How to increase spark.worker.timeout?

classic Classic list List threaded Threaded
1 message Options
Reply | Threaded
Open this post in threaded view
|

How to increase spark.worker.timeout?

Jyun-Fan Tsai
Hi
We use standalone deploy master.  When we running jobs we usually see some worker Dead.  To resolve the problem we set "spark.worker.timeout" to 600 seconds in spark-env.sh:

SPARK_JAVA_OPTS+=" -Dspark.worker.timeout=600"

However, we still see such error:

"14/01/09 09:28:27 WARN master.Master: Removing worker-ABC because we got no heartbeat in 60 seconds"

The error message shows that the timeout is still 60 seconds, the default value.  Any advice to set timeout correctly?

The spark version we use is 0.8.1

--
Regards,
Jyun-Fan Tsai