condor_config settings

Global Configuration

The following settings go into condor_config on all hosts in the pool. Currently, it is extremely liberal in what hosts it allows access from, which will change in the future.

CONDOR_HOST = condor.chem.indiana.edu
CONDOR_ADMIN = superadmin@indiana.edu
UID_DOMAIN = $(FULL_HOSTNAME)
FILESYSTEM_DOMAIN = $(FULL_HOSTNAME)
FLOCK_TO = condor.chem.indiana.edu
FLOCK_FROM = condor-n*.chem.indiana.edu
HOSTALLOW_READ = condor-n*.chem.indiana.edu
HOSTALLOW_WRITE = condor-n*.chem.indiana.edu

Local Configuration (testing)

The following can go into condor_config.local on each host during the testing phase. It allows for all jobs to run uninterrupted. Only the first node really needs to have a local config file.

START = LoadAvg < 0.2
SUSPEND = False
PREEMPT = False
VACATE = False
CONTINUE = True
KILL = False

In /usr/local/condor/local.condorhostname-n1/condor_config.local be sure to put the following:

DAEMON_LIST = COLLECTOR, MASTER, NEGOTIATOR, SCHEDD

You only want to start the COLLECTOR on the first controller node - none of the others.

Local Configuration (production)

We are still working out what kind of start and finish policies to use in production. More information can be seen in 3.6: Startd Policy in the Condor manual.

Most likely we will go with the default of After 15 minutes of no console activity and low (<30%) CPU activity.