Hello All,

I'm using NiFi 1.4.0 and was playing with the ExecuteStreamCommand processor to 
submit spark jobs. I wrapped the spark-submit command in a shell script and 
tested it first with ExecuteProcess and everything works as expected, however 
according to my use case I need to use many attributes from the flow before it 
as well as trigger few jobs only if flowfiles with certain attributes are 
received. So, I started considering ExecuteStreamCommand processor and I used 
the same shell script to submit the spark job. The problem is the job gets hung 
somewhere and never finishes, stopping the processor didn't help and I even 
killed the job on my Linux box (with it's PID) and the processor is still 
running. The only way that I could start this hung processor is by restarting. 
It's not making sense why this is not working while ExecuteProcess works 
perfectly. Does anyone have similar problems? Below are my configurations

(Also, the box that I'm running NiFi is an edge node)

NiFi version - 1.4.0 (not HDP)
Spark version -  2.1.1.2.6.2.0-205 (HDP)
Spark-submit client - YARN

Thanks
Karthik

Reply via email to