Summary PHYSGI01 Meeting 5/23/94

B. Gibbard

Next Meeting:

Next meeting will be Monday, June 6, 11:30 - 12:00, in Room 2-160. Notice that we are skipping a week since Monday, May 30, is a laboratory holiday. A meeting could be arranged for Tuesday, June 1, if there is need but otherwise the plan described below for this week will be in effect for the following week as well.

CPU Intensive Users:

Groups who announced that they would be doing production processing during the week of May 23 are:

E891 (Saulys et al.) - ~2 CPU intensive Monte Carlo job.
E852 (Weygend et al.) - ~6 CPU intensive jobs (AGS dependent) and a long running non-CPU intensive dispatch process.
E881 (Gushue et al.) - ~6 CPU intensive jobs (AGS dependent).

All experiments which anticipate running more than an single occasional production job during a given week are required to attend the scheduling meeting. If that is not possible they should notify Bruce Gibbard (gibbard@bnl.gov) and Tom Throwe (throwe@rsgi02.rhic.gov) of their needs by Email prior to the Monday 11:30 meeting.

Plan for week beginning 5/23/94:

The rules will be the same as for last week:

A production job is defined as a job which would accrue an hour or more of CPU time in 8 hours of running in the absence of contention.
Each experiment is still allowed to run up to 6 concurrent "production" jobs but all except 1 of those six must be run "niced" to "1".

Other issues:

Experiment E881 reported a situation over the weekend in which the script which automatically deletes files on the scratch disk down to 50% free when that disk becomes more than 75% full, apparently decided to delete all files on the scratch disk. A check will be made to see if any changes have been made in this script recently.

There are problems with the 8 mm tape drives on PHYSGI01. The problems seem to be associated with doing very long backups. They have prevented the recent successful completion of full backups and seem to affect about 50% of the incremental backups. Since these incrementals are incremental since the last successful full backup they are themselves rather long. E852 reports the same problem when trying to backup large data sets. The onset of the problem seems to coincide with the move to IRIX 5.2. This problem will be investigate via SGI and TTI.

Users of PHYSGI01 are still strongly encourage to set a mail forward from their working account on PHYSGI01 to an account on a machine where they frequently check their mail so that they can be easily contacted.