Search code examples
apache-sparkh2osparkling-water

getOrCreate deployment failing randomly

When attempting to call H2OContext.getOrCreate with a valid SparkContext, randomly we keep seeing failures to deploy:

17/04/21 17:21:32 ERROR TaskSchedulerImpl: Lost executor 0 on 172.17.0.4: Remote RPC client disassociated. Likely due to containers exceeding thresholds, or network issues. Check driver logs for WARN messages.
17/04/21 17:21:38 ERROR LiveListenerBus: Listener ExecutorAddNotSupportedListener threw an exception
java.lang.IllegalArgumentException: Executor without H2O instance discovered, killing the cloud!
    at org.apache.spark.listeners.ExecutorAddNotSupportedListener.onExecutorAdded(H2OSparkListener.scala:27)
    at org.apache.spark.scheduler.SparkListenerBus$class.doPostEvent(SparkListenerBus.scala:61)
    at org.apache.spark.scheduler.LiveListenerBus.doPostEvent(LiveListenerBus.scala:36)
    at org.apache.spark.scheduler.LiveListenerBus.doPostEvent(LiveListenerBus.scala:36)
    at org.apache.spark.util.ListenerBus$class.postToAll(ListenerBus.scala:63)
    at org.apache.spark.scheduler.LiveListenerBus.postToAll(LiveListenerBus.scala:36)
    at org.apache.spark.scheduler.LiveListenerBus$$anon$1$$anonfun$run$1$$anonfun$apply$mcV$sp$1.apply$mcV$sp(LiveListenerBus.scala:94)
    at org.apache.spark.scheduler.LiveListenerBus$$anon$1$$anonfun$run$1$$anonfun$apply$mcV$sp$1.apply(LiveListenerBus.scala:79)
    at org.apache.spark.scheduler.LiveListenerBus$$anon$1$$anonfun$run$1$$anonfun$apply$mcV$sp$1.apply(LiveListenerBus.scala:79)
    at scala.util.DynamicVariable.withValue(DynamicVariable.scala:58)
    at org.apache.spark.scheduler.LiveListenerBus$$anon$1$$anonfun$run$1.apply$mcV$sp(LiveListenerBus.scala:78)
    at org.apache.spark.util.Utils$.tryOrStopSparkContext(Utils.scala:1252)
    at org.apache.spark.scheduler.LiveListenerBus$$anon$1.run(LiveListenerBus.scala:77)

The H2OContext.getOrCreate causes the error:

Context.spark_session = SparkSession.builder.getOrCreate()
Context.h2o_context = H2OContext.getOrCreate(Context.spark_session)

Any thoughts from the H2O Crew?

Solution

  • this is a known behaviour of Sparkling Water internal backend at the moment. To avoid this, the external Sparkling Water backend can be used. More information about this can be found here https://github.com/h2oai/sparkling-water/blob/master/doc/backends.md

    I'm currently working on this JIRA which should eliminate the behaviour above as well. It's work in progress, this JIRA https://0xdata.atlassian.net/browse/SW-369 can be tracked to get the status of the task.