[HADOOP-11333] Fix deadlock in DomainSocketWatcher when the notification pipe is full

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 2.6.1, 2.7.0, 3.0.0-alpha1
• Labels: 2.6.1-candidate
• Target Version/s: 2.7.0
• Assignee: yunjiong zhao
• Reporter: yunjiong zhao
• Created: 24/Nov/14 03:55
• Updated: 30/Aug/16 01:31
• Resolved: 01/Dec/14 19:50

Description

I found some of our DataNodes will run "exceeds the limit of concurrent xciever", the limit is 4K.

After check the stack, I suspect that org.apache.hadoop.net.unix.DomainSocket.writeArray0 which called by DomainSocketWatcher.kick stuck:

"DataXceiver for client unix:/var/run/hadoop-hdfs/dn Waiting for operation #1" daemon prio=10 tid=0x00007f55c5576000 nid=0x385d waiting on condition [0x00007f558d5d4000]
java.lang.Thread.State: WAITING (parking)
at sun.misc.Unsafe.park(Native Method)

• parking to wait for <0x0000000740df9c90> (a java.util.concurrent.locks.ReentrantLock$NonfairSync)
  at java.util.concurrent.locks.LockSupport.park(LockSupport.java:186)
  at java.util.concurrent.locks.AbstractQueuedSynchronizer.parkAndCheckInterrupt(AbstractQueuedSynchronizer.java:834)
  at java.util.concurrent.locks.AbstractQueuedSynchronizer.acquireQueued(AbstractQueuedSynchronizer.java:867)
  at java.util.concurrent.locks.AbstractQueuedSynchronizer.acquire(AbstractQueuedSynchronizer.java:1197)
  at java.util.concurrent.locks.ReentrantLock$NonfairSync.lock(ReentrantLock.java:214)
  at java.util.concurrent.locks.ReentrantLock.lock(ReentrantLock.java:290)
  at org.apache.hadoop.net.unix.DomainSocketWatcher.add(DomainSocketWatcher.java:286)
  at org.apache.hadoop.hdfs.server.datanode.ShortCircuitRegistry.createNewMemorySegment(ShortCircuitRegistry.java:283)
  at org.apache.hadoop.hdfs.server.datanode.DataXceiver.requestShortCircuitShm(DataXceiver.java:413)
  at org.apache.hadoop.hdfs.protocol.datatransfer.Receiver.opRequestShortCircuitShm(Receiver.java:172)
  at org.apache.hadoop.hdfs.protocol.datatransfer.Receiver.processOp(Receiver.java:92)
  at org.apache.hadoop.hdfs.server.datanode.DataXceiver.run(DataXceiver.java:232)
  –
  "DataXceiver for client unix:/var/run/hadoop-hdfs/dn Waiting for operation #1" daemon prio=10 tid=0x00007f7de034c800 nid=0x7b7 runnable [0x00007f7db06c5000]
  java.lang.Thread.State: RUNNABLE
  at org.apache.hadoop.net.unix.DomainSocket.writeArray0(Native Method)
  at org.apache.hadoop.net.unix.DomainSocket.access$300(DomainSocket.java:45)
  at org.apache.hadoop.net.unix.DomainSocket$DomainOutputStream.write(DomainSocket.java:589)
  at org.apache.hadoop.net.unix.DomainSocketWatcher.kick(DomainSocketWatcher.java:350)
  at org.apache.hadoop.net.unix.DomainSocketWatcher.add(DomainSocketWatcher.java:303)
  at org.apache.hadoop.hdfs.server.datanode.ShortCircuitRegistry.createNewMemorySegment(ShortCircuitRegistry.java:283)
  at org.apache.hadoop.hdfs.server.datanode.DataXceiver.requestShortCircuitShm(DataXceiver.java:413)
  at org.apache.hadoop.hdfs.protocol.datatransfer.Receiver.opRequestShortCircuitShm(Receiver.java:172)
  at org.apache.hadoop.hdfs.protocol.datatransfer.Receiver.processOp(Receiver.java:92)
  at org.apache.hadoop.hdfs.server.datanode.DataXceiver.run(DataXceiver.java:232)
  at java.lang.Thread.run(Thread.java:745)

"DataXceiver for client unix:/var/run/hadoop-hdfs/dn Waiting for operation #1" daemon prio=10 tid=0x00007f55c5574000 nid=0x377a waiting on condition [0x00007f558d7d6000]
java.lang.Thread.State: WAITING (parking)
at sun.misc.Unsafe.park(Native Method)

• parking to wait for <0x0000000740df9cb0> (a java.util.concurrent.locks.AbstractQueuedSynchronizer$ConditionObject)
  at java.util.concurrent.locks.LockSupport.park(LockSupport.java:186)
  at java.util.concurrent.locks.AbstractQueuedSynchronizer$ConditionObject.await(AbstractQueuedSynchronizer.java:2043)
  at org.apache.hadoop.net.unix.DomainSocketWatcher.add(DomainSocketWatcher.java:306)
  at org.apache.hadoop.hdfs.server.datanode.ShortCircuitRegistry.createNewMemorySegment(ShortCircuitRegistry.java:283)
  at org.apache.hadoop.hdfs.server.datanode.DataXceiver.requestShortCircuitShm(DataXceiver.java:413)
  at org.apache.hadoop.hdfs.protocol.datatransfer.Receiver.opRequestShortCircuitShm(Receiver.java:