Custom Query (101 matches)

The jobarchived process seems to consume large amount of memory on an Xeon 65 bits with Linux (CentOS 4.5).

It uses 1Gb after an execution of 2 days.

My /etc/jobarchived.conf :

[DEFAULT]

DAEMONIZE : 1

DEBUG_LEVEL : 0

USE_SYSLOG : 1

SYSLOG_LEVEL : 0

SYSLOG_FACILITY : DAEMON

GMETAD_CONF : /etc/gmetad.conf

ARCHIVE_XMLSOURCE : localhost:8651

ARCHIVE_DATASOURCES : "mycluster"

ARCHIVE_HOURS_PER_RRD : 12

ARCHIVE_EXCLUDE_METRICS : ".*Temp.*", ".*RPM.*", ".*Version.*", ".*Tag$", "boott ime", "gexec", "os.*", "machine_type"

ARCHIVE_PATH : /var/lib/jobarch

JOB_SQL_DBASE : localhost/jobarchive

JOB_TIMEOUT : 168

RRDTOOL : /usr/bin/rrdtool

Question about setting up job archive. If I have jobmond.py running on multiple clusters, but the data is only viewable through the web on a central repository web site, do I have to set up the archive database on each of the clusters or just on the central repository server? Also where would I run the two .py scripts - on each cluster or just the central one?

Thanks. Jeff

Thanks to Stijn De Weirdt from University Gent for discovering this:

Add support for Torque job arrays to Job Monarch.

A job array is currently displayed as only 1 job and causes massive memory usage by jobmond