[
https://issues.apache.org/jira/browse/SPARK-22752?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16287305#comment-16287305
]
Marco Gaido commented on SPARK-22752:
-------------------------------------
Hi [~zsxwing], thanks for looking at this. The checkpointDir is on HDFS. Inside
it we have:
{noformat}
metadata . 0.1 kB
commits
offsets
sources
state
{noformat}
where {{commits}} contains files from {{0}} to {{13}}, {{offsets}} from {{0}}
to {{14}}, {{sources}} contains only a file which is {{0/0}} and {{state}}
contains only the directory {{0/0}} which is empty.
Do you need other information? Have you idea of which is the root cause of the
issue? Thanks.
> FileNotFoundException while reading from Kafka
> ----------------------------------------------
>
> Key: SPARK-22752
> URL: https://issues.apache.org/jira/browse/SPARK-22752
> Project: Spark
> Issue Type: Bug
> Components: Structured Streaming
> Affects Versions: 2.2.0
> Reporter: Marco Gaido
>
> We are running a stateful structured streaming job which reads from Kafka and
> writes to HDFS. And we are hitting this exception:
> {noformat}
> 17/12/08 05:20:12 ERROR FileFormatWriter: Aborting job null.
> org.apache.spark.SparkException: Job aborted due to stage failure: Task 0 in
> stage 1.0 failed 4 times, most recent failure: Lost task 0.3 in stage 1.0
> (TID 4, hcube1-1n03.eng.hortonworks.com, executor 1):
> java.lang.IllegalStateException: Error reading delta file
> /checkpointDir/state/0/0/1.delta of HDFSStateStoreProvider[id = (op=0,
> part=0), dir = /checkpointDir/state/0/0]: /checkpointDir/state/0/0/1.delta
> does not exist
> at
> org.apache.spark.sql.execution.streaming.state.HDFSBackedStateStoreProvider.org$apache$spark$sql$execution$streaming$state$HDFSBackedStateStoreProvider$$updateFromDeltaFile(HDFSBackedStateStoreProvider.scala:410)
> at
> org.apache.spark.sql.execution.streaming.state.HDFSBackedStateStoreProvider$$anonfun$org$apache$spark$sql$execution$streaming$state$HDFSBackedStateStoreProvider$$loadMap$1$$anonfun$6.apply(HDFSBackedStateStoreProvider.scala:362)
> at
> org.apache.spark.sql.execution.streaming.state.HDFSBackedStateStoreProvider$$anonfun$org$apache$spark$sql$execution$streaming$state$HDFSBackedStateStoreProvider$$loadMap$1$$anonfun$6.apply(HDFSBackedStateStoreProvider.scala:359)
> at scala.Option.getOrElse(Option.scala:121)
> at
> org.apache.spark.sql.execution.streaming.state.HDFSBackedStateStoreProvider$$anonfun$org$apache$spark$sql$execution$streaming$state$HDFSBackedStateStoreProvider$$loadMap$1.apply(HDFSBackedStateStoreProvider.scala:359)
> at
> org.apache.spark.sql.execution.streaming.state.HDFSBackedStateStoreProvider$$anonfun$org$apache$spark$sql$execution$streaming$state$HDFSBackedStateStoreProvider$$loadMap$1.apply(HDFSBackedStateStoreProvider.scala:358)
> at scala.Option.getOrElse(Option.scala:121)
> at
> org.apache.spark.sql.execution.streaming.state.HDFSBackedStateStoreProvider.org$apache$spark$sql$execution$streaming$state$HDFSBackedStateStoreProvider$$loadMap(HDFSBackedStateStoreProvider.scala:358)
> at
> org.apache.spark.sql.execution.streaming.state.HDFSBackedStateStoreProvider$$anonfun$org$apache$spark$sql$execution$streaming$state$HDFSBackedStateStoreProvider$$loadMap$1$$anonfun$6.apply(HDFSBackedStateStoreProvider.scala:360)
> at
> org.apache.spark.sql.execution.streaming.state.HDFSBackedStateStoreProvider$$anonfun$org$apache$spark$sql$execution$streaming$state$HDFSBackedStateStoreProvider$$loadMap$1$$anonfun$6.apply(HDFSBackedStateStoreProvider.scala:359)
> at scala.Option.getOrElse(Option.scala:121)
> {noformat}
> Of course, the file doesn't exist in HDFS. And in the {{state/0/0}} directory
> there is no file at all. While we have some files in the commits and offsets
> folders. I am not sure about the reason of this behavior. It seems to happen
> on the second time the job is started, after the first one failed. So it
> looks like task failures can generate it. Or it might be related to
> watermarks, since there are some problems related to the incoming data for
> which the watermark was filtering all the incoming data.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
