attending the GridPP meeting in Lancaster? We could discuss more (or maybe finalise things) then. Regards, Catalin From: gridpp-dirac-users-bounces@imperial.ac.uk [mailto:gridpp-dirac-users-bounces@imperial.ac.uk] On Behalf Of WHITE Darren Sent: 04 September 2017 16:52 To: gridpp-dirac-users@imperial.ac.uk Subject: Re: [Gridpp-Dirac-Users] Sites with git Hi Catalin, Even before this project got underway, we were warned to not do that for code that is in such active development. The time it takes for CVMFS updates to filter down can take far too much time for testing. I spent the last few days in Southampton last week and we were pushing changes to GitHub several times an hour that could instantly be available for testing. That’s simply not possible from what I understand/was told about CVMFS. Also, the Lyon repo is not available at all sites for LSST folks, as I found out the hard way. The specified repo I was told after many days of failure is apparently /cvmfs/lsst.opensciencegrid.org<http://lsst.opensciencegrid.org>, which we successfully use to set up EUPS, which we then, in turn, use to set up the many packages needed for this specific code (which are also not available on either the LYON or opensciencegrid repos and so packaged into a tarball on an SE). This is another issue for new folks trying to work on GridPP. There’s no real clarity of what CVMFS repos are available, what sites you can run at, what tools they include both at the site or on the repos, etc. Cheers Darren ------------------------------------------------------------------------- Dr Darren J. White Applications Developer EPCC, University of Edinburgh Email: d.white@epcc.ed.ac.uk<mailto:d.white@epcc.ed.ac.uk> Telephone: +44 (0) 131 6513415 Room 2411, James Clerk Maxwell Building, Kings Buildings, Mayfield Road, Edinburgh EH9 3JZ, United Kingdom ------------------------------------------------------------------------- On 4 Sep 2017, at 16:38, catalin.condurache@stfc.ac.uk<mailto:catalin.condurache@stfc.ac.uk> wrote: Hi Darren, Just a question: why you do not stick with the cvmfs repository? There is ‘lsst.in2p3.fr<http://lsst.in2p3.fr/>’ repository (maintained at Lyon and replicated by Stratum1 at RAL) where I’d assume all software should be provisioned from. Is this code you’re talking about only interesting the UK sites? In this case you may want to have another (UK regional) ‘lsst’ CVMFS repository. Anyway we at RAL have the policy to install a minimal distribution on our WNs (and git is not part of it), assuming the VOs are provisioning their software via CVMFS. Regards, Catalin Condurache RAL Tier-1 Services From: gridpp-dirac-users-bounces@imperial.ac.uk<mailto:gridpp-dirac-users-bounces@imperial.ac.uk> [mailto:gridpp-dirac-users-bounces@imperial.ac.uk] On Behalf Of WHITE Darren Sent: 04 September 2017 15:54 To: gridpp-dirac-users@imperial.ac.uk<mailto:gridpp-dirac-users@imperial.ac.uk> Subject: [Gridpp-Dirac-Users] Sites with git Hi there, I’m working with LSST folks in Southampton who have code that is updated quite regularly. A while ago, I tested to see whether git could clone at time of execution (I cannot remember which site this was at), and it was a success. However, RAL does not have git installed, as similar code fails at that site. I’ve asked this before and didn’t really get a full answer. Is there any way of knowing which sites have which software installed? It would save new users a lot of pain when trying to get up and running, or even just testing. Cheers Darren ------------------------------------------------------------------------- Dr Darren J. White Applications Developer EPCC, University of Edinburgh Email: d.white@epcc.ed.ac.uk<mailto:d.white@epcc.ed.ac.uk> Telephone: +44 (0) 131 6513415 Room 2411, James Clerk Maxwell Building, Kings Buildings, Mayfield Road, Edinburgh EH9 3JZ, United Kingdom ------------------------------------------------------------------------- -- _______________________________________________ Gridpp-Dirac-Users mailing list Gridpp-Dirac-Users@imperial.ac.uk<mailto:Gridpp-Dirac-Users@imperial.ac.uk> https://mailman.ic.ac.uk/mailman/listinfo/gridpp-dirac-users
Hi everyone, Thanks for the list Steve. It’s a good place to start, but there isn’t a place for people within those VOs to easily find what repos/software are available, which is the bigger issue. Thanks for the reply Catalin. The options sound useful, but I’m not sure how much time I can put towards this. The work I’m doing is now coming to a close, and I’m moving away from the GridPP side of my LSST work. This project with Southampton was primarily a use case to see how useful GridPP will be for general use from within LSST. There is a new mailing list for LSST-UK folks to contact if they have issues, so it may be left to that from now on. I’m not sure who will be in charge of the setup of LSST-GridPP how you describe. For now, Southampton’s setup works as needed, so for the short term no work is needed now, but longer-term there is a need for a more detailed overview of how LSST can properly utilise GridPP and to give new users a better understanding of how GridPP actually works. I will discuss with my project manager about how this could be done at our next meeting. Thanks again for everybody’s help. Darren ------------------------------------------------------------------------- Dr Darren J. White Applications Developer EPCC, University of Edinburgh Email: d.white@epcc.ed.ac.uk<mailto:d.white@epcc.ed.ac.uk> Telephone: +44 (0) 131 6513415 Room 2411, James Clerk Maxwell Building, Kings Buildings, Mayfield Road, Edinburgh EH9 3JZ, United Kingdom ------------------------------------------------------------------------- On 5 Sep 2017, at 09:00, catalin.condurache@stfc.ac.uk<mailto:catalin.condurache@stfc.ac.uk> wrote: Good morning Darren, 1) At WLCG level there used to be some time ago few repositories replicated every 10 mins (so-called fast replicas), so it can be done for LSST up to an extent (or at least one could give a try with a local repository). 2) Regarding LSST community, it is quite obvious that there is no consistency there with two repositories (opensciencegrid and in2p3). However these two repos are replicated by the Stratum-1 at RAL, so they should be accessible by any client with correct settings. If there is interest, we could work together and make lsst.in2p3.fr<http://lsst.in2p3.fr/> available via the CVMFS_CONFIG_REPOSITORY feature. 3) Another option I see is to add git tool to one of the existing lsst repos or to the grid.cern.ch<http://grid.cern.ch/> repository (which should be available everywhere), but there might be other implications… Anyway at GridPP level we definitely could work to provide a consistent solution. I just didn’t know anyone involved in LSST from any UK site. But this could be the start. BTW are you attending the GridPP meeting in Lancaster? We could discuss more (or maybe finalise things) then. Regards, Catalin From: gridpp-dirac-users-bounces@imperial.ac.uk<mailto:gridpp-dirac-users-bounces@imperial.ac.uk> [mailto:gridpp-dirac-users-bounces@imperial.ac.uk] On Behalf Of WHITE Darren Sent: 04 September 2017 16:52 To: gridpp-dirac-users@imperial.ac.uk<mailto:gridpp-dirac-users@imperial.ac.uk> Subject: Re: [Gridpp-Dirac-Users] Sites with git Hi Catalin, Even before this project got underway, we were warned to not do that for code that is in such active development. The time it takes for CVMFS updates to filter down can take far too much time for testing. I spent the last few days in Southampton last week and we were pushing changes to GitHub several times an hour that could instantly be available for testing. That’s simply not possible from what I understand/was told about CVMFS. Also, the Lyon repo is not available at all sites for LSST folks, as I found out the hard way. The specified repo I was told after many days of failure is apparently /cvmfs/lsst.opensciencegrid.org<http://lsst.opensciencegrid.org/>, which we successfully use to set up EUPS, which we then, in turn, use to set up the many packages needed for this specific code (which are also not available on either the LYON or opensciencegrid repos and so packaged into a tarball on an SE). This is another issue for new folks trying to work on GridPP. There’s no real clarity of what CVMFS repos are available, what sites you can run at, what tools they include both at the site or on the repos, etc. Cheers Darren ------------------------------------------------------------------------- Dr Darren J. White Applications Developer EPCC, University of Edinburgh Email: d.white@epcc.ed.ac.uk<mailto:d.white@epcc.ed.ac.uk> Telephone: +44 (0) 131 6513415 Room 2411, James Clerk Maxwell Building, Kings Buildings, Mayfield Road, Edinburgh EH9 3JZ, United Kingdom ------------------------------------------------------------------------- On 4 Sep 2017, at 16:38, catalin.condurache@stfc.ac.uk<mailto:catalin.condurache@stfc.ac.uk> wrote: Hi Darren, Just a question: why you do not stick with the cvmfs repository? There is ‘lsst.in2p3.fr<http://lsst.in2p3.fr/>’ repository (maintained at Lyon and replicated by Stratum1 at RAL) where I’d assume all software should be provisioned from. Is this code you’re talking about only interesting the UK sites? In this case you may want to have another (UK regional) ‘lsst’ CVMFS repository. Anyway we at RAL have the policy to install a minimal distribution on our WNs (and git is not part of it), assuming the VOs are provisioning their software via CVMFS. Regards, Catalin Condurache RAL Tier-1 Services From: gridpp-dirac-users-bounces@imperial.ac.uk<mailto:gridpp-dirac-users-bounces@imperial.ac.uk> [mailto:gridpp-dirac-users-bounces@imperial.ac.uk] On Behalf Of WHITE Darren Sent: 04 September 2017 15:54 To: gridpp-dirac-users@imperial.ac.uk<mailto:gridpp-dirac-users@imperial.ac.uk> Subject: [Gridpp-Dirac-Users] Sites with git Hi there, I’m working with LSST folks in Southampton who have code that is updated quite regularly. A while ago, I tested to see whether git could clone at time of execution (I cannot remember which site this was at), and it was a success. However, RAL does not have git installed, as similar code fails at that site. I’ve asked this before and didn’t really get a full answer. Is there any way of knowing which sites have which software installed? It would save new users a lot of pain when trying to get up and running, or even just testing. Cheers Darren ------------------------------------------------------------------------- Dr Darren J. White Applications Developer EPCC, University of Edinburgh Email: d.white@epcc.ed.ac.uk<mailto:d.white@epcc.ed.ac.uk> Telephone: +44 (0) 131 6513415 Room 2411, James Clerk Maxwell Building, Kings Buildings, Mayfield Road, Edinburgh EH9 3JZ, United Kingdom ------------------------------------------------------------------------- -- _______________________________________________ Gridpp-Dirac-Users mailing list Gridpp-Dirac-Users@imperial.ac.uk<mailto:Gridpp-Dirac-Users@imperial.ac.uk> https://mailman.ic.ac.uk/mailman/listinfo/gridpp-dirac-users -- _______________________________________________ Gridpp-Dirac-Users mailing list Gridpp-Dirac-Users@imperial.ac.uk<mailto:Gridpp-Dirac-Users@imperial.ac.uk> https://mailman.ic.ac.uk/mailman/listinfo/gridpp-dirac-users The University of Edinburgh is a charitable body, registered in Scotland, with registration number SC005336.
Hi Darren, it is cumbersome and difficult to maintain to list any type of software a shared site might have installed. Sites on the grid typically support a number of VOs. This is why VOs tend to package what they need in their repositories (CVMFS, git or other). Software distribution is not only a gridpp problem. GridPP offers the means to install and distribute software, in particular cvmfs, but the experiments should decide what they want to distribute. LHC experiments for example have a team of people managing the core of their software upon which the users build their code. Users have then just to maintain their own limited number of files. In UKLSST case it is true that the code development made CVMFS unusable for the code that changes often. However it could be used for established packages that you are not sure you may find at every site and don't change. The UK had a dedicate subdir in the OSG CVMFS repository for this which we created when Joe Zuntz was the testing user. cheers alessandra On 05/09/2017 10:16, WHITE Darren wrote:
Hi everyone,
Thanks for the list Steve. It’s a good place to start, but there isn’t a place for people within those VOs to easily find what repos/software are available, which is the bigger issue.
