You are viewing a plain text version of this content. The canonical link for it is here.
Posted to user@spark.apache.org by Kalpit Shah <sh...@gmail.com> on 2014/04/12 18:31:35 UTC

Re: Changing number of workers for benchmarking purposes

In spark release 0.7.1, I added support for running multiple worker processes
on a single slave machine. I built it for performance testing multiple
workers on a single machine in standalone mode.

Set the following in conf/spark-env.sh and bounce your cluster :

export SPARK_WORKER_INSTANCES=3

This will create 3 worker processes on every slave machine.

- Kalpit



--
View this message in context: http://apache-spark-user-list.1001560.n3.nabble.com/Changing-number-of-workers-for-benchmarking-purposes-tp2606p4169.html
Sent from the Apache Spark User List mailing list archive at Nabble.com.

Re: Changing number of workers for benchmarking purposes

Posted by Kalpit Shah <sh...@gmail.com>.
I think "SPARK_WORKER_INSTANCES" is deprecated.

This should work: "export SPARK_EXECUTOR_INSTANCES=2"



--
View this message in context: http://apache-spark-user-list.1001560.n3.nabble.com/Changing-number-of-workers-for-benchmarking-purposes-tp2606p26491.html
Sent from the Apache Spark User List mailing list archive at Nabble.com.

---------------------------------------------------------------------
To unsubscribe, e-mail: user-unsubscribe@spark.apache.org
For additional commands, e-mail: user-help@spark.apache.org

Re: Changing number of workers for benchmarking purposes

Posted by lisak <li...@gmail.com>.
Hey,

I'm using this setup in a single m4.4xlarge node in order to utilize it : 

https://github.com/gettyimages/docker-spark/blob/master/docker-compose.yml

but setting : 

    SPARK_WORKER_INSTANCES: 2
    SPARK_WORKER_CORES: 2  

still creates only one worker. One JVM process that utilizes up to 200% CPU

Do I have to also start 2 org.apache.spark.deploy.worker.Worker instances
explicitly ? Or should I rather stick with : 

    SPARK_WORKER_INSTANCES: 1
    SPARK_WORKER_CORES: 8

?  



--
View this message in context: http://apache-spark-user-list.1001560.n3.nabble.com/Changing-number-of-workers-for-benchmarking-purposes-tp2606p26488.html
Sent from the Apache Spark User List mailing list archive at Nabble.com.

---------------------------------------------------------------------
To unsubscribe, e-mail: user-unsubscribe@spark.apache.org
For additional commands, e-mail: user-help@spark.apache.org