Minutes of the GridPP Storage Phone conf 10 Aug 2005 Present: Imperial: Mona, Dave RAL Tier1: Andrew, Jeremy Lancaster: Matt Edinburgh: Phil, Jamie RAL Storage: Owen, Jens (chair+secretary) Apologies: RAL Storage: Jiri 0. Review of actions 4 09/02/2005 Send external URLs to Jens ALL Ongoing -> Unchanged 16 09/03/2005 Check gridftp monitoring with James Casey and Kors Bos Jens Ongoing -> No longer needed See for example http://lxgate24.cern.ch/GRIDVIEW/ and http://lxb2009.cern.ch:8080/R-GMA/BrowserServlet/getQueryForm.do?tableName=GridftpMonitor 27 04/05/2005 Send storage and capacity info to Jamie ALL Ongoing -> Unchanged Note that it's now linked to on the main page: http://www.gridpp.ac.uk/ 8 09/02/2005 Resurrect SRM client API Jens Open -> Now allocated to Jiri - but still low priority 22 09/03/2005 Post perf numbers and ganglia links to list Steve Open -> Shouldn't be needed anymore either, given the extensive monitoring that's being done for SC3. I will close it unless I hear objections. https://uimon.cern.ch/twiki/bin/view/LCG/ServiceChallengeMonitoringInformation 28 18/05/2005 Clarify support for SRM 2.1 in dCache Jens Open -> No news 29 18/05/2005 Clarify VOMS support for dCache Jens Open -> No news 30 15/06/2005 Start here howto - recommended install Owen Open -> Now owned by Greig 31 15/06/2005 BDII howto required. Owen Open -> Some information on web site already. Alessandra and Steve can provide input. 33 13/07/2005 Add HOWTO to GridPP web site Owen Open -> This is an Everything HOWTO. Open. 35 13/07/2005 Find out correct behaviour for what to do on full system Owen Open -> Some discussion was required to pin down the problem. It appears to be that advisoryDeleted files remain in trash ` and have to be deleted by admin. Part of the problem may be to advise users to use storage type correctly, but experiments will require permanent storage also at Tier 2s, and then we have the problem. 1. There seem to be some site-specific issues from last meeting. We should follow up (briefly!) in this meeting, at least to define the actions. Edinburgh - "This has led to an action on Owen to give them some help to avoid clobbering the SAN with excessive IO." -> Problem is with NFS mounted SAN: large UDP cannot pass through firewall, and NFS was persuaded to use TCP instead. Owen says he's got some information in the HOWTO. Edinburgh will do further tests with DPM. Imperial - "excessive open socket connections and want more tests" - -> Mona reported several bugs. 1. Open socket connections grow with file transfers and eventually crash the pool. 2. srmcp (the cmd line client) hangs doing srmPut or srmGet. Need to delete the SRM request from the postgres database. 3. Need to bulk move files from a pool. Derek has written a script to move *a* file, but what if there are 1000 files? 4. BDII problem. IC's admin interface not upgraded to LCG 2.6.0, but everything else is, so there is an error about a missing GlueServiceUniqueID attribute. => Need for information how to upgrade admin interface to 2.6.0. We will need to try to replicate these bugs. Thus, if anyone else recognise or can otherwise replicate 1 or 2 (which look more time consuming to replicate and Jiri is on leave), please inform the list. Ta. Glasgow - DPM dual homing. DPM source. -> Nobody attending from Glasgow. Action Jens: Check whether we can redistribute the source to DPM. J-P told me it should go to a public CVS eventually. Lancaster - about to test dCache dual homing with UKLight. 2. New dCAche release http://www.dcache.org/manuals/experts_docs/rel-dcache-1-6-5.html -> Not tested yet. If you are the first to test it and it works (or doesn't), please let the list know. Looks like one to try for all you dualhome fans (Jiri is our local dualhome fan but is on leave). 3. dCache workshop There will be a dCache workshop at DESY at the end of this month. http://www.dcache.org/downloads/dCache-Workshop.html Alessandra and Mona will go, possibly also Greig and Derek. This reminds me that it is not obvious what the SRM content is in the workshop. Action Jens: Check with Michael and Patrick what is planned about the SRM. 4. Bug tracking Owen managed to get a Savannah project page at CERN. Nee "teir2srmsupport", it is now called https://savannah.cern.ch/projects/srmsupportuk/ Owen signed up most or all of the usual suspects who already have Savannah accounts. Currently, Owen, Jiri, and Jens are admins. There was some discussion about whether there is a communication problem, and, if so, what it is. Mona reported that there is a real need for problems to be resolved and post installation instructions (e.g. testing). Some people in the storage community are speaking directly to the dCache developers, despite reports to the contrary. This is fine, but it is best if the mailing list is cc'ed, or if they summarise to the list, so everybody can benefit. As a general rule for support, Phil proposed elsewhere the following support procedure (slightly edited by me): (1) look at the faq and doc pages first (see below); (2) If problem not answered post a mail to gridpp storage list. (3) If not answered then contact Greig (tier2) Jiri or Owen (tier1+2) (4) Greig/Jiri file bug to the coders bug tracking system and our new Savannah. (5) Well known issues and answers are harvested from GRIDPP-STORAGE and added back into (1) Optionally, (3') If not answered, contact LCG support (or similar), or upstream support (e.g. developers), cc'ing gridpp storage list. 5. Very technical support Is it worth asking Patrick and/or Michael to attend some of our phone conferences? -> Action ALL: send proposed issues for discussion to Jens by end of Friday 12 August. Also, someone (forgot who, sorry) once suggested that we extend the conferences so everyone who wants to discuss technical stuff can hang on after those who don't have left. -> We did that today, except in the middle of the conference, mainly covering Mona's bug reports. While it may have been useful, in the future we should move it to the end so non-dCache experts, or non-DPM experts, can leave. 6. AOB -> Jamie is almost ready to release additional DPM documentation, some for the sysadmin and some for the user. -> Related to that, Glasgow has some wiki pages including one for DPM. GridPP will set up a Wiki and the DPM pages at Glasgow will then move. http://www.physics.gla.ac.uk/gridpp/datamanagement/index.php/DiskPoolManager Andrew reported that Steve had pointed out the need for wikification, so perhaps we can all use the new GridPP wiki. Should be available by next week. -> AHM. We should do something storagey for the GridPP stand, at least a poster. Jens wrote this one for Institute of Physics, High Energy Particle Physics meeting Dublin 21-23 March 2005: http://www.gridpp.ac.uk/posters/march_05/storage_march05.png Action Jens: check with Becki what's planned. Dave very kindly offered his visualisation toolkit plus potentially someone at IC to work with us if we wish to provide live visualisation of something storagey. Which would be a good thing to do because everybody remembers Dave's RB and the little boxes representing jobs. Perhaps it can be integrated with Jamie's monitoring. New Actions ----------- (numbers may be wrong if Owen's added some that I have missed) Number Opened Description Owner Status Closed 37 10/08/2005 Check DPM source availability with J,-P. Jens Open 38 10/08/2005 Send dCache phone conf issues to Jens by 12 Aug 2005 ALL Open 12/08/2005 39 10/08/2005 Check with Becki what's planned for AHM Jens Open 40 10/08/2005 Check with Michael & Patrick SRM plans for DESY workshop Jens Open