Thanks for the reply Catalin. The options sound useful, but I’m not sure how much time I can put towards this. The work I’m doing is now coming to a close, and I’m moving away from the GridPP side of my LSST work. This project with Southampton was primarily a use case to see how useful GridPP will be for general use from within LSST. There is a new mailing list for LSST-UK folks to contact if they have issues, so it may be left to that from now on. I’m not sure who will be in charge of the setup of LSST-GridPP how you describe. For now, Southampton’s setup works as needed, so for the short term no work is needed now, but longer-term there is a need for a more detailed overview of how LSST can properly utilise GridPP and to give new users a better understanding of how GridPP actually works. I will discuss with my project manager about how this could be done at our next meeting.
Thanks again for everybody’s help.
Darren ------------------------------------------------------------------------- Dr Darren J. White Applications Developer EPCC, University of Edinburgh
Email: d.white@epcc.ed.ac.uk <mailto:d.white@epcc.ed.ac.uk> Telephone: +44 (0) 131 6513415
Room 2411, James Clerk Maxwell Building, Kings Buildings, Mayfield Road, Edinburgh EH9 3JZ, United Kingdom -------------------------------------------------------------------------
On 5 Sep 2017, at 09:00, catalin.condurache@stfc.ac.uk <mailto:catalin.condurache@stfc.ac.uk> wrote:
Good morning Darren, 1) At WLCG level there used to be some time ago few repositories replicated every 10 mins (so-called fast replicas), so it can be done for LSST up to an extent (or at least one could give a try with a local repository). 2) Regarding LSST community, it is quite obvious that there is no consistency there with two repositories (opensciencegrid and in2p3). However these two repos are replicated by the Stratum-1 at RAL, so they should be accessible by any client with correct settings. If there is interest, we could work together and makelsst.in2p3.fr <http://lsst.in2p3.fr/>available via the CVMFS_CONFIG_REPOSITORY feature. 3) Another option I see is to add git tool to one of the existing lsst repos or to thegrid.cern.ch <http://grid.cern.ch/>repository (which should be available everywhere), but there might be other implications… Anyway at GridPP level we definitely could work to provide a consistent solution. I just didn’t know anyone involved in LSST from any UK site. But this could be the start. BTW are you attending the GridPP meeting in Lancaster? We could discuss more (or maybe finalise things) then. Regards, Catalin *From:*gridpp-dirac-users-bounces@imperial.ac.uk <mailto:gridpp-dirac-users-bounces@imperial.ac.uk> [mailto:gridpp-dirac-users-bounces@imperial.ac.uk]*On Behalf Of*WHITE Darren *Sent:*04 September 2017 16:52 *To:*gridpp-dirac-users@imperial.ac.uk <mailto:gridpp-dirac-users@imperial.ac.uk> *Subject:*Re: [Gridpp-Dirac-Users] Sites with git Hi Catalin, Even before this project got underway, we were warned to not do that for code that is in such active development. The time it takes for CVMFS updates to filter down can take far too much time for testing. I spent the last few days in Southampton last week and we were pushing changes to GitHub several times an hour that could instantly be available for testing. That’s simply not possible from what I understand/was told about CVMFS. Also, the Lyon repo is not available at all sites for LSST folks, as I found out the hard way. The specified repo I was told after many days of failure is apparently /cvmfs/lsst.opensciencegrid.org <http://lsst.opensciencegrid.org/>, which we successfully use to set up EUPS, which we then, in turn, use to set up the many packages needed for this specific code (which are also not available on either the LYON or opensciencegrid repos and so packaged into a tarball on an SE). This is another issue for new folks trying to work on GridPP. There’s no real clarity of what CVMFS repos are available, what sites you can run at, what tools they include both at the site or on the repos, etc. Cheers Darren ------------------------------------------------------------------------- Dr Darren J. White Applications Developer EPCC, University of Edinburgh
Email: d.white@epcc.ed.ac.uk <mailto:d.white@epcc.ed.ac.uk> Telephone: +44 (0) 131 6513415
Room 2411, James Clerk Maxwell Building, Kings Buildings, Mayfield Road, Edinburgh EH9 3JZ, United Kingdom -------------------------------------------------------------------------
On 4 Sep 2017, at 16:38,catalin.condurache@stfc.ac.uk <mailto:catalin.condurache@stfc.ac.uk>wrote: Hi Darren, Just a question: why you do not stick with the cvmfs repository? There is ‘lsst.in2p3.fr <http://lsst.in2p3.fr/>’ repository (maintained at Lyon and replicated by Stratum1 at RAL) where I’d assume all software should be provisioned from. Is this code you’re talking about only interesting the UK sites? In this case you may want to have another (UK regional) ‘lsst’ CVMFS repository. Anyway we at RAL have the policy to install a minimal distribution on our WNs (and git is not part of it), assuming the VOs are provisioning their software via CVMFS. Regards, Catalin Condurache RAL Tier-1 Services *From:*gridpp-dirac-users-bounces@imperial.ac.uk <mailto:gridpp-dirac-users-bounces@imperial.ac.uk>[mailto:gridpp-dirac-users-bounces@imperial.ac.uk]*On Behalf Of*WHITE Darren *Sent:*04 September 2017 15:54 *To:*gridpp-dirac-users@imperial.ac.uk <mailto:gridpp-dirac-users@imperial.ac.uk> *Subject:*[Gridpp-Dirac-Users] Sites with git Hi there, I’m working with LSST folks in Southampton who have code that is updated quite regularly. A while ago, I tested to see whether git could clone at time of execution (I cannot remember which site this was at), and it was a success. However, RAL does not have git installed, as similar code fails at that site. I’ve asked this before and didn’t really get a full answer. Is there any way of knowing which sites have which software installed? It would save new users a lot of pain when trying to get up and running, or even just testing. Cheers Darren ------------------------------------------------------------------------- Dr Darren J. White Applications Developer EPCC, University of Edinburgh
Email: d.white@epcc.ed.ac.uk <mailto:d.white@epcc.ed.ac.uk> Telephone: +44 (0) 131 6513415
Room 2411, James Clerk Maxwell Building, Kings Buildings, Mayfield Road, Edinburgh EH9 3JZ, United Kingdom ------------------------------------------------------------------------- -- _______________________________________________ Gridpp-Dirac-Users mailing list Gridpp-Dirac-Users@imperial.ac.uk <mailto:Gridpp-Dirac-Users@imperial.ac.uk> https://mailman.ic.ac.uk/mailman/listinfo/gridpp-dirac-users
-- _______________________________________________ Gridpp-Dirac-Users mailing list Gridpp-Dirac-Users@imperial.ac.uk <mailto:Gridpp-Dirac-Users@imperial.ac.uk> https://mailman.ic.ac.uk/mailman/listinfo/gridpp-dirac-users
The University of Edinburgh is a charitable body, registered in Scotland, with registration number SC005336.
-- Respect is a rational process. \\// Fatti non foste a viver come bruti, ma per seguir virtute e canoscenza(Dante) For Ur-Fascism, disagreement is treason. (U. Eco) But but but her emails... covfefe!
Hi everyone, Thanks for the discussion. I’ve learned quite a bit about GridPP over the past couple of days, much of it useful. I understand that it is not the place for GridPP support to list every single software package. The changes made for each individual experiment make that impossible. But I think it would be useful to have some list of what cvmfs repositories are available for each project, so people can then go and check what is available in those repos. That was the biggest issue I came to first. I wrote my software linking to the Lyon cvmfs on my local machine, based on the suggestions of other people within GridPP at the time, only to find that it wasn’t mounted as default on the sites, but that the OSG version was. Now, in the past couple of months, the Lyon cvmfs is being mounted, but only at certain sites. There’s no way of being able to find that out unless you are in that small group of people 'in the know’ or by emailing support (which I’m sure you guys would prefer to not be the case for each new user from a fledgling project). From my short time working with LSST it seems like it isn’t a singular organisation like some of the other projects using GridPP, or at least not mature enough for that to be the case, and it’s made up primarily of small subgroups tackling individual problems. Those groups (like Southampton) are happy to use GridPP as long as they have some idea of what to expect and what they have to work with (in terms of what repos are available, how they can bring across their own software or install additional dependencies, for example). Right now, that’s not quite the case from our recent experience. There’s no documentation on what cvmfs repos are available, there’s no documentation on how to package your own software onto SEs manually (like we had to do since the repos didn’t have the right versions of software we needed), etc. Decent, simple beginners guides to these things would go a long way to making the experience smoother. If I am able to contribute more time to this within my allotted time on LSST, I’m happy to help, but I feel it is something that does need to be done if there is to be better support for new/small/inexperienced groups to use GridPP. Thanks again, Darren ------------------------------------------------------------------------- Dr Darren J. White Applications Developer EPCC, University of Edinburgh Email: d.white@epcc.ed.ac.uk<mailto:d.white@epcc.ed.ac.uk> Telephone: +44 (0) 131 6513415 Room 2411, James Clerk Maxwell Building, Kings Buildings, Mayfield Road, Edinburgh EH9 3JZ, United Kingdom ------------------------------------------------------------------------- On 6 Sep 2017, at 11:56, Alessandra Forti <Alessandra.Forti@cern.ch<mailto:Alessandra.Forti@cern.ch>> wrote: Hi Darren, it is cumbersome and difficult to maintain to list any type of software a shared site might have installed. Sites on the grid typically support a number of VOs. This is why VOs tend to package what they need in their repositories (CVMFS, git or other). Software distribution is not only a gridpp problem. GridPP offers the means to install and distribute software, in particular cvmfs, but the experiments should decide what they want to distribute. LHC experiments for example have a team of people managing the core of their software upon which the users build their code. Users have then just to maintain their own limited number of files. In UKLSST case it is true that the code development made CVMFS unusable for the code that changes often. However it could be used for established packages that you are not sure you may find at every site and don't change. The UK had a dedicate subdir in the OSG CVMFS repository for this which we created when Joe Zuntz was the testing user. cheers alessandra On 05/09/2017 10:16, WHITE Darren wrote: Hi everyone, Thanks for the list Steve. It’s a good place to start, but there isn’t a place for people within those VOs to easily find what repos/software are available, which is the bigger issue. Thanks for the reply Catalin. The options sound useful, but I’m not sure how much time I can put towards this. The work I’m doing is now coming to a close, and I’m moving away from the GridPP side of my LSST work. This project with Southampton was primarily a use case to see how useful GridPP will be for general use from within LSST. There is a new mailing list for LSST-UK folks to contact if they have issues, so it may be left to that from now on. I’m not sure who will be in charge of the setup of LSST-GridPP how you describe. For now, Southampton’s setup works as needed, so for the short term no work is needed now, but longer-term there is a need for a more detailed overview of how LSST can properly utilise GridPP and to give new users a better understanding of how GridPP actually works. I will discuss with my project manager about how this could be done at our next meeting. Thanks again for everybody’s help. Darren ------------------------------------------------------------------------- Dr Darren J. White Applications Developer EPCC, University of Edinburgh Email: d.white@epcc.ed.ac.uk<mailto:d.white@epcc.ed.ac.uk> Telephone: +44 (0) 131 6513415 Room 2411, James Clerk Maxwell Building, Kings Buildings, Mayfield Road, Edinburgh EH9 3JZ, United Kingdom ------------------------------------------------------------------------- On 5 Sep 2017, at 09:00, catalin.condurache@stfc.ac.uk<mailto:catalin.condurache@stfc.ac.uk> wrote: Good morning Darren, 1) At WLCG level there used to be some time ago few repositories replicated every 10 mins (so-called fast replicas), so it can be done for LSST up to an extent (or at least one could give a try with a local repository). 