ZEUS Offline Meeting     24. March 2000


Present: Chris Cormack, Massimo Corradi, Adrian Fox-Murphy, Tobias Haas, Marek Kowal, Rainer Mankel, Ingo Martens, Matthew Wing, Sergej Zotkin, Krzysztof Wrona

Code Management

The new software release 2000a has been announced on Monday (Matthew). It contains all modifications before upgrading to the new OS versions SuSE-Linux 6.2 and Solaris 2.7. AFS group space had become short during the build. The zephyr modification which fixes the compression error has gone into a special zephyr release v1998b.4. A corresponding global version will be created as well.

The computer center has increased the quota on the 2 busiest volumes (Rainer). For the time being, we cannot have the repository on a single (~10 GB) volume. The biggest volume size presently covered by the backup system is 2 GB.

Data Processing

Reconstruction of current data is running without problems (Sergej). One file of the 99 positron data is corrupted and has to be reprocessed. This will wait until the 99p copying process is finished, in case there are more files to be redone.

However, stageing of tapes is very slow (about 1 hour wait!). The problem started yesterday, there is a long list of requests distributed roughly equally over experiments ZEUS, H1 and HERMES (Marek & Krzysztof).
 
 

Workgroup Servers & Computing Infrastructure

The new web server and the "ZOW replacement" PC are in the computer center and connected to the network (Ingo). There is the question of operating system: SuSE 6.2 is available on CD's but a bit old, the 6.3 has problems with this installation mode. Ingo will look for a 6.4 release on CDs.

The setup of the AFS test machine has still problems. Knut Woller is aware of it & promised to improve the installation on his server.
 
 

ZARAH

The PC farm (destination "zenith") will be announced for public use (Marek). The monitor queue is also planned to move to zenith starting from 10-April. It may be useful to change the processing scheme such that one waits first for all files of a run to be on disk before initiating processing. (This would delay feedback to data-taking a bit). The data corruption monitor, which also runs on ZARAH is a special case.

The infoseek local search machine, recently available on the DESY web server, has been tailored to the ZARAH & ZEUS web pages. The date-wise reconstruction status plots are now also available on the ZARAH web server.

Copying of the 99p MDSTs had to be paused, it degraded the ZARAH performance because of the stageing and HIPPI network problems (Krzysztof). 320 of ~600 runs are copied. The old MDST2/MDST3 files from 1996 have been deleted from disk.
 
 

ZES

Update of event offsets in the 1996 tag database is finished (Adrian). Copying to the mirrored location took only 1h45m within the doener circumventing the HIPPI interface.

The ZESLIB has to be updated to comply with the new set of ZES versions (2.0...2.3) in the updated set of n-tuples. This will take about 1 week.
 
 
 

Monte Carlo

The output disk of the zandsak had filled up completely & production had stopped because the data cannot be written to tape fast enough because of the robot problem already mentioned (Massimo). It is running again after Marek gave them more disk space. It would be more efficient to use multiple file copy in one request instead of issueing one osmcp command fore each file (Marek).

The modified ZEPHYR with the correct treatment of compressed tables works fine (Massimo). There is a problem with the monitoring jobs, as funnel frequently closes incorrectly at the end of the job. 5 Linux streams are established (4 on Italian, 1 on a Canadian machine).

It was not yet possible to exchange executables between different Linux versions (Chris). It will be investigated if this problem is due to shared libraries.

The whole Monte Carlo production chain is currently ported & tested under SuSE 6.3 (Chris). He will also investigate if the port is backward compatible to older Linux. The "pow" problem reported earlier has been explained in the desy.computing newsgroup.