xrootd & objectids
Alastair presented some slides about a problem they've encountered with accessing object via xrootd because xrootd expects a // after the host. The code will be modified to accomodate this. Experiments will need to add an extra / in their root URLs. ATLAS and LHCb will not have a problem adding it in AGIS and the LHCb equivalent, CMS will need to use the TFC (Trivial File Catalog) for this. Since the TFC seems more dynamic it could be use to test things before changing the code and to cover the rolling out period.
the proxy delegation documentation has been finally released to dcache it is still in googledoc format so people can comment. Al Rossi (FNAL) is working on the dcache implementation. The google doc has been now distributed to the group so other people can comment as well. We also discussed rolling out delegation, particularly at sites that are based on native xrootd. For DPM it should be matter of upgrading the rpm and then use the new version of puppet.
We all agree the smoke tests are a valuable addition to the test we do and allow the sites to run by themselves and possibly debug silently. We also agreed to run the http smoke tests in a cron and email the results once a day to the group as a prod to fix problems on top of the heat map. We will review if the cron is useful at the next meeting. Paul asked if we also want checksum tests added to it, we all agreed it was a good idea. Fabrizio is waiting for a go ahead to upgrade the DPM RC instance which is still in legacy mode and doesn't support checksum. Alessandra says we should aim at the final configuration and perhaps it is time to set a timeline to achieve this so that sites know what they need to do when. We have 2 extra sites among which DESY production which will be used for the stress tests instead of the testbed. There were problems with LRZ and AGLT2 but they have been solved.
We discussed the heat map which doesn't look that good, but the problem here seems we are victims of our success and we have reached the limit of sites kibana can cope with so it will not display all we have. Thomas will add the stress tests link to the twiki. This stress tests heat map should now be in the production instance of kibana and be accessible by everyone