[HDFS-15219] DFS Client will stuck when ResponseProcessor.run throw Error

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 2.7.3
• Fix Version/s: 3.3.0, 3.1.4, 3.2.2
• Component/s: hdfs-client
• Hadoop Flags: Reviewed
• Assignee: Chenyu Zheng
• Reporter: Chenyu Zheng
• Created: 12/Mar/20 02:55
• Updated: 09/Apr/20 06:11
• Resolved: 24/Mar/20 17:49

Description

In my case, a Tez application stucked more than 2 hours util we kill this applicaiton. The Reason is a task attempt stucked, becuase speculative execution is disable. 

Then Exception like this:

2020-03-11 01:23:59,141 [INFO] [TezChild] |exec.MapOperator|: MAP[4]: records read - 100000
2020-03-11 01:24:50,294 [INFO] [TezChild] |exec.FileSinkOperator|: FS[3]: records written - 1000000
2020-03-11 01:24:50,294 [INFO] [TezChild] |exec.MapOperator|: MAP[4]: records read - 1000000
2020-03-11 01:29:02,967 [FATAL] [ResponseProcessor for block BP-1856561198-172.16.6.67-1421842461517:blk_15177828027_14109212073] |yarn.YarnUncaughtExceptionHandler|: Thread Thread[ResponseProcessor for block BP-1856561198-172.16.6.67-1421842461517:blk_15177828027_14109212073,5,main] threw an Error. Shutting down now...
java.lang.NoClassDefFoundError: com/google/protobuf/TextFormat
 at org.apache.hadoop.hdfs.protocol.datatransfer.PipelineAck.toString(PipelineAck.java:253)
 at java.lang.String.valueOf(String.java:2847)
 at java.lang.StringBuilder.append(StringBuilder.java:128)
 at org.apache.hadoop.hdfs.DFSOutputStream$DataStreamer$ResponseProcessor.run(DFSOutputStream.java:737)
Caused by: java.lang.ClassNotFoundException: com.google.protobuf.TextFormat
 at java.net.URLClassLoader$1.run(URLClassLoader.java:363)
 at java.net.URLClassLoader$1.run(URLClassLoader.java:355)
 at java.security.AccessController.doPrivileged(Native Method)
 at java.net.URLClassLoader.findClass(URLClassLoader.java:354)
 at java.lang.ClassLoader.loadClass(ClassLoader.java:425)
 at sun.misc.Launcher$AppClassLoader.loadClass(Launcher.java:308)
 at java.lang.ClassLoader.loadClass(ClassLoader.java:358)
 ... 4 more
Caused by: java.util.zip.ZipException: error reading zip file
 at java.util.zip.ZipFile.read(Native Method)
 at java.util.zip.ZipFile.access$1400(ZipFile.java:56)
 at java.util.zip.ZipFile$ZipFileInputStream.read(ZipFile.java:679)
 at java.util.zip.ZipFile$ZipFileInflaterInputStream.fill(ZipFile.java:415)
 at java.util.zip.InflaterInputStream.read(InflaterInputStream.java:158)
 at sun.misc.Resource.getBytes(Resource.java:124)
 at java.net.URLClassLoader.defineClass(URLClassLoader.java:444)
 at java.net.URLClassLoader.access$100(URLClassLoader.java:71)
 at java.net.URLClassLoader$1.run(URLClassLoader.java:361)
 ... 10 more
2020-03-11 01:29:02,970 [INFO] [ResponseProcessor for block BP-1856561198-172.16.6.67-1421842461517:blk_15177828027_14109212073] |util.ExitUtil|: Exiting with status -1
2020-03-11 03:27:26,833 [INFO] [TaskHeartbeatThread] |task.TaskReporter|: Received should die response from AM
2020-03-11 03:27:26,834 [INFO] [TaskHeartbeatThread] |task.TaskReporter|: Asked to die via task heartbeat
2020-03-11 03:27:26,839 [INFO] [TaskHeartbeatThread] |task.TezTaskRunner2|: Attempting to abort attempt_1583335296048_917815_3_01_000704_0 due to an invocation of shutdownRequested

Reason is UncaughtException. When time is 01:29, a disk was error, so throw NoClassDefFoundError. ResponseProcessor.run only catch Exception, can't catch NoClassDefFoundError. So the ReponseProcessor didn't set errorState. Then DataStream didn't know ReponseProcessor was dead, and can't trigger closeResponder, so stucked in DataStream.run.

 I tested in unit-test TestDataStream.testDfsClient. When I throw NoClassDefFoundError in ResponseProcessor.run, the TestDataStream.testDfsClient will failed bacause of timeout.

I think we should catch Throwable but not Exception in ReponseProcessor.run.

 

Issue Links

• links to: GitHub Pull Request #1902