2) Regarding LSST community, it is quite obvious that there is no consistency there with two repositories (opensciencegrid and in2p3). However these two repos are replicated by the Stratum-1 at RAL, so they should be accessible by any client with correct settings. If there is interest, we could work together and make lsst.in2p3.fr<http://lsst.in2p3.fr/> available via the CVMFS_CONFIG_REPOSITORY feature. 3) Another option I see is to add git tool to one of the existing lsst repos or to the grid.cern.ch<http://grid.cern.ch/> repository (which should be available everywhere), but there might be other implications… Anyway at GridPP level we definitely could work to provide a consistent solution. I just didn’t know anyone involved in LSST from any UK site. But this could be the start. BTW are you attending the GridPP meeting in Lancaster? We could discuss more (or maybe finalise things) then. Regards, Catalin From: gridpp-dirac-users-bounces@imperial.ac.uk<mailto:gridpp-dirac-users-bounces@imperial.ac.uk> [mailto:gridpp-dirac-users-bounces@imperial.ac.uk] On Behalf Of WHITE Darren Sent: 04 September 2017 16:52 To: gridpp-dirac-users@imperial.ac.uk<mailto:gridpp-dirac-users@imperial.ac.uk> Subject: Re: [Gridpp-Dirac-Users] Sites with git Hi Catalin, Even before this project got underway, we were warned to not do that for code that is in such active development. The time it takes for CVMFS updates to filter down can take far too much time for testing. I spent the last few days in Southampton last week and we were pushing changes to GitHub several times an hour that could instantly be available for testing. That’s simply not possible from what I understand/was told about CVMFS. Also, the Lyon repo is not available at all sites for LSST folks, as I found out the hard way. The specified repo I was told after many days of failure is apparently /cvmfs/lsst.opensciencegrid.org<http://lsst.opensciencegrid.org/>, which we successfully use to set up EUPS, which we then, in turn, use to set up the many packages needed for this specific code (which are also not available on either the LYON or opensciencegrid repos and so packaged into a tarball on an SE). This is another issue for new folks trying to work on GridPP. There’s no real clarity of what CVMFS repos are available, what sites you can run at, what tools they include both at the site or on the repos, etc. Cheers Darren ------------------------------------------------------------------------- Dr Darren J. White Applications Developer EPCC, University of Edinburgh Email: d.white@epcc.ed.ac.uk<mailto:d.white@epcc.ed.ac.uk> Telephone: +44 (0) 131 6513415 Room 2411, James Clerk Maxwell Building, Kings Buildings, Mayfield Road, Edinburgh EH9 3JZ, United Kingdom ------------------------------------------------------------------------- On 4 Sep 2017, at 16:38, catalin.condurache@stfc.ac.uk<mailto:catalin.condurache@stfc.ac.uk> wrote: Hi Darren, Just a question: why you do not stick with the cvmfs repository? There is ‘lsst.in2p3.fr<http://lsst.in2p3.fr/>’ repository (maintained at Lyon and replicated by Stratum1 at RAL) where I’d assume all software should be provisioned from. Is this code you’re talking about only interesting the UK sites? In this case you may want to have another (UK regional) ‘lsst’ CVMFS repository. Anyway we at RAL have the policy to install a minimal distribution on our WNs (and git is not part of it), assuming the VOs are provisioning their software via CVMFS. Regards, Catalin Condurache RAL Tier-1 Services From: gridpp-dirac-users-bounces@imperial.ac.uk<mailto:gridpp-dirac-users-bounces@imperial.ac.uk> [mailto:gridpp-dirac-users-bounces@imperial.ac.uk] On Behalf Of WHITE Darren Sent: 04 September 2017 15:54 To: gridpp-dirac-users@imperial.ac.uk<mailto:gridpp-dirac-users@imperial.ac.uk> Subject: [Gridpp-Dirac-Users] Sites with git Hi there, I’m working with LSST folks in Southampton who have code that is updated quite regularly. A while ago, I tested to see whether git could clone at time of execution (I cannot remember which site this was at), and it was a success. However, RAL does not have git installed, as similar code fails at that site. I’ve asked this before and didn’t really get a full answer. Is there any way of knowing which sites have which software installed? It would save new users a lot of pain when trying to get up and running, or even just testing. Cheers Darren ------------------------------------------------------------------------- Dr Darren J. White Applications Developer EPCC, University of Edinburgh Email: d.white@epcc.ed.ac.uk<mailto:d.white@epcc.ed.ac.uk> Telephone: +44 (0) 131 6513415 Room 2411, James Clerk Maxwell Building, Kings Buildings, Mayfield Road, Edinburgh EH9 3JZ, United Kingdom ------------------------------------------------------------------------- -- _______________________________________________ Gridpp-Dirac-Users mailing list Gridpp-Dirac-Users@imperial.ac.uk<mailto:Gridpp-Dirac-Users@imperial.ac.uk> https://mailman.ic.ac.uk/mailman/listinfo/gridpp-dirac-users -- _______________________________________________ Gridpp-Dirac-Users mailing list Gridpp-Dirac-Users@imperial.ac.uk<mailto:Gridpp-Dirac-Users@imperial.ac.uk> https://mailman.ic.ac.uk/mailman/listinfo/gridpp-dirac-users The University of Edinburgh is a charitable body, registered in Scotland, with registration number SC005336. -- Respect is a rational process. \\// Fatti non foste a viver come bruti, ma per seguir virtute e canoscenza(Dante) For Ur-Fascism, disagreement is treason. (U. Eco) But but but her emails... covfefe! -- _______________________________________________ Gridpp-Dirac-Users mailing list Gridpp-Dirac-Users@imperial.ac.uk<mailto:Gridpp-Dirac-Users@imperial.ac.uk> https://mailman.ic.ac.uk/mailman/listinfo/gridpp-dirac-users The University of Edinburgh is a charitable body, registered in Scotland, with registration number SC005336.
Hi Darren, thank you for the feedback. We (Imperial) just got half a post for 'small experiment support', so I will forward your email to that person. I would also encourage other users to please let us ('Gridpp') know if something doesn't work as expected, so we can fix it. You can use this list, we'll redirect it if necessary. Regards, Daniela On 6 September 2017 at 12:53, WHITE Darren <d.white@epcc.ed.ac.uk> wrote:
Hi everyone,
Thanks for the discussion. I’ve learned quite a bit about GridPP over the past couple of days, much of it useful. I understand that it is not the place for GridPP support to list every single software package. The changes made for each individual experiment make that impossible. But I think it would be useful to have some list of what cvmfs repositories are available for each project, so people can then go and check what is available in those repos.
That was the biggest issue I came to first. I wrote my software linking to the Lyon cvmfs on my local machine, based on the suggestions of other people within GridPP at the time, only to find that it wasn’t mounted as default on the sites, but that the OSG version was. Now, in the past couple of months, the Lyon cvmfs is being mounted, but only at certain sites. There’s no way of being able to find that out unless you are in that small group of people 'in the know’ or by emailing support (which I’m sure you guys would prefer to not be the case for each new user from a fledgling project).
From my short time working with LSST it seems like it isn’t a singular organisation like some of the other projects using GridPP, or at least not mature enough for that to be the case, and it’s made up primarily of small subgroups tackling individual problems. Those groups (like Southampton) are happy to use GridPP as long as they have some idea of what to expect and what they have to work with (in terms of what repos are available, how they can bring across their own software or install additional dependencies, for example). Right now, that’s not quite the case from our recent experience. There’s no documentation on what cvmfs repos are available, there’s no documentation on how to package your own software onto SEs manually (like we had to do since the repos didn’t have the right versions of software we needed), etc. Decent, simple beginners guides to these things would go a long way to making the experience smoother. If I am able to contribute more time to this within my allotted time on LSST, I’m happy to help, but I feel it is something that does need to be done if there is to be better support for new/small/inexperienced groups to use GridPP.
Thanks again,
Darren
------------------------------------------------------------------------- Dr Darren J. White Applications Developer EPCC, University of Edinburgh
Email: d.white@epcc.ed.ac.uk Telephone: +44 (0) 131 6513415
Room 2411, James Clerk Maxwell Building, Kings Buildings, Mayfield Road, Edinburgh EH9 3JZ, United Kingdom -------------------------------------------------------------------------
On 6 Sep 2017, at 11:56, Alessandra Forti <Alessandra.Forti@cern.ch> wrote:
Hi Darren,
it is cumbersome and difficult to maintain to list any type of software a shared site might have installed. Sites on the grid typically support a number of VOs. This is why VOs tend to package what they need in their repositories (CVMFS, git or other).
Software distribution is not only a gridpp problem. GridPP offers the means to install and distribute software, in particular cvmfs, but the experiments should decide what they want to distribute. LHC experiments for example have a team of people managing the core of their software upon which the users build their code. Users have then just to maintain their own limited number of files.
In UKLSST case it is true that the code development made CVMFS unusable for the code that changes often. However it could be used for established packages that you are not sure you may find at every site and don't change. The UK had a dedicate subdir in the OSG CVMFS repository for this which we created when Joe Zuntz was the testing user.
cheers alessandra
On 05/09/2017 10:16, WHITE Darren wrote:
Hi everyone,
Thanks for the list Steve. It’s a good place to start, but there isn’t a place for people within those VOs to easily find what repos/software are available, which is the bigger issue.
Thanks for the reply Catalin. The options sound useful, but I’m not sure how much time I can put towards this. The work I’m doing is now coming to a close, and I’m moving away from the GridPP side of my LSST work. This project with Southampton was primarily a use case to see how useful GridPP will be for general use from within LSST. There is a new mailing list for LSST-UK folks to contact if they have issues, so it may be left to that from now on. I’m not sure who will be in charge of the setup of LSST-GridPP how you describe. For now, Southampton’s setup works as needed, so for the short term no work is needed now, but longer-term there is a need for a more detailed overview of how LSST can properly utilise GridPP and to give new users a better understanding of how GridPP actually works. I will discuss with my project manager about how this could be done at our next meeting.
Thanks again for everybody’s help.
Darren ------------------------------------------------------------------------- Dr Darren J. White Applications Developer EPCC, University of Edinburgh
Email: d.white@epcc.ed.ac.uk Telephone: +44 (0) 131 6513415
Room 2411, James Clerk Maxwell Building, Kings Buildings, Mayfield Road, Edinburgh EH9 3JZ, United Kingdom -------------------------------------------------------------------------
On 5 Sep 2017, at 09:00, catalin.condurache@stfc.ac.uk wrote:
Good morning Darren,
1) At WLCG level there used to be some time ago few repositories replicated every 10 mins (so-called fast replicas), so it can be done for LSST up to an extent (or at least one could give a try with a local repository).
2) Regarding LSST community, it is quite obvious that there is no consistency there with two repositories (opensciencegrid and in2p3). However these two repos are replicated by the Stratum-1 at RAL, so they should be accessible by any client with correct settings. If there is interest, we could work together and make lsst.in2p3.fr available via the CVMFS_CONFIG_REPOSITORY feature.
3) Another option I see is to add git tool to one of the existing lsst repos or to the grid.cern.ch repository (which should be available everywhere), but there might be other implications…
Anyway at GridPP level we definitely could work to provide a consistent solution. I just didn’t know anyone involved in LSST from any UK site. But this could be the start.
BTW are you attending the GridPP meeting in Lancaster? We could discuss more (or maybe finalise things) then.
Regards, Catalin
From: gridpp-dirac-users-bounces@imperial.ac.uk [mailto:gridpp-dirac-users-bounces@imperial.ac.uk] On Behalf Of WHITE Darren Sent: 04 September 2017 16:52 To: gridpp-dirac-users@imperial.ac.uk Subject: Re: [Gridpp-Dirac-Users] Sites with git
Hi Catalin,
Even before this project got underway, we were warned to not do that for code that is in such active development. The time it takes for CVMFS updates to filter down can take far too much time for testing. I spent the last few days in Southampton last week and we were pushing changes to GitHub several times an hour that could instantly be available for testing. That’s simply not possible from what I understand/was told about CVMFS.
Also, the Lyon repo is not available at all sites for LSST folks, as I found out the hard way. The specified repo I was told after many days of failure is apparently /cvmfs/lsst.opensciencegrid.org, which we successfully use to set up EUPS, which we then, in turn, use to set up the many packages needed for this specific code (which are also not available on either the LYON or opensciencegrid repos and so packaged into a tarball on an SE). This is another issue for new folks trying to work on GridPP. There’s no real clarity of what CVMFS repos are available, what sites you can run at, what tools they include both at the site or on the repos, etc.
Cheers
Darren ------------------------------------------------------------------------- Dr Darren J. White Applications Developer EPCC, University of Edinburgh
Email: d.white@epcc.ed.ac.uk Telephone: +44 (0) 131 6513415
Room 2411, James Clerk Maxwell Building, Kings Buildings, Mayfield Road, Edinburgh EH9 3JZ, United Kingdom -------------------------------------------------------------------------
On 4 Sep 2017, at 16:38, catalin.condurache@stfc.ac.uk wrote:
Hi Darren,
Just a question: why you do not stick with the cvmfs repository? There is ‘lsst.in2p3.fr’ repository (maintained at Lyon and replicated by Stratum1 at RAL) where I’d assume all software should be provisioned from.
Is this code you’re talking about only interesting the UK sites? In this case you may want to have another (UK regional) ‘lsst’ CVMFS repository.
Anyway we at RAL have the policy to install a minimal distribution on our WNs (and git is not part of it), assuming the VOs are provisioning their software via CVMFS.
Regards, Catalin Condurache RAL Tier-1 Services
From: gridpp-dirac-users-bounces@imperial.ac.uk [mailto:gridpp-dirac-users-bounces@imperial.ac.uk] On Behalf Of WHITE Darren Sent: 04 September 2017 15:54 To: gridpp-dirac-users@imperial.ac.uk Subject: [Gridpp-Dirac-Users] Sites with git
Hi there,
I’m working with LSST folks in Southampton who have code that is updated quite regularly. A while ago, I tested to see whether git could clone at time of execution (I cannot remember which site this was at), and it was a success. However, RAL does not have git installed, as similar code fails at that site. I’ve asked this before and didn’t really get a full answer. Is there any way of knowing which sites have which software installed? It would save new users a lot of pain when trying to get up and running, or even just testing.
Cheers
Darren ------------------------------------------------------------------------- Dr Darren J. White Applications Developer EPCC, University of Edinburgh
Email: d.white@epcc.ed.ac.uk Telephone: +44 (0) 131 6513415
Room 2411, James Clerk Maxwell Building, Kings Buildings, Mayfield Road, Edinburgh EH9 3JZ, United Kingdom -------------------------------------------------------------------------
-- _______________________________________________ Gridpp-Dirac-Users mailing list Gridpp-Dirac-Users@imperial.ac.uk https://mailman.ic.ac.uk/mailman/listinfo/gridpp-dirac-users
-- _______________________________________________ Gridpp-Dirac-Users mailing list Gridpp-Dirac-Users@imperial.ac.uk https://mailman.ic.ac.uk/mailman/listinfo/gridpp-dirac-users
The University of Edinburgh is a charitable body, registered in Scotland, with registration number SC005336.
-- Respect is a rational process. \\// Fatti non foste a viver come bruti, ma per seguir virtute e canoscenza(Dante) For Ur-Fascism, disagreement is treason. (U. Eco) But but but her emails... covfefe!
-- _______________________________________________ Gridpp-Dirac-Users mailing list Gridpp-Dirac-Users@imperial.ac.uk https://mailman.ic.ac.uk/mailman/listinfo/gridpp-dirac-users
The University of Edinburgh is a charitable body, registered in Scotland, with registration number SC005336.
-- _______________________________________________ Gridpp-Dirac-Users mailing list Gridpp-Dirac-Users@imperial.ac.uk https://mailman.ic.ac.uk/mailman/listinfo/gridpp-dirac-users
-- Sent from the pit of despair ----------------------------------------------------------- daniela.bauer@imperial.ac.uk HEP Group/Physics Dep Imperial College London, SW7 2BW Tel: +44-(0)20-75947810 http://www.hep.ph.ic.ac.uk/~dbauer/
Hi Darren, we have an LSST_UK wiki page for the moment is mostly for sites but we can add a User section or we can start another wiki page for users and add the information there. https://www.gridpp.ac.uk/wiki/LSST_UK cheers alessandra On 06/09/2017 12:53, WHITE Darren wrote:
Hi everyone,
Thanks for the discussion. I’ve learned quite a bit about GridPP over the past couple of days, much of it useful. I understand that it is not the place for GridPP support to list every single software package. The changes made for each individual experiment make that impossible. But I think it would be useful to have some list of what cvmfs repositories are available for each project, so people can then go and check what is available in those repos.
That was the biggest issue I came to first. I wrote my software linking to the Lyon cvmfs on my local machine, based on the suggestions of other people within GridPP at the time, only to find that it wasn’t mounted as default on the sites, but that the OSG version was. Now, in the past couple of months, the Lyon cvmfs is being mounted, but only at certain sites. There’s no way of being able to find that out unless you are in that small group of people 'in the know’ or by emailing support (which I’m sure you guys would prefer to not be the case for each new user from a fledgling project).
From my short time working with LSST it seems like it isn’t a singular organisation like some of the other projects using GridPP, or at least not mature enough for that to be the case, and it’s made up primarily of small subgroups tackling individual problems. Those groups (like Southampton) are happy to use GridPP as long as they have some idea of what to expect and what they have to work with (in terms of what repos are available, how they can bring across their own software or install additional dependencies, for example). Right now, that’s not quite the case from our recent experience. There’s no documentation on what cvmfs repos are available, there’s no documentation on how to package your own software onto SEs manually (like we had to do since the repos didn’t have the right versions of software we needed), etc. Decent, simple beginners guides to these things would go a long way to making the experience smoother. If I am able to contribute more time to this within my allotted time on LSST, I’m happy to help, but I feel it is something that does need to be done if there is to be better support for new/small/inexperienced groups to use GridPP.
Thanks again,
Darren
------------------------------------------------------------------------- Dr Darren J. White Applications Developer EPCC, University of Edinburgh
Email: d.white@epcc.ed.ac.uk <mailto:d.white@epcc.ed.ac.uk> Telephone: +44 (0) 131 6513415
Room 2411, James Clerk Maxwell Building, Kings Buildings, Mayfield Road, Edinburgh EH9 3JZ, United Kingdom -------------------------------------------------------------------------
On 6 Sep 2017, at 11:56, Alessandra Forti <Alessandra.Forti@cern.ch <mailto:Alessandra.Forti@cern.ch>> wrote:
Hi Darren,
it is cumbersome and difficult to maintain to list any type of software a shared site might have installed. Sites on the grid typically support a number of VOs. This is why VOs tend to package what they need in their repositories (CVMFS, git or other).
Software distribution is not only a gridpp problem. GridPP offers the means to install and distribute software, in particular cvmfs, but the experiments should decide what they want to distribute. LHC experiments for example have a team of people managing the core of their software upon which the users build their code. Users have then just to maintain their own limited number of files.
In UKLSST case it is true that the code development made CVMFS unusable for the code that changes often. However it could be used for established packages that you are not sure you may find at every site and don't change. The UK had a dedicate subdir in the OSG CVMFS repository for this which we created when Joe Zuntz was the testing user.
cheers alessandra
On 05/09/2017 10:16, WHITE Darren wrote:
Hi everyone,
Thanks for the list Steve. It’s a good place to start, but there isn’t a place for people within those VOs to easily find what repos/software are available, which is the bigger issue.
Thanks for the reply Catalin. The options sound useful, but I’m not sure how much time I can put towards this. The work I’m doing is now coming to a close, and I’m moving away from the GridPP side of my LSST work. This project with Southampton was primarily a use case to see how useful GridPP will be for general use from within LSST. There is a new mailing list for LSST-UK folks to contact if they have issues, so it may be left to that from now on. I’m not sure who will be in charge of the setup of LSST-GridPP how you describe. For now, Southampton’s setup works as needed, so for the short term no work is needed now, but longer-term there is a need for a more detailed overview of how LSST can properly utilise GridPP and to give new users a better understanding of how GridPP actually works. I will discuss with my project manager about how this could be done at our next meeting.
Thanks again for everybody’s help.
