change the slave configuration, so that the nodes are brought online only on demand

This will:
 * reduce the amount of used memory: currently any given slave.jar
   process takes something between 200 and 350 MB of resident memory,
   which easily goes over the 10 GB mark…
 * in general, reduce the number of running processes on the master,
   which can't possibly be a bad thing
 * make possible to not even start jobs on the slaves if a host goes offline
   for whatever reason
At the same time, starting slave.jar is not such an expensive process,
and the delay in starting the jobs from an offline node is totally
negligible
Signed-off-by: Mattia Rizzolo's avatarMattia Rizzolo <mattia@debian.org>
parent 78e3a51a
......@@ -105,7 +105,9 @@ Process to follow to add a new node to jenkins:
* 'Usage': select "Only build jobs with label expressions matching this node"
* 'Launch method': select "Launch agent via execution of command on the master"
* 'Launch command': `/srv/jenkins/bin/start-slave.sh`
* 'Availability': select "Keep this agent online as much as possible"
* 'Availability': select "Take this agent online when in demand, and offline when idle"
* 'In demand delay': 0 (so that builds will start right away)
* 'Idle delay': 5 (this is an arbitrary amount of time)
The slave setup is done so that the slave.jar program doesn't get run on the remote nodes,
to avoid needing Java available in there.
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment