[
https://issues.apache.org/jira/browse/SPARK-6847?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15008993#comment-15008993
]
Glyton Camilleri commented on SPARK-6847:
-----------------------------------------
Hi Yunjie,
Whether 50 seconds is good or not as a checkpoint interval depends largely on
the time-window the stream is acting on; so if the stream is set to execute
jobs every 10 seconds, then 50 seconds could be fine.
In my example code, {{certainConditionsAreMet}} was just a place-holder: the
conditions met were application-specific in that case; so in other words, there
were conditions under which we would perform the side-effect on the stream,
which in our case ((1) above) was saving the contents of the stream to HDFS. So
the fix looked something like this:
{code}
def isTimeToSave: Boolean = ... // this function decides whether it's time to
store the contents of the stream to HDFS
def saveData[A](stream: DStream[A]) = if (isTimeToSave) stream.foreachRDD {
... // write data in HDFS
} else stream.foreachRDD {
_.foreachPartition { _ => () } // just do nothing
}
{code}
The {{else}} part is what i'm referring to above.
> Stack overflow on updateStateByKey which followed by a dstream with
> checkpoint set
> ----------------------------------------------------------------------------------
>
> Key: SPARK-6847
> URL: https://issues.apache.org/jira/browse/SPARK-6847
> Project: Spark
> Issue Type: Bug
> Components: Streaming
> Affects Versions: 1.3.0
> Reporter: Jack Hu
> Labels: StackOverflowError, Streaming
>
> The issue happens with the following sample code: uses {{updateStateByKey}}
> followed by a {{map}} with checkpoint interval 10 seconds
> {code}
> val sparkConf = new SparkConf().setAppName("test")
> val streamingContext = new StreamingContext(sparkConf, Seconds(10))
> streamingContext.checkpoint("""checkpoint""")
> val source = streamingContext.socketTextStream("localhost", 9999)
> val updatedResult = source.map(
> (1,_)).updateStateByKey(
> (newlist : Seq[String], oldstate : Option[String]) =>
> newlist.headOption.orElse(oldstate))
> updatedResult.map(_._2)
> .checkpoint(Seconds(10))
> .foreachRDD((rdd, t) => {
> println("Deep: " + rdd.toDebugString.split("\n").length)
> println(t.toString() + ": " + rdd.collect.length)
> })
> streamingContext.start()
> streamingContext.awaitTermination()
> {code}
> From the output, we can see that the dependency will be increasing time over
> time, the {{updateStateByKey}} never get check-pointed, and finally, the
> stack overflow will happen.
> Note:
> * The rdd in {{updatedResult.map(_._2)}} get check-pointed in this case, but
> not the {{updateStateByKey}}
> * If remove the {{checkpoint(Seconds(10))}} from the map result (
> {{updatedResult.map(_._2)}} ), the stack overflow will not happen
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