Darren ------------------------------------------------------------------------- Dr Darren J. White Applications Developer EPCC, University of Edinburgh
Email: d.white@epcc.ed.ac.uk <mailto:d.white@epcc.ed.ac.uk> Telephone: +44 (0) 131 6513415
Room 2411, James Clerk Maxwell Building, Kings Buildings, Mayfield Road, Edinburgh EH9 3JZ, United Kingdom -------------------------------------------------------------------------
On 5 Sep 2017, at 09:00, catalin.condurache@stfc.ac.uk <mailto:catalin.condurache@stfc.ac.uk> wrote:
Good morning Darren, 1) At WLCG level there used to be some time ago few repositories replicated every 10 mins (so-called fast replicas), so it can be done for LSST up to an extent (or at least one could give a try with a local repository). 2) Regarding LSST community, it is quite obvious that there is no consistency there with two repositories (opensciencegrid and in2p3). However these two repos are replicated by the Stratum-1 at RAL, so they should be accessible by any client with correct settings. If there is interest, we could work together and makelsst.in2p3.fr <http://lsst.in2p3.fr/>available via the CVMFS_CONFIG_REPOSITORY feature. 3) Another option I see is to add git tool to one of the existing lsst repos or to thegrid.cern.ch <http://grid.cern.ch/>repository (which should be available everywhere), but there might be other implications… Anyway at GridPP level we definitely could work to provide a consistent solution. I just didn’t know anyone involved in LSST from any UK site. But this could be the start. BTW are you attending the GridPP meeting in Lancaster? We could discuss more (or maybe finalise things) then. Regards, Catalin *From:*gridpp-dirac-users-bounces@imperial.ac.uk <mailto:gridpp-dirac-users-bounces@imperial.ac.uk> [mailto:gridpp-dirac-users-bounces@imperial.ac.uk]*On Behalf Of*WHITE Darren *Sent:*04 September 2017 16:52 *To:*gridpp-dirac-users@imperial.ac.uk <mailto:gridpp-dirac-users@imperial.ac.uk> *Subject:*Re: [Gridpp-Dirac-Users] Sites with git Hi Catalin, Even before this project got underway, we were warned to not do that for code that is in such active development. The time it takes for CVMFS updates to filter down can take far too much time for testing. I spent the last few days in Southampton last week and we were pushing changes to GitHub several times an hour that could instantly be available for testing. That’s simply not possible from what I understand/was told about CVMFS. Also, the Lyon repo is not available at all sites for LSST folks, as I found out the hard way. The specified repo I was told after many days of failure is apparently /cvmfs/lsst.opensciencegrid.org <http://lsst.opensciencegrid.org/>, which we successfully use to set up EUPS, which we then, in turn, use to set up the many packages needed for this specific code (which are also not available on either the LYON or opensciencegrid repos and so packaged into a tarball on an SE). This is another issue for new folks trying to work on GridPP. There’s no real clarity of what CVMFS repos are available, what sites you can run at, what tools they include both at the site or on the repos, etc. Cheers Darren ------------------------------------------------------------------------- Dr Darren J. White Applications Developer EPCC, University of Edinburgh
Email: d.white@epcc.ed.ac.uk <mailto:d.white@epcc.ed.ac.uk> Telephone: +44 (0) 131 6513415
Room 2411, James Clerk Maxwell Building, Kings Buildings, Mayfield Road, Edinburgh EH9 3JZ, United Kingdom -------------------------------------------------------------------------
On 4 Sep 2017, at 16:38,catalin.condurache@stfc.ac.uk <mailto:catalin.condurache@stfc.ac.uk>wrote: Hi Darren, Just a question: why you do not stick with the cvmfs repository? There is ‘lsst.in2p3.fr <http://lsst.in2p3.fr/>’ repository (maintained at Lyon and replicated by Stratum1 at RAL) where I’d assume all software should be provisioned from. Is this code you’re talking about only interesting the UK sites? In this case you may want to have another (UK regional) ‘lsst’ CVMFS repository. Anyway we at RAL have the policy to install a minimal distribution on our WNs (and git is not part of it), assuming the VOs are provisioning their software via CVMFS. Regards, Catalin Condurache RAL Tier-1 Services *From:*gridpp-dirac-users-bounces@imperial.ac.uk <mailto:gridpp-dirac-users-bounces@imperial.ac.uk>[mailto:gridpp-dirac-users-bounces@imperial.ac.uk]*On Behalf Of*WHITE Darren *Sent:*04 September 2017 15:54 *To:*gridpp-dirac-users@imperial.ac.uk <mailto:gridpp-dirac-users@imperial.ac.uk> *Subject:*[Gridpp-Dirac-Users] Sites with git Hi there, I’m working with LSST folks in Southampton who have code that is updated quite regularly. A while ago, I tested to see whether git could clone at time of execution (I cannot remember which site this was at), and it was a success. However, RAL does not have git installed, as similar code fails at that site. I’ve asked this before and didn’t really get a full answer. Is there any way of knowing which sites have which software installed? It would save new users a lot of pain when trying to get up and running, or even just testing. Cheers Darren ------------------------------------------------------------------------- Dr Darren J. White Applications Developer EPCC, University of Edinburgh
Email: d.white@epcc.ed.ac.uk <mailto:d.white@epcc.ed.ac.uk> Telephone: +44 (0) 131 6513415
Room 2411, James Clerk Maxwell Building, Kings Buildings, Mayfield Road, Edinburgh EH9 3JZ, United Kingdom ------------------------------------------------------------------------- -- _______________________________________________ Gridpp-Dirac-Users mailing list Gridpp-Dirac-Users@imperial.ac.uk <mailto:Gridpp-Dirac-Users@imperial.ac.uk> https://mailman.ic.ac.uk/mailman/listinfo/gridpp-dirac-users
-- _______________________________________________ Gridpp-Dirac-Users mailing list Gridpp-Dirac-Users@imperial.ac.uk <mailto:Gridpp-Dirac-Users@imperial.ac.uk> https://mailman.ic.ac.uk/mailman/listinfo/gridpp-dirac-users
The University of Edinburgh is a charitable body, registered in Scotland, with registration number SC005336.
-- Respect is a rational process. \\// Fatti non foste a viver come bruti, ma per seguir virtute e canoscenza(Dante) For Ur-Fascism, disagreement is treason. (U. Eco) But but but her emails... covfefe! -- _______________________________________________ Gridpp-Dirac-Users mailing list Gridpp-Dirac-Users@imperial.ac.uk <mailto:Gridpp-Dirac-Users@imperial.ac.uk> https://mailman.ic.ac.uk/mailman/listinfo/gridpp-dirac-users
The University of Edinburgh is a charitable body, registered in Scotland, with registration number SC005336.
-- Respect is a rational process. \\// Fatti non foste a viver come bruti, ma per seguir virtute e canoscenza(Dante) For Ur-Fascism, disagreement is treason. (U. Eco) But but but her emails... covfefe!
Hi, Is there a simple way to find out what repositories exist at which sites from a job running as gridpp or would you have to be a member of each VO to see what repositories that VO had. If the former then I could run jobs to build a list (what would the command be? ls /cvmfs.. ?) if the latter I can't help. Cheers +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ Steve Lloyd Emeritus Professor, School of Physics and Astronomy Queen Mary University of London, Mile End Road, London E1 4NS, UK E-mail: s.l.lloyd@qmul.ac.uk +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
On 6 Sep 2017, at 13:56, Alessandra Forti <Alessandra.Forti@cern.ch> wrote:
Hi Darren,
we have an LSST_UK wiki page for the moment is mostly for sites but we can add a User section or we can start another wiki page for users and add the information there.
https://www.gridpp.ac.uk/wiki/LSST_UK
cheers alessandra
On 06/09/2017 12:53, WHITE Darren wrote:
Hi everyone,
Thanks for the discussion. I’ve learned quite a bit about GridPP over the past couple of days, much of it useful. I understand that it is not the place for GridPP support to list every single software package. The changes made for each individual experiment make that impossible. But I think it would be useful to have some list of what cvmfs repositories are available for each project, so people can then go and check what is available in those repos.
That was the biggest issue I came to first. I wrote my software linking to the Lyon cvmfs on my local machine, based on the suggestions of other people within GridPP at the time, only to find that it wasn’t mounted as default on the sites, but that the OSG version was. Now, in the past couple of months, the Lyon cvmfs is being mounted, but only at certain sites. There’s no way of being able to find that out unless you are in that small group of people 'in the know’ or by emailing support (which I’m sure you guys would prefer to not be the case for each new user from a fledgling project).
From my short time working with LSST it seems like it isn’t a singular organisation like some of the other projects using GridPP, or at least not mature enough for that to be the case, and it’s made up primarily of small subgroups tackling individual problems. Those groups (like Southampton) are happy to use GridPP as long as they have some idea of what to expect and what they have to work with (in terms of what repos are available, how they can bring across their own software or install additional dependencies, for example). Right now, that’s not quite the case from our recent experience. There’s no documentation on what cvmfs repos are available, there’s no documentation on how to package your own software onto SEs manually (like we had to do since the repos didn’t have the right versions of software we needed), etc. Decent, simple beginners guides to these things would go a long way to making the experience smoother. If I am able to contribute more time to this within my allotted time on LSST, I’m happy to help, but I feel it is something that does need to be done if there is to be better support for new/small/inexperienced groups to use GridPP.
Thanks again,
Darren
------------------------------------------------------------------------- Dr Darren J. White Applications Developer EPCC, University of Edinburgh
Email: d.white@epcc.ed.ac.uk Telephone: +44 (0) 131 6513415
Room 2411, James Clerk Maxwell Building, Kings Buildings, Mayfield Road, Edinburgh EH9 3JZ, United Kingdom -------------------------------------------------------------------------
On 6 Sep 2017, at 11:56, Alessandra Forti <Alessandra.Forti@cern.ch> wrote:
Hi Darren,
it is cumbersome and difficult to maintain to list any type of software a shared site might have installed. Sites on the grid typically support a number of VOs. This is why VOs tend to package what they need in their repositories (CVMFS, git or other).
Software distribution is not only a gridpp problem. GridPP offers the means to install and distribute software, in particular cvmfs, but the experiments should decide what they want to distribute. LHC experiments for example have a team of people managing the core of their software upon which the users build their code. Users have then just to maintain their own limited number of files.
In UKLSST case it is true that the code development made CVMFS unusable for the code that changes often. However it could be used for established packages that you are not sure you may find at every site and don't change. The UK had a dedicate subdir in the OSG CVMFS repository for this which we created when Joe Zuntz was the testing user.
cheers alessandra
On 05/09/2017 10:16, WHITE Darren wrote:
Hi everyone,
Thanks for the list Steve. It’s a good place to start, but there isn’t a place for people within those VOs to easily find what repos/software are available, which is the bigger issue.
Thanks for the reply Catalin. The options sound useful, but I’m not sure how much time I can put towards this. The work I’m doing is now coming to a close, and I’m moving away from the GridPP side of my LSST work. This project with Southampton was primarily a use case to see how useful GridPP will be for general use from within LSST. There is a new mailing list for LSST-UK folks to contact if they have issues, so it may be left to that from now on. I’m not sure who will be in charge of the setup of LSST-GridPP how you describe. For now, Southampton’s setup works as needed, so for the short term no work is needed now, but longer-term there is a need for a more detailed overview of how LSST can properly utilise GridPP and to give new users a better understanding of how GridPP actually works. I will discuss with my project manager about how this could be done at our next meeting.
Thanks again for everybody’s help.
Darren ------------------------------------------------------------------------- Dr Darren J. White Applications Developer EPCC, University of Edinburgh
Email: d.white@epcc.ed.ac.uk Telephone: +44 (0) 131 6513415
Room 2411, James Clerk Maxwell Building, Kings Buildings, Mayfield Road, Edinburgh EH9 3JZ, United Kingdom -------------------------------------------------------------------------
On 5 Sep 2017, at 09:00, catalin.condurache@stfc.ac.uk wrote:
Good morning Darren,
1) At WLCG level there used to be some time ago few repositories replicated every 10 mins (so-called fast replicas), so it can be done for LSST up to an extent (or at least one could give a try with a local repository).
2) Regarding LSST community, it is quite obvious that there is no consistency there with two repositories (opensciencegrid and in2p3). However these two repos are replicated by the Stratum-1 at RAL, so they should be accessible by any client with correct settings. If there is interest, we could work together and make lsst.in2p3.fr available via the CVMFS_CONFIG_REPOSITORY feature.
3) Another option I see is to add git tool to one of the existing lsst repos or to the grid.cern.ch repository (which should be available everywhere), but there might be other implications…
Anyway at GridPP level we definitely could work to provide a consistent solution. I just didn’t know anyone involved in LSST from any UK site. But this could be the start.
BTW are you attending the GridPP meeting in Lancaster? We could discuss more (or maybe finalise things) then.
Regards, Catalin
From: gridpp-dirac-users-bounces@imperial.ac.uk [mailto:gridpp-dirac-users-bounces@imperial.ac.uk] On Behalf Of WHITE Darren Sent: 04 September 2017 16:52 To: gridpp-dirac-users@imperial.ac.uk Subject: Re: [Gridpp-Dirac-Users] Sites with git
Hi Catalin,
Even before this project got underway, we were warned to not do that for code that is in such active development. The time it takes for CVMFS updates to filter down can take far too much time for testing. I spent the last few days in Southampton last week and we were pushing changes to GitHub several times an hour that could instantly be available for testing. That’s simply not possible from what I understand/was told about CVMFS.
Also, the Lyon repo is not available at all sites for LSST folks, as I found out the hard way. The specified repo I was told after many days of failure is apparently /cvmfs/lsst.opensciencegrid.org, which we successfully use to set up EUPS, which we then, in turn, use to set up the many packages needed for this specific code (which are also not available on either the LYON or opensciencegrid repos and so packaged into a tarball on an SE). This is another issue for new folks trying to work on GridPP. There’s no real clarity of what CVMFS repos are available, what sites you can run at, what tools they include both at the site or on the repos, etc.
