Why spark executor receives SIGTERM?

Question

I am using Spark API (Spark core API, not Stream,SQL etc.) I often see this kind of error in spark dumped log: Spark environment: 1.3.1 yarn-client

ERROR executor.CoarseGrainedExecutorBackend: RECEIVED SIGNAL 15: SIGTERM

1. Who triggers SIGTERM. YARN,Spark or myself?
2. Will this signal terminate Spark Executor? If not, wow will it affect spark program.

I do press Ctrl+c, but that whould be SIGINT. If YARN kill executor, that would be SIGKILL.

Answer 1

You will likely find the reason in yarn logs. If you activated log aggregation, you can type

yarn logs -applicationId [app_id]

and lookup for exceptions.