Pile of Dross

Monday 29th

Bank Holiday

Tuesday 30
(Privilege Holiday, but not for me)

dCache Workshop DESY Day 1

Patrick Fuhrmann

Multiple mover queues
– allows splitting up of queues by indvidual door they use

idle threshold on cost
– use same file replica , so others get older and can be removed

cm set magic on in PoolManager
should fix unbalanced transfers
fixed in 1.6.6

Resiliency Manager
– can’t use with HSM
– allows draining a pool to many other for maintenance

Internal Copy Manager
– In 1.6..5
– instructions in book
– for draining pools

Pnfs using PostGres
– Coming Soon
– To overcome 2GB gdbm, file limit

New pnfs
– optimise dcache access to pnf over nfs interface

VOMS
- DOne buit being tested
- Not yet integrated with GSI-DCAP
- Not in next release

Timur Perelmutov

Lambda Station for managing OPN

Cleared out testfiles from nfs60-63 in preparation for rejigging pools.
Tried transfers to Glasgow using fts, but seem to be stuck
Started SC3 outcomes document
Investigated Durham UI problem, they can upload via gridftp to the RB manually.

Added grid.niif.hu to RGMA registry, check fraunhofer.de were still okay with their changed mon box
Restarted myproxy
TB dropping different files in /tmp onto lcgui02, poked when he asked about connections dropping due to network glitch we had this morning
Started RPC.rquotad on dcache head nodes dcache.gridpp and jrad1dch01.gridpp to stop irritating messages from quota check on login to lcgui01.
RB saga continues – someone in Durham can’t submit jobs to it using their UI, but can using ours. I can using ours and lxplus @CERN, UCL-Hep reported that it wasn’t working then said it had started working again
Helping CK use srmcp to test against dpm on decv06
Moved csfnfs56 to new ipaddress added dcache pool to CMS

Attended SC meeting,
Checked RB working,
Checked that VO_SC3_* was added to the job environment, need to upload and install new edgonCSF rpm to complete process

Needed to get new version of Config::IniFiles rpm – one I had was built againsst wrong perl version, added script to kickstart to remove unwanted rpms from syslogger

RB acting up – jobs go into Running state, skipping Scheduled state and never reach Done state, I suspect its related to the disk full we had on the rb on Friday

-Technical Discussion Group – CPU Technology: Dual Cores
-Installed dcache on csfnfs56, need to change IP address, add entry into dcache exports, add into CMS-pgroup
-270, adjusted partitions, added perl-Config-IniFiles to yum repository to fix yumit dependency
-PhEDEx on lcgui02 using all inodes in /tmp, mailed TB
-dCache PinManager hung again this morning, restarted

Debugging srmcp problem between castorgrid and RAL (CMS problem in SC3), looks like a weird name caching issue, possibly the java libraries used by dCache are implementing an older version of globus and occasionally we get lucky due to DNS caching?
Turned 266 into a script and did repeated installs of 270 to iron out bugs, pretty much done but the partition layout could do with a touch more work.

Logrotate of 266 using wrong pid filename for syslog-ng, meant logs were rotated but syslog-ng not restarted to reopen files.
Updated all dcacheSetup files on dCache disk servers to new java path and restarted.
Booked flights to Hamburg

Requested new certificates for the two jra1 box and the new CMS dCache disk. Phone support to CB, who’s installing/configuring dCache. Helpdesk box issue /tmp filled with far too many files – stopping e-mail getting through as RT couldn’t create tmp file, cleared out, need to install tmpwatch or something. Request to do testing to dpm at Glasgow from Jeremy, do some srmcp’s tomorrow, but FTS server is at risk ths week and people at Glasgow more available for monitoring next week, so leave mass transfers until then.
New syslog box is configured with DHCP address, which is wrong, should only access DHCP on re-install, need to look at installation scripts.
Edited dCache PoolManager to add settings for H1, fixed small mistake made by ST when creating pools on nfs58

In line with all corporate guidelines on rebranding, after the merger comes the rename. So this blog is now officially a Pile of Dross. Next comes a new logo.

Sad thing is I’m not kidding either. Anyone have any ideas of things that evoke the spirit or nature of dross?