Cheers
Darren ------------------------------------------------------------------------- Dr Darren J. White Applications Developer EPCC, University of Edinburgh
Email: d.white@epcc.ed.ac.uk Telephone: +44 (0) 131 6513415
Room 2411, James Clerk Maxwell Building, Kings Buildings, Mayfield Road, Edinburgh EH9 3JZ, United Kingdom -------------------------------------------------------------------------
On 4 Sep 2017, at 16:38, catalin.condurache@stfc.ac.uk wrote:
Hi Darren,
Just a question: why you do not stick with the cvmfs repository? There is ‘lsst.in2p3.fr’ repository (maintained at Lyon and replicated by Stratum1 at RAL) where I’d assume all software should be provisioned from.
Is this code you’re talking about only interesting the UK sites? In this case you may want to have another (UK regional) ‘lsst’ CVMFS repository.
Anyway we at RAL have the policy to install a minimal distribution on our WNs (and git is not part of it), assuming the VOs are provisioning their software via CVMFS.
Regards, Catalin Condurache RAL Tier-1 Services
From: gridpp-dirac-users-bounces@imperial.ac.uk [mailto:gridpp-dirac-users-bounces@imperial.ac.uk] On Behalf Of WHITE Darren Sent: 04 September 2017 15:54 To: gridpp-dirac-users@imperial.ac.uk Subject: [Gridpp-Dirac-Users] Sites with git
Hi there,
I’m working with LSST folks in Southampton who have code that is updated quite regularly. A while ago, I tested to see whether git could clone at time of execution (I cannot remember which site this was at), and it was a success. However, RAL does not have git installed, as similar code fails at that site. I’ve asked this before and didn’t really get a full answer. Is there any way of knowing which sites have which software installed? It would save new users a lot of pain when trying to get up and running, or even just testing.
Cheers
Darren ------------------------------------------------------------------------- Dr Darren J. White Applications Developer EPCC, University of Edinburgh
Email: d.white@epcc.ed.ac.uk Telephone: +44 (0) 131 6513415
Room 2411, James Clerk Maxwell Building, Kings Buildings, Mayfield Road, Edinburgh EH9 3JZ, United Kingdom -------------------------------------------------------------------------
-- _______________________________________________ Gridpp-Dirac-Users mailing list Gridpp-Dirac-Users@imperial.ac.uk https://mailman.ic.ac.uk/mailman/listinfo/gridpp-dirac-users
-- _______________________________________________ Gridpp-Dirac-Users mailing list Gridpp-Dirac-Users@imperial.ac.uk https://mailman.ic.ac.uk/mailman/listinfo/gridpp-dirac-users
The University of Edinburgh is a charitable body, registered in Scotland, with registration number SC005336.
-- Respect is a rational process. \\// Fatti non foste a viver come bruti, ma per seguir virtute e canoscenza(Dante) For Ur-Fascism, disagreement is treason. (U. Eco) But but but her emails... covfefe!
-- _______________________________________________ Gridpp-Dirac-Users mailing list Gridpp-Dirac-Users@imperial.ac.uk https://mailman.ic.ac.uk/mailman/listinfo/gridpp-dirac-users
The University of Edinburgh is a charitable body, registered in Scotland, with registration number SC005336.
-- Respect is a rational process. \\// Fatti non foste a viver come bruti, ma per seguir virtute e canoscenza(Dante) For Ur-Fascism, disagreement is treason. (U. Eco) But but but her emails... covfefe!
-- _______________________________________________ Gridpp-Dirac-Users mailing list Gridpp-Dirac-Users@imperial.ac.uk https://mailman.ic.ac.uk/mailman/listinfo/gridpp-dirac-users
Hi, You got this the wrong way around. If a site supports a given VO, they support the cvmfs repos the VO needs. It is however the *responsibility of the VO* to communicate the repositories they need to GridPP (sticking them into the VO id card would be a good start *). Now I think we worked out what went wrong with lsst and why the information was not propagated, but this is not a reason to mess with a principally working system, VOs also need to be strongly discouraged to make use of cvmfs repos that are not theirs and just happen to be present coincidentally, as other VOs might withdraw software at any time. * I actually asked for a dedicated field for this, but it can go anywhere: https://ggus.eu/?mode=ticket_info&ticket_id=127663 Regards, Daniela On 7 September 2017 at 10:11, Steve Lloyd <s.l.lloyd@qmul.ac.uk> wrote:
Hi, Is there a simple way to find out what repositories exist at which sites from a job running as gridpp or would you have to be a member of each VO to see what repositories that VO had. If the former then I could run jobs to build a list (what would the command be? ls /cvmfs.. ?) if the latter I can't help. Cheers +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ Steve Lloyd Emeritus Professor, School of Physics and Astronomy Queen Mary University of London, Mile End Road, London E1 4NS, UK E-mail: s.l.lloyd@qmul.ac.uk +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
On 6 Sep 2017, at 13:56, Alessandra Forti <Alessandra.Forti@cern.ch> wrote:
Hi Darren,
we have an LSST_UK wiki page for the moment is mostly for sites but we can add a User section or we can start another wiki page for users and add the information there.
https://www.gridpp.ac.uk/wiki/LSST_UK
cheers alessandra
On 06/09/2017 12:53, WHITE Darren wrote:
Hi everyone,
Thanks for the discussion. I’ve learned quite a bit about GridPP over the past couple of days, much of it useful. I understand that it is not the place for GridPP support to list every single software package. The changes made for each individual experiment make that impossible. But I think it would be useful to have some list of what cvmfs repositories are available for each project, so people can then go and check what is available in those repos.
That was the biggest issue I came to first. I wrote my software linking to the Lyon cvmfs on my local machine, based on the suggestions of other people within GridPP at the time, only to find that it wasn’t mounted as default on the sites, but that the OSG version was. Now, in the past couple of months, the Lyon cvmfs is being mounted, but only at certain sites. There’s no way of being able to find that out unless you are in that small group of people 'in the know’ or by emailing support (which I’m sure you guys would prefer to not be the case for each new user from a fledgling project).
From my short time working with LSST it seems like it isn’t a singular organisation like some of the other projects using GridPP, or at least not mature enough for that to be the case, and it’s made up primarily of small subgroups tackling individual problems. Those groups (like Southampton) are happy to use GridPP as long as they have some idea of what to expect and what they have to work with (in terms of what repos are available, how they can bring across their own software or install additional dependencies, for example). Right now, that’s not quite the case from our recent experience. There’s no documentation on what cvmfs repos are available, there’s no documentation on how to package your own software onto SEs manually (like we had to do since the repos didn’t have the right versions of software we needed), etc. Decent, simple beginners guides to these things would go a long way to making the experience smoother. If I am able to contribute more time to this within my allotted time on LSST, I’m happy to help, but I feel it is something that does need to be done if there is to be better support for new/small/inexperienced groups to use GridPP.
Thanks again,
Darren
------------------------------------------------------------------------- Dr Darren J. White Applications Developer EPCC, University of Edinburgh
Email: d.white@epcc.ed.ac.uk Telephone: +44 (0) 131 6513415
Room 2411, James Clerk Maxwell Building, Kings Buildings, Mayfield Road, Edinburgh EH9 3JZ, United Kingdom -------------------------------------------------------------------------
On 6 Sep 2017, at 11:56, Alessandra Forti <Alessandra.Forti@cern.ch> wrote:
Hi Darren,
it is cumbersome and difficult to maintain to list any type of software a shared site might have installed. Sites on the grid typically support a number of VOs. This is why VOs tend to package what they need in their repositories (CVMFS, git or other).
Software distribution is not only a gridpp problem. GridPP offers the means to install and distribute software, in particular cvmfs, but the experiments should decide what they want to distribute. LHC experiments for example have a team of people managing the core of their software upon which the users build their code. Users have then just to maintain their own limited number of files.
In UKLSST case it is true that the code development made CVMFS unusable for the code that changes often. However it could be used for established packages that you are not sure you may find at every site and don't change. The UK had a dedicate subdir in the OSG CVMFS repository for this which we created when Joe Zuntz was the testing user.
cheers alessandra
On 05/09/2017 10:16, WHITE Darren wrote:
Hi everyone,
Thanks for the list Steve. It’s a good place to start, but there isn’t a place for people within those VOs to easily find what repos/software are available, which is the bigger issue.
Thanks for the reply Catalin. The options sound useful, but I’m not sure how much time I can put towards this. The work I’m doing is now coming to a close, and I’m moving away from the GridPP side of my LSST work. This project with Southampton was primarily a use case to see how useful GridPP will be for general use from within LSST. There is a new mailing list for LSST-UK folks to contact if they have issues, so it may be left to that from now on. I’m not sure who will be in charge of the setup of LSST-GridPP how you describe. For now, Southampton’s setup works as needed, so for the short term no work is needed now, but longer-term there is a need for a more detailed overview of how LSST can properly utilise GridPP and to give new users a better understanding of how GridPP actually works. I will discuss with my project manager about how this could be done at our next meeting.
Thanks again for everybody’s help.
Darren ------------------------------------------------------------------------- Dr Darren J. White Applications Developer EPCC, University of Edinburgh
Email: d.white@epcc.ed.ac.uk Telephone: +44 (0) 131 6513415
Room 2411, James Clerk Maxwell Building, Kings Buildings, Mayfield Road, Edinburgh EH9 3JZ, United Kingdom -------------------------------------------------------------------------
On 5 Sep 2017, at 09:00, catalin.condurache@stfc.ac.uk wrote:
Good morning Darren,
1) At WLCG level there used to be some time ago few repositories replicated every 10 mins (so-called fast replicas), so it can be done for LSST up to an extent (or at least one could give a try with a local repository).
2) Regarding LSST community, it is quite obvious that there is no consistency there with two repositories (opensciencegrid and in2p3). However these two repos are replicated by the Stratum-1 at RAL, so they should be accessible by any client with correct settings. If there is interest, we could work together and make lsst.in2p3.fr available via the CVMFS_CONFIG_REPOSITORY feature.
3) Another option I see is to add git tool to one of the existing lsst repos or to the grid.cern.ch repository (which should be available everywhere), but there might be other implications…
Anyway at GridPP level we definitely could work to provide a consistent solution. I just didn’t know anyone involved in LSST from any UK site. But this could be the start.
BTW are you attending the GridPP meeting in Lancaster? We could discuss more (or maybe finalise things) then.
Regards, Catalin
From: gridpp-dirac-users-bounces@imperial.ac.uk [mailto:gridpp-dirac-users-bounces@imperial.ac.uk] On Behalf Of WHITE Darren Sent: 04 September 2017 16:52 To: gridpp-dirac-users@imperial.ac.uk Subject: Re: [Gridpp-Dirac-Users] Sites with git
Hi Catalin,
Even before this project got underway, we were warned to not do that for code that is in such active development. The time it takes for CVMFS updates to filter down can take far too much time for testing. I spent the last few days in Southampton last week and we were pushing changes to GitHub several times an hour that could instantly be available for testing. That’s simply not possible from what I understand/was told about CVMFS.
Also, the Lyon repo is not available at all sites for LSST folks, as I found out the hard way. The specified repo I was told after many days of failure is apparently /cvmfs/lsst.opensciencegrid.org, which we successfully use to set up EUPS, which we then, in turn, use to set up the many packages needed for this specific code (which are also not available on either the LYON or opensciencegrid repos and so packaged into a tarball on an SE). This is another issue for new folks trying to work on GridPP. There’s no real clarity of what CVMFS repos are available, what sites you can run at, what tools they include both at the site or on the repos, etc.
