NDPF TODO List

Enable backup on hooimijt

Het startup script staat er wel (/etc/init.d/adsm), maar zodra did gestart wordt gaat de load naar oneindig omdat de tsmc eerst een expiry van de /export/cache gaat doen. Dat duurt volgens mij een paar uur waarin de hele machine unresponsive is (het tsmc process zit dan al die tijd in een "D" state). Daar kan de rest van de farm dus niet tegen.

Eerst even bij SARA de oude backup laten moven, en dan nog eens proberen op een rustig moment? (niet weggooien! :-)

Upgrade Torque

New version (2.0) of Torque is out. This one includes TMPDIR patch IIRC. Time to upgrade. Misschien moeten we JT de Torque server laten vernietigen. De laatste RPMs hiervoor (van SteveT met de TMPDIR patches &c) staan nu op

 http://hepunx.rl.ac.uk/~traylens/rpms/torque

maar SteveT waarschuwde nog wel: "The newer ones have had less than a day of testing so be warned."

VOor SteveT's versie zijn de startup scripts verschillend (niet meer een enkele "/etc/init.d/pbs", maar een setje "pbs_{mom,sched,sever}". Daarop moeten de startup configs in de Quattor config wel worden aangepast (nu staat daar nog een manual override in de local/ config). Zie voor de sources ook:

 http://www.gridpp.rl.ac.uk/viewcvs/viewcvs.cgi/torque/

Check pool accounts

Apparently things can go wrong if we have e.g.

dteamsm01

and

dteam001

as pool accounts for 'dteamsm' and 'dteam' ... because 'dteamsm01' is a valid pool account for .dteam. Check and repair.

VOBOX installation

LCG now has an official "VOBOX" profile. We need to install one of these, evaluate it and based on what we see, maybe install one more.

Fixes needed for information published to BDII

There are a number of new attributes in the GlueVOView blocks that are not yet being published, like the software dir and data dir. This is going to require some serious quattor work and is not a task to be taken lightly.

Also there is a warning in the GIIS monitor that our publishing of teras.sara.nl as a close SE is failing some sanity checks. This last one may be a fault in the test, someone needs to looks carefully at this.

Update of Resource Broker

See message on LCG-ROLLOUT

R-GMA GIN Update

See message on LCG-ROLLOUT

LFC Updates

The DPM updates mentioned in the link below have already been installed. I don't know about the others. Someone please check and when it's all done, please remove this item! Link is these updates

FTS client

Once we get a DPM we can start doing service challenge stuff. For this we will need an FTS client. Here is some info from Gavin McCance:

Clients:

Either manually: [ https://uimon.cern.ch/twiki/bin/view/LCG/FtsClientInstall13 Link]

or with whatever comes with LCG-2.6.0 (yaim?)

The difficulty is the FTS server URL (i.e. "where does my client commandline
tool point to"). Currently, we have not finished our integration with the
BDII (or indeed any information system) so we use a file in the local
filesystem. An XML file as well.

[ https://uimon.cern.ch/twiki/bin/view/LCG/FtsClientInstall13 Link]

describes the format of this services.xml file - but it really only suitable
to point to client commandline tool to *one* FTS server. It is possible to
point it to multiple servers and use the "-s" option of the commandlinme
tool to select between them, but it's a hack that will have to go away once
we do it properly with the BDII, so I wouldn't rely on it.

Clients at NIKHEF: point to the SARA server.

Clients at SARA: point to either CERN T0 server or the local SARA server. If
the same client on a given machine needs to point to both servers (the CERN
one to manage the T0-T1 and the SARA on to manage the T1-T2) then use the
"-s" option with two entries in the service.xml file.

hope this helps,
cheers,
gav

Cleanpool Script

The bad version of the cleanpool script is in /export/perm/adm/bin. It needs to have the apostrophes removed, be put back into test mode. This means to replace the "rm -fr" and "rmdir" stuff with "ls -l" and see what it thinks it will do. If this all looks OK, it needs to be tested with the "rm" commands put back, but in a sandboxed environment, say copy a pool directory somewhere and try there.

hooibaal OS upgrade

"hooibaal" is nu de laatste RH73 machine :-)

ganglia monitoring

ganglia monitoring op tbn06 is zo goed als dood. Tijd voor een nieuwe machine die ganglia/syslog/auditing op zich gaat nemen. Dat moet maar een 'nieuwe' bak zijn (en geen oude pizza0 class doos).

more LCG updates

Two updates have been made available to LCG-2_6_0.

edg-utils-system-1.8.0-1_sl3.noarch.rpm
edg-utils-user-1.8.0-1_sl3.noarch.rpm

They are in the apt/yum updates folder

[ http://grid-deployment.web.cern.ch/grid-deployment/gis/apt/LCG-2_6_0/sl3/en/i386/RPMS.lcg_sl3.updates/ link ]

Install GEANT4 VO

Contact Patricia Mendez Lorenzo