Manchester Storage Tests
Contents
Slashgrid
07/08/2007
I've inserted the install rule for slashgrid clients in cfengine. Tomorrow the rpms hould be installed on all the WN. We need to install the servers to on some of the nodes in the cluster but for now we can start to play with the nodes Andrew has already installed on the niels machines.
08/08/2007
The rpms are installed, we started the slashgrid service which I didn't get I had to do yesterday. We are wiating for niels016-18 new certificate because there is a small problem in how the certificates are handled in SL4 and SL3 by curl. However in insecure mode it works
htll --no-verify https://niels016.tier2.hep.manchester.ac.uk/aforti/ 5 Aug 8 16:16 pippo
09/08/2007
curl has been updated together with the new gridsite rpms that use the new version. Unfortunately the UI cannot be updated without messing around because other glite tools depend on the previous version of curl. gridsite and curl have been upgraded, hopefully, on all the WN via cfengine.
Roger got permission denied because the wrong DN was in the gacl lists, however this has been solved and he can now read/write on the slashgrid testbed.
Andrew submitted a test job to the cluster which was successful. Next step is to install slashgrid on few WN rather than on the shuttles to have a uniform testbed.
Xrootd
Dcache
09/08/2007
See progress on RT ticket
https://rt.tier2.hep.manchester.ac.uk/Ticket/Display.html?id=382
Application
Roger john barlow 10:09, 9 Aug 2007 (BST)
I am using the BaBar Run 1 Off peak Tau User ntuples, which is a small but significant amount of data. 1.5 million events in 29 files each around 80 MB. More data will be added.
As well as the 3 sensible methods, I have an afs version which reads the files from a server in the physics machine room.
The current status is
afs Wall time 1203 sec, CPU time 357 sec
xrootd Wall time 556 sec, CPU time 340 sec
dcache fail to read files from dcache
slashgrid havn't managed to copy files into slashgrid
I will repeat the afs and xrootd jobs and see how much the numbers change on different runs