Cheers
Darren ------------------------------------------------------------------------- Dr Darren J. White Applications Developer EPCC, University of Edinburgh
Email: d.white@epcc.ed.ac.uk Telephone: +44 (0) 131 6513415
Room 2411, James Clerk Maxwell Building, Kings Buildings, Mayfield Road, Edinburgh EH9 3JZ, United Kingdom -------------------------------------------------------------------------
On 4 Sep 2017, at 16:38, catalin.condurache@stfc.ac.uk wrote:
Hi Darren,
Just a question: why you do not stick with the cvmfs repository? There is ‘lsst.in2p3.fr’ repository (maintained at Lyon and replicated by Stratum1 at RAL) where I’d assume all software should be provisioned from.
Is this code you’re talking about only interesting the UK sites? In this case you may want to have another (UK regional) ‘lsst’ CVMFS repository.
Anyway we at RAL have the policy to install a minimal distribution on our WNs (and git is not part of it), assuming the VOs are provisioning their software via CVMFS.
Regards, Catalin Condurache RAL Tier-1 Services
From: gridpp-dirac-users-bounces@imperial.ac.uk [mailto:gridpp-dirac-users-bounces@imperial.ac.uk] On Behalf Of WHITE Darren Sent: 04 September 2017 15:54 To: gridpp-dirac-users@imperial.ac.uk Subject: [Gridpp-Dirac-Users] Sites with git
Hi there,
I’m working with LSST folks in Southampton who have code that is updated quite regularly. A while ago, I tested to see whether git could clone at time of execution (I cannot remember which site this was at), and it was a success. However, RAL does not have git installed, as similar code fails at that site. I’ve asked this before and didn’t really get a full answer. Is there any way of knowing which sites have which software installed? It would save new users a lot of pain when trying to get up and running, or even just testing.
Cheers
Darren ------------------------------------------------------------------------- Dr Darren J. White Applications Developer EPCC, University of Edinburgh
Email: d.white@epcc.ed.ac.uk Telephone: +44 (0) 131 6513415
Room 2411, James Clerk Maxwell Building, Kings Buildings, Mayfield Road, Edinburgh EH9 3JZ, United Kingdom -------------------------------------------------------------------------
-- _______________________________________________ Gridpp-Dirac-Users mailing list Gridpp-Dirac-Users@imperial.ac.uk https://mailman.ic.ac.uk/mailman/listinfo/gridpp-dirac-users
-- _______________________________________________ Gridpp-Dirac-Users mailing list Gridpp-Dirac-Users@imperial.ac.uk https://mailman.ic.ac.uk/mailman/listinfo/gridpp-dirac-users
The University of Edinburgh is a charitable body, registered in Scotland, with registration number SC005336.
-- Respect is a rational process. \\// Fatti non foste a viver come bruti, ma per seguir virtute e canoscenza(Dante) For Ur-Fascism, disagreement is treason. (U. Eco) But but but her emails... covfefe!
-- _______________________________________________ Gridpp-Dirac-Users mailing list Gridpp-Dirac-Users@imperial.ac.uk https://mailman.ic.ac.uk/mailman/listinfo/gridpp-dirac-users
The University of Edinburgh is a charitable body, registered in Scotland, with registration number SC005336.
-- Respect is a rational process. \\// Fatti non foste a viver come bruti, ma per seguir virtute e canoscenza(Dante) For Ur-Fascism, disagreement is treason. (U. Eco) But but but her emails... covfefe!
-- _______________________________________________ Gridpp-Dirac-Users mailing list Gridpp-Dirac-Users@imperial.ac.uk https://mailman.ic.ac.uk/mailman/listinfo/gridpp-dirac-users
-- _______________________________________________ Gridpp-Dirac-Users mailing list Gridpp-Dirac-Users@imperial.ac.uk https://mailman.ic.ac.uk/mailman/listinfo/gridpp-dirac-users
-- Sent from the pit of despair ----------------------------------------------------------- daniela.bauer@imperial.ac.uk HEP Group/Physics Dep Imperial College London, SW7 2BW Tel: +44-(0)20-75947810 http://www.hep.ph.ic.ac.uk/~dbauer/
Hi Catalin, wrt cvmfs: I thought it was agreed that LSST uses: /cvmfs/lsst.opensciencegrid.org Has this changed ? regards, Daniela On 4 September 2017 at 16:38, <catalin.condurache@stfc.ac.uk> wrote:
Hi Darren,
Just a question: why you do not stick with the cvmfs repository? There is ‘lsst.in2p3.fr’ repository (maintained at Lyon and replicated by Stratum1 at RAL) where I’d assume all software should be provisioned from.
Is this code you’re talking about only interesting the UK sites? In this case you may want to have another (UK regional) ‘lsst’ CVMFS repository.
Anyway we at RAL have the policy to install a minimal distribution on our WNs (and git is not part of it), assuming the VOs are provisioning their software via CVMFS.
Regards,
Catalin Condurache
RAL Tier-1 Services
From: gridpp-dirac-users-bounces@imperial.ac.uk [mailto:gridpp-dirac-users-bounces@imperial.ac.uk] On Behalf Of WHITE Darren Sent: 04 September 2017 15:54 To: gridpp-dirac-users@imperial.ac.uk Subject: [Gridpp-Dirac-Users] Sites with git
Hi there,
I’m working with LSST folks in Southampton who have code that is updated quite regularly. A while ago, I tested to see whether git could clone at time of execution (I cannot remember which site this was at), and it was a success. However, RAL does not have git installed, as similar code fails at that site. I’ve asked this before and didn’t really get a full answer. Is there any way of knowing which sites have which software installed? It would save new users a lot of pain when trying to get up and running, or even just testing.
Cheers
Darren
------------------------------------------------------------------------- Dr Darren J. White Applications Developer EPCC, University of Edinburgh
Email: d.white@epcc.ed.ac.uk Telephone: +44 (0) 131 6513415
Room 2411, James Clerk Maxwell Building, Kings Buildings, Mayfield Road, Edinburgh EH9 3JZ, United Kingdom -------------------------------------------------------------------------
-- _______________________________________________ Gridpp-Dirac-Users mailing list Gridpp-Dirac-Users@imperial.ac.uk https://mailman.ic.ac.uk/mailman/listinfo/gridpp-dirac-users
-- Sent from the pit of despair ----------------------------------------------------------- daniela.bauer@imperial.ac.uk HEP Group/Physics Dep Imperial College London, SW7 2BW Tel: +44-(0)20-75947810 http://www.hep.ph.ic.ac.uk/~dbauer/
Hi Daniela, It might be that the agreement was between LSST users (at some point in the past), however it should not matter which 'lsst' repos are replicated. Anyway in July we tried to enable 'lsst' jobs at RAL and following discussions with few people involved in LSST I started to replicate the lsst.in2p3.fr repo from Lyon. Once this replica was available, the 'lsst' jobs started to succeed at RAL. So, as I highlighted in my previous message, these inconsistencies should be sorted out internally within LSST, at least in my opinion. Regards, Catalin
-----Original Message----- From: gridpp-dirac-users-bounces@imperial.ac.uk [mailto:gridpp-dirac-users- bounces@imperial.ac.uk] On Behalf Of Daniela Bauer Sent: 05 September 2017 12:09 To: gridpp-dirac-users@imperial.ac.uk Subject: Re: [Gridpp-Dirac-Users] Sites with git
Hi Catalin,
wrt cvmfs: I thought it was agreed that LSST uses: /cvmfs/lsst.opensciencegrid.org Has this changed ?
regards, Daniela
On 4 September 2017 at 16:38, <catalin.condurache@stfc.ac.uk> wrote:
Hi Darren,
Just a question: why you do not stick with the cvmfs repository? There is ‘lsst.in2p3.fr’ repository (maintained at Lyon and replicated by Stratum1 at RAL) where I’d assume all software should be provisioned from.
Is this code you’re talking about only interesting the UK sites? In this case you may want to have another (UK regional) ‘lsst’ CVMFS repository.
Anyway we at RAL have the policy to install a minimal distribution on our WNs (and git is not part of it), assuming the VOs are provisioning their software via CVMFS.
Regards,
Catalin Condurache
RAL Tier-1 Services
From: gridpp-dirac-users-bounces@imperial.ac.uk [mailto:gridpp-dirac-users-bounces@imperial.ac.uk] On Behalf Of WHITE Darren Sent: 04 September 2017 15:54 To: gridpp-dirac-users@imperial.ac.uk Subject: [Gridpp-Dirac-Users] Sites with git
Hi there,
I’m working with LSST folks in Southampton who have code that is updated quite regularly. A while ago, I tested to see whether git could clone at time of execution (I cannot remember which site this was at), and it was a success. However, RAL does not have git installed, as similar code fails at that site. I’ve asked this before and didn’t really get a full answer. Is there any way of knowing which sites have which software installed? It would save new users a lot of pain when trying to get up and running, or even just testing.
Cheers
Darren
---------------------------------------------------------------------- --- Dr Darren J. White Applications Developer EPCC, University of Edinburgh
Email: d.white@epcc.ed.ac.uk Telephone: +44 (0) 131 6513415
Room 2411, James Clerk Maxwell Building, Kings Buildings, Mayfield Road, Edinburgh EH9 3JZ, United Kingdom ---------------------------------------------------------------------- ---
-- _______________________________________________ Gridpp-Dirac-Users mailing list Gridpp-Dirac-Users@imperial.ac.uk https://mailman.ic.ac.uk/mailman/listinfo/gridpp-dirac-users
-- Sent from the pit of despair
----------------------------------------------------------- daniela.bauer@imperial.ac.uk HEP Group/Physics Dep Imperial College London, SW7 2BW Tel: +44-(0)20-75947810 http://www.hep.ph.ic.ac.uk/~dbauer/
-- _______________________________________________ Gridpp-Dirac-Users mailing list Gridpp-Dirac-Users@imperial.ac.uk https://mailman.ic.ac.uk/mailman/listinfo/gridpp-dirac-users
HI Catalin, I agree this is an lsst internal problem, but if you quietly enable a repo, the jobs will run at RAL, but we then have to debug every other site who didn't know, and that's where it all goes wrong. Regards, Daniela On 5 September 2017 at 15:44, <catalin.condurache@stfc.ac.uk> wrote:
Hi Daniela,
It might be that the agreement was between LSST users (at some point in the past), however it should not matter which 'lsst' repos are replicated. Anyway in July we tried to enable 'lsst' jobs at RAL and following discussions with few people involved in LSST I started to replicate the lsst.in2p3.fr repo from Lyon. Once this replica was available, the 'lsst' jobs started to succeed at RAL.
So, as I highlighted in my previous message, these inconsistencies should be sorted out internally within LSST, at least in my opinion.
Regards, Catalin
-----Original Message----- From: gridpp-dirac-users-bounces@imperial.ac.uk [mailto:gridpp-dirac-users- bounces@imperial.ac.uk] On Behalf Of Daniela Bauer Sent: 05 September 2017 12:09 To: gridpp-dirac-users@imperial.ac.uk Subject: Re: [Gridpp-Dirac-Users] Sites with git
Hi Catalin,
wrt cvmfs: I thought it was agreed that LSST uses: /cvmfs/lsst.opensciencegrid.org Has this changed ?
regards, Daniela
On 4 September 2017 at 16:38, <catalin.condurache@stfc.ac.uk> wrote:
Hi Darren,
Just a question: why you do not stick with the cvmfs repository? There is ‘lsst.in2p3.fr’ repository (maintained at Lyon and replicated by Stratum1 at RAL) where I’d assume all software should be provisioned from.
Is this code you’re talking about only interesting the UK sites? In this case you may want to have another (UK regional) ‘lsst’ CVMFS repository.
Anyway we at RAL have the policy to install a minimal distribution on our WNs (and git is not part of it), assuming the VOs are provisioning their software via CVMFS.
