6

I have a spark streaming application that was running absolutely fine until yesterday and all of a sudden running into these warnings. I have the same environment and using the same code. Here are the warnings:

05/09 17:13:03 INFO ExecutorAllocationManager: Requesting 16 new executors because tasks are backlogged (new desired total will be 31) 16/05/09 17:13:03 INFO ExecutorAllocationManager: Requesting 19 new executors because tasks are backlogged (new desired total will be 50)

16/05/09 17:13:12 WARN YarnScheduler: Initial job has not accepted any resources; check your cluster UI to ensure that workers are registered and have sufficient resources

16/05/09 17:13:27 WARN YarnScheduler: Initial job has not accepted any resources; check your cluster UI to ensure that workers are registered and have sufficient resources

I am using apache spark 1.6 on cloudera 5.5. quickstart VM. There are no applications running on the cluster to consume available resources.

Is there any config to be made?

Thanks!

Useme Alehosaini
  • 2,998
  • 6
  • 18
  • 26
Abhi
  • 1,153
  • 1
  • 23
  • 38
  • Possible duplicate of [TaskSchedulerImpl: Initial job has not accepted any resources;](http://stackoverflow.com/questions/29469462/taskschedulerimpl-initial-job-has-not-accepted-any-resources). If that question doesn't solve your problem, please edit to explain what you tried and what happened. – jpmc26 May 10 '16 at 00:23
  • Have you enabled dynamic executor allocation? – Joe Widen May 10 '16 at 00:54
  • Yes, I added following properties yarn.nodemanager.aux-services spark_shuffle yarn.nodemanager.aux-services.spark_shuffle.class org.apache.spark.network.yarn.YarnShuffleService in spark-env.sh – Abhi May 10 '16 at 00:56
  • Do I need any additional parameters to enable Dynamic allocation? I also added SPARK_WORKER_MEMORY, CORES in the spark-env but that did not help!! – Abhi May 10 '16 at 01:00
  • @Abhi, you found any solution to the problem? – lu5er Jan 14 '19 at 08:41

1 Answers1

0

The quickstart vm was designed to give a quick taste of the functionality, and is likely configured to 'fit in the box' rather than for long term stability.

In general, my recommendation for problems with sandboxes and quickstarts is to just set up a new one if it is needed.

In this specific case, the quickstart is for CDH, for which the successor (CDP) is already out.

CDP, the Cloudera Data Platform is the successor to both CDH and HDP.

If you want to check out the latest version, there is a trial which should serve for most purposes that you may wanted to use the quickstart. This can be downloaded here.

Full disclosure: I am an employee of Cloudera, the company behind both CDH and CDP.

Dennis Jaheruddin
  • 21,208
  • 8
  • 66
  • 122