Changes between Version 6 and Version 7 of developers-meeting-20100623

06/23/10 17:54:30 (11 years ago)



  • developers-meeting-20100623

    v6 v7  
    5050Direct 1.9.8-nn and skip 1.9.7 for now. 
     52Current testing 1.9.7 HEAD.  Tigran to finish off testing 1.9.7 then test 1.9.8-HEAD. 
     54Paul: send an enquiry email if someone is running a non-golden on production for testing purposes. 
     56Next Tuesday: ask what they are testing. 
     58Paul: send email to ask what people are testing in Tier-1 centres. 
     60Oleg tried 1.9.6-? because he wanted to try WebDAV, but the webdav clients were broken. 
    5462= Trunk activity = 
    5866== gPlazma == 
     68Come back to discussion on what would be considered condition for one particular plugin within a set of plugins. 
     70When we do verification of correctness. 
     72Consider absence of exception ==> success, 
     74presence of exception ==> failure, 
     76Propose we add a special kind of plugin: or strategy.  Gets login reply as an argument and can fail the login before we reply. 
     78===  Root and home === 
     80These appear in two places: commons and gPlazma. 
     82The ones in common cannot be used in gPlazma, so should be deleted. 
     84=== gPlazma class === 
     86Please review Timur's patch. 
     88== SRM deadlocks == 
     90Timur not had a chance to look at this yet. 
     92Seems to affect 1.9.6 ... 1.9.9 and Trunk. 
     94Not a problem with 1.9.5 because that branch doesn't have synchronisation there --> comes from Terracotta patches. 
     96Timur to look into this. 
    6098= Debian packages = 
    62 Thomas, 
     100Do we want to have "real" Debian packages. 
     102Proper dependency management.   
     104Owen: when we put dependencies into RPM we got many problems with sites. 
     106deb-config-common: for example, automatically set up database.  dCacheConfigure written to make this easy. 
     108How best to split dCache?  Do we want dcache-pool, dcache-srm, dcache-?? 
     110In the past it was dCache-core, dCache-srm, etc.  But now everything depends on globas code, so splitting it would only be config. files. 
     112How complex is the process? 
     114gLite ported to Debian?  No, not completely yet. 
     116Tigran: if you do that, fine with me. 
     118Thomas: I've done this before, so it shouldn't be too much effort. 
     120One idea is to have packages that do configuration for a particular usage: dcache-pool would configure dCache for a pool. 
     122Maybe it doesn't make sense if no one will use. 
     124Please are increasingly using Debian for workstation / laptop machines.  This may be the place where people would be testing dCache for the first time. 
     126Thomas to do a proof-of-concept version. 
    64128= S2 = 
    66 Owen 
     130Three methods put in the mailing list:  AbortRequest, GetSpaceMetadata, GetSpaceTokens. 
    68 = Favour = 
     132After Tigrans patch (on the last two).  It seems that these look like bugs in the test-suite. 
    70 Patrick 
     134Tigrans patch for AbortRequest fixes ... 
     136Timur will look at material that  
     138SRM returns SRM_SUCCESS.  Test expected SRM_ABORTED, so the test fails.  Spec says SRM_SUCCESS is the correct response, so the test is wrong. 
     140Owen is testing 1.9.5-19 
     142Ron reported that SRM reservations failed for non-VOMS proxy (RT ticket #).  Could you try without a VOMS-proxy? 
     144Yesterday, Gerard confirms that 1.9.5-20rc2 works for him. 
     146Plea to Owen: please check whether one can reserve space with non-VOMS proxy. 
     148Patch on review to fix problem reported by  
     150Latest official S2 as RPM from Flavia (2 years old). 
     152Old tar-ball, which is installed on lucan, where the tests pass. 
     154The main point is that no one is maintaining S2.  Officially supported by CERN/WLCG, but Patrick needs to confirm that this is still the case. 
     156= A favour = 
     160There is a conversation between Evano and Xavia on "one pool running amok".  They're talking about setting the number of ?? to a ridiculous high number. 
     162Paul's description on Chimera's filesystem loop. Please tidy up and email this to dCache-forum. 
    72164= dCache configure = 
    74 Owen 
     166Just to update people: there's a new wrapper around dCache configure so we can write modules to configure dCache in Python ^ BASH. 
     168A little bit annoying that commenting out and uncommenting services in the layout file is awkward. 
     170Owen: propose Yaml or adding a single keyword to enable/disable a service. 
    76172= Golden Release #2 = 
    78 Paul 
     176= Issues from [wiki:developers-meeting-20100622 Add link to yesterday's Tier-1 meeting] = 
    81 = Issues from [wiki:developers-meeting-20100622 Add link to yesterday's Tier-1 meeting] = 
    83180= Outstanding RT Tickets =