Regards,
Catalin Condurache
RAL Tier-1 Services
From: gridpp-dirac-users-bounces@imperial.ac.uk [mailto:gridpp-dirac-users-bounces@imperial.ac.uk] On Behalf Of WHITE Darren Sent: 04 September 2017 15:54 To: gridpp-dirac-users@imperial.ac.uk Subject: [Gridpp-Dirac-Users] Sites with git
Hi there,
I’m working with LSST folks in Southampton who have code that is updated quite regularly. A while ago, I tested to see whether git could clone at time of execution (I cannot remember which site this was at), and it was a success. However, RAL does not have git installed, as similar code fails at that site. I’ve asked this before and didn’t really get a full answer. Is there any way of knowing which sites have which software installed? It would save new users a lot of pain when trying to get up and running, or even just testing.
Cheers
Darren
---------------------------------------------------------------------- --- Dr Darren J. White Applications Developer EPCC, University of Edinburgh
Email: d.white@epcc.ed.ac.uk Telephone: +44 (0) 131 6513415
Room 2411, James Clerk Maxwell Building, Kings Buildings, Mayfield Road, Edinburgh EH9 3JZ, United Kingdom ---------------------------------------------------------------------- ---
-- _______________________________________________ Gridpp-Dirac-Users mailing list Gridpp-Dirac-Users@imperial.ac.uk https://mailman.ic.ac.uk/mailman/listinfo/gridpp-dirac-users
-- Sent from the pit of despair
----------------------------------------------------------- daniela.bauer@imperial.ac.uk HEP Group/Physics Dep Imperial College London, SW7 2BW Tel: +44-(0)20-75947810 http://www.hep.ph.ic.ac.uk/~dbauer/
-- _______________________________________________ Gridpp-Dirac-Users mailing list Gridpp-Dirac-Users@imperial.ac.uk https://mailman.ic.ac.uk/mailman/listinfo/gridpp-dirac-users --
Gridpp-Dirac-Users mailing list Gridpp-Dirac-Users@imperial.ac.uk https://mailman.ic.ac.uk/mailman/listinfo/gridpp-dirac-users
-- Sent from the pit of despair ----------------------------------------------------------- daniela.bauer@imperial.ac.uk HEP Group/Physics Dep Imperial College London, SW7 2BW Tel: +44-(0)20-75947810 http://www.hep.ph.ic.ac.uk/~dbauer/
Hi Daniela,
I agree this is an lsst internal problem, but if you quietly enable a repo, the jobs will run at RAL, but we then have to debug every other site who didn't know, and that's where it all goes wrong.
It wasn't a quiet enabling of the repos, since several people were aware of it. And as long as LSST contacts (and RAL people) knew about it, didn't consider to circulate more (and using what channels?). About debugging, I assume each job environment contains some info about CVMFS access, but then it will be in the form of lsst.<domain1> or lsst.<domain2>, so easy to spot a inconsistency. Definitely a job won't randomly access one or another 'lsst' repo. Cheers, Catalin
Regards, Daniela
On 5 September 2017 at 15:44, <catalin.condurache@stfc.ac.uk> wrote:
Hi Daniela,
It might be that the agreement was between LSST users (at some point in the past), however it should not matter which 'lsst' repos are replicated. Anyway in July we tried to enable 'lsst' jobs at RAL and following discussions with few people involved in LSST I started to replicate the lsst.in2p3.fr repo from Lyon. Once this replica was available, the 'lsst' jobs started to succeed at RAL.
So, as I highlighted in my previous message, these inconsistencies should be sorted out internally within LSST, at least in my opinion.
Regards, Catalin
-----Original Message----- From: gridpp-dirac-users-bounces@imperial.ac.uk [mailto:gridpp-dirac-users- bounces@imperial.ac.uk] On Behalf Of Daniela Bauer Sent: 05 September 2017 12:09 To: gridpp-dirac-users@imperial.ac.uk Subject: Re: [Gridpp-Dirac-Users] Sites with git
Hi Catalin,
wrt cvmfs: I thought it was agreed that LSST uses: /cvmfs/lsst.opensciencegrid.org Has this changed ?
regards, Daniela
On 4 September 2017 at 16:38, <catalin.condurache@stfc.ac.uk> wrote:
Hi Darren,
Just a question: why you do not stick with the cvmfs repository? There is ‘lsst.in2p3.fr’ repository (maintained at Lyon and replicated by Stratum1 at RAL) where I’d assume all software should be provisioned from.
Is this code you’re talking about only interesting the UK sites? In this case you may want to have another (UK regional) ‘lsst’ CVMFS repository.
Anyway we at RAL have the policy to install a minimal distribution on our WNs (and git is not part of it), assuming the VOs are provisioning their software via CVMFS.
Regards,
Catalin Condurache
RAL Tier-1 Services
From: gridpp-dirac-users-bounces@imperial.ac.uk [mailto:gridpp-dirac-users-bounces@imperial.ac.uk] On Behalf Of WHITE Darren Sent: 04 September 2017 15:54 To: gridpp-dirac-users@imperial.ac.uk Subject: [Gridpp-Dirac-Users] Sites with git
Hi there,
I’m working with LSST folks in Southampton who have code that is updated quite regularly. A while ago, I tested to see whether git could clone at time of execution (I cannot remember which site this was at), and it was a success. However, RAL does not have git installed, as similar code fails at that site. I’ve asked this before and didn’t really get a full answer. Is there any way of knowing which sites have which software installed? It would save new users a lot of pain when trying to get up and running, or even just testing.
Cheers
Darren
------------------------------------------------------------------- --- --- Dr Darren J. White Applications Developer EPCC, University of Edinburgh
Email: d.white@epcc.ed.ac.uk Telephone: +44 (0) 131 6513415
Room 2411, James Clerk Maxwell Building, Kings Buildings, Mayfield Road, Edinburgh EH9 3JZ, United Kingdom ------------------------------------------------------------------- --- ---
-- _______________________________________________ Gridpp-Dirac-Users mailing list Gridpp-Dirac-Users@imperial.ac.uk https://mailman.ic.ac.uk/mailman/listinfo/gridpp-dirac-users
-- Sent from the pit of despair
----------------------------------------------------------- daniela.bauer@imperial.ac.uk HEP Group/Physics Dep Imperial College London, SW7 2BW Tel: +44-(0)20-75947810 http://www.hep.ph.ic.ac.uk/~dbauer/
-- _______________________________________________ Gridpp-Dirac-Users mailing list Gridpp-Dirac-Users@imperial.ac.uk https://mailman.ic.ac.uk/mailman/listinfo/gridpp-dirac-users --
Gridpp-Dirac-Users mailing list Gridpp-Dirac-Users@imperial.ac.uk https://mailman.ic.ac.uk/mailman/listinfo/gridpp-dirac-users
-- Sent from the pit of despair
----------------------------------------------------------- daniela.bauer@imperial.ac.uk HEP Group/Physics Dep Imperial College London, SW7 2BW Tel: +44-(0)20-75947810 http://www.hep.ph.ic.ac.uk/~dbauer/
-- _______________________________________________ Gridpp-Dirac-Users mailing list Gridpp-Dirac-Users@imperial.ac.uk https://mailman.ic.ac.uk/mailman/listinfo/gridpp-dirac-users
Hi,
wrt cvmfs: I thought it was agreed that LSST uses: /cvmfs/lsst.opensciencegrid.org Has this changed ?
it hasn't, but LSST is at the beginning and don't have an established infrastructure yet and there are competing repositories. For example ATM the OSG one contains the PhoSim packages the USLSST wants to use for their Data Challenge, while the IN2P3 one contains data management packages the French are working on which also will be needed for the DC so they asked eventually for both. Catalin did a replication work that allows UK sites not to do anything to have the in2p3 repo installed. cheers alessandra On 05/09/2017 15:44, catalin.condurache@stfc.ac.uk wrote:
Hi Daniela,
It might be that the agreement was between LSST users (at some point in the past), however it should not matter which 'lsst' repos are replicated. Anyway in July we tried to enable 'lsst' jobs at RAL and following discussions with few people involved in LSST I started to replicate the lsst.in2p3.fr repo from Lyon. Once this replica was available, the 'lsst' jobs started to succeed at RAL.
So, as I highlighted in my previous message, these inconsistencies should be sorted out internally within LSST, at least in my opinion.
Regards, Catalin
-----Original Message----- From: gridpp-dirac-users-bounces@imperial.ac.uk [mailto:gridpp-dirac-users- bounces@imperial.ac.uk] On Behalf Of Daniela Bauer Sent: 05 September 2017 12:09 To: gridpp-dirac-users@imperial.ac.uk Subject: Re: [Gridpp-Dirac-Users] Sites with git
Hi Catalin,
wrt cvmfs: I thought it was agreed that LSST uses: /cvmfs/lsst.opensciencegrid.org Has this changed ?
regards, Daniela
On 4 September 2017 at 16:38, <catalin.condurache@stfc.ac.uk> wrote:
Hi Darren,
Just a question: why you do not stick with the cvmfs repository? There is ‘lsst.in2p3.fr’ repository (maintained at Lyon and replicated by Stratum1 at RAL) where I’d assume all software should be provisioned from.
Is this code you’re talking about only interesting the UK sites? In this case you may want to have another (UK regional) ‘lsst’ CVMFS repository.
Anyway we at RAL have the policy to install a minimal distribution on our WNs (and git is not part of it), assuming the VOs are provisioning their software via CVMFS.
Regards,
Catalin Condurache
RAL Tier-1 Services
From: gridpp-dirac-users-bounces@imperial.ac.uk [mailto:gridpp-dirac-users-bounces@imperial.ac.uk] On Behalf Of WHITE Darren Sent: 04 September 2017 15:54 To: gridpp-dirac-users@imperial.ac.uk Subject: [Gridpp-Dirac-Users] Sites with git
Hi there,
I’m working with LSST folks in Southampton who have code that is updated quite regularly. A while ago, I tested to see whether git could clone at time of execution (I cannot remember which site this was at), and it was a success. However, RAL does not have git installed, as similar code fails at that site. I’ve asked this before and didn’t really get a full answer. Is there any way of knowing which sites have which software installed? It would save new users a lot of pain when trying to get up and running, or even just testing.
Cheers
Darren
---------------------------------------------------------------------- --- Dr Darren J. White Applications Developer EPCC, University of Edinburgh
Email: d.white@epcc.ed.ac.uk Telephone: +44 (0) 131 6513415
Room 2411, James Clerk Maxwell Building, Kings Buildings, Mayfield Road, Edinburgh EH9 3JZ, United Kingdom ---------------------------------------------------------------------- ---
-- _______________________________________________ Gridpp-Dirac-Users mailing list Gridpp-Dirac-Users@imperial.ac.uk https://mailman.ic.ac.uk/mailman/listinfo/gridpp-dirac-users
-- Sent from the pit of despair
----------------------------------------------------------- daniela.bauer@imperial.ac.uk HEP Group/Physics Dep Imperial College London, SW7 2BW Tel: +44-(0)20-75947810 http://www.hep.ph.ic.ac.uk/~dbauer/
-- _______________________________________________ Gridpp-Dirac-Users mailing list Gridpp-Dirac-Users@imperial.ac.uk https://mailman.ic.ac.uk/mailman/listinfo/gridpp-dirac-users
-- Respect is a rational process. \\// Fatti non foste a viver come bruti, ma per seguir virtute e canoscenza(Dante) For Ur-Fascism, disagreement is treason. (U. Eco) But but but her emails... covfefe!
participants (6)
-
Alessandra Forti
-
catalin.condurache＠stfc.ac.uk
-
Daniela Bauer
-
Robert Currie
-
Steve Lloyd
-
WHITE Darren