Skip to content

Federate your cluster

Requirements

  • Two working HTCondor clusters instatiated either with DODAS (this guide) or with plain HELM (this other guide).
  • Working kubectl for the 2 clusters
    • either logging on the k8s master node or having a valid one on you computer

We will call as master the cluster that will distribute the jobs to the second one (slave)

Configure the master cluster

Using the k8s context of the master cluster, edit its schedd configmap:

kubectl edit configmap scheddconfigd

Then add the following lines to the content:

FLOCK_TO = <public address of the slave Collector node>
FLOCK_COLLECTOR_HOSTS = $(FLOCK_TO)
FLOCK_NEGOTIATOR_HOSTS = $(FLOCK_TO)
HOSTALLOW_NEGOTIATOR_SCHEDD = $(COLLECTOR_HOST), $(FLOCK_NEGOTIATOR_HOSTS)

N.B: after the change you have to restart the schedd pod

Configure the master cluster

Using the k8s context of the slave cluster, edit its collector configmap:

kubectl edit configmap ccbconfigd

Then add the following lines to the content:

FLOCK_FROM = <public address of the master schedd>

N.B: after the change you have to restart the collector pod

All done now. Give the system few minutes to adapt and you should be able to see jobs coming to the slave clusters soon.