Date: Tue, 16 Apr 2002 10:16:35 -0500 (CDT)
From: Gustaaf Brooijmans <gusbroo@fnal.gov>
To: level3 <D0-l3alg@fnal.gov>, D0-l3soft@fnal.gov
Subject: comments on the 4/10 minutes

Hi,

A few additions to the minutes of the 4/10 meeting:

-Events which lead to filtershell timeouts (i.e. for which processing is
aborted after 30 s) will be kept on disk on the L3 farm in raw binary
format.  A little program that wraps this up in an edm event and sends the
event off to some error stream needs to be written, but manpower has not
been identified.

-During the test run, memory consumption went up in steps of about 20 MB,
starting from the 50 MB the program usually uses.  A scan of 3 nodes (out
of 24) showed one executable had grown to 110 MB after 15 minutes of cpu
time.  This looks like a memory leak, or a strange approach to memory
management.

Cheers,

Gustaaf