Difference between revisions of "RAL HTCondor Multicore Jobs Configuration"

Revision as of 08:33, 21 May 2014

Defrag daemon setup

Our current configuration for the defrag daemon is:

DAEMON_LIST = $(DAEMON_LIST) DEFRAG
DEFRAG_INTERVAL = 600
DEFRAG_DRAINING_MACHINES_PER_HOUR = 30.0
DEFRAG_MAX_CONCURRENT_DRAINING = 60
DEFRAG_MAX_WHOLE_MACHINES = 300
DEFRAG_SCHEDULE = graceful

## Allow some defrag configuration to be settable
DEFRAG.SETTABLE_ATTRS_ADMINISTRATOR = DEFRAG_MAX_CONCURRENT_DRAINING,DEFRAG_DRAINING_MACHINES_PER_HOUR,DEFRAG_MAX_WHOLE_MACHINES
ENABLE_RUNTIME_CONFIG = TRUE

## Which machines are more desirable to drain
DEFRAG_RANK = ifThenElse(Cpus >= 8, -10, (TotalCpus - Cpus)/(8.0 - Cpus))

# Definition of a "whole" machine:
# - anything with 8 cores (since multicore jobs only need 8 cores, don't need to drain whole machines with > 8 cores)
# - must be configured to actually start new jobs (otherwise machines which are deliberately being drained will be included)
DEFRAG_WHOLE_MACHINE_EXPR = ((Cpus == TotalCpus) || (Cpus >= 8)) && StartJobs =?= True

# Decide which machines to drain
# - must not be cloud machines
# - must be healthy
# - must be configured to actually start new jobs
DEFRAG_REQUIREMENTS = PartitionableSlot && Offline =!= True && RalCluster =!= "wn-cloud" && StartJobs =?= True && NODE_IS_HEALTHY =?= True

## Logs
MAX_DEFRAG_LOG = 104857600
MAX_NUM_DEFRAG_LOG = 10

Adjusting draining based on demand

The following script is run as a cron every 20 minutes.

#!/bin/bash
# Change condor_defrag daemon parameters depending on how many idle and running multicore jobs
# there are.

function setDefrag () {
   # Get the address of the defrag daemon
   defrag_address=$(condor_status -any -autoformat MyAddress -constraint 'MyType =?= "Defrag"')

   # Log
   echo "Setting DEFRAG_MAX_CONCURRENT_DRAINING=$3, DEFRAG_DRAINING_MACHINES_PER_HOUR=$4, DEFRAG_MAX_WHOLE_MACHINES=$5 (idle multicore=$1, running multicore=$2)"

   # Set configuration
   /usr/bin/condor_config_val -address "$defrag_address" -rset "DEFRAG_MAX_CONCURRENT_DRAINING = $3" >& /dev/null
   /usr/bin/condor_config_val -address "$defrag_address" -rset "DEFRAG_DRAINING_MACHINES_PER_HOUR = $4" >& /dev/null
   /usr/bin/condor_config_val -address "$defrag_address" -rset "DEFRAG_MAX_WHOLE_MACHINES = $5" >& /dev/null
   /usr/sbin/condor_reconfig -daemon defrag >& /dev/null
}

# Get total number of idle multicore jobs
idle_jobs=$(condor_q -global -constraint 'RequestCpus == 8 && JobStatus == 1' -autoformat ClusterId | wc -l)

# Get total numbe of running multicore jobs
running_jobs=$(condor_q -global -constraint 'RequestCpus == 8 && JobStatus == 2' -autoformat ClusterId | wc -l)

if [ $idle_jobs -gt 20 ] && [ $running_jobs -lt 190 ]
then
   setDefrag $idle_jobs $running_jobs 60 40 300
elif [ $idle_jobs -gt 20 ] && [ $running_jobs -gt 190 ]
then
   setDefrag $idle_jobs $running_jobs 8 8 300
else
   setDefrag $idle_jobs $running_jobs 1 1 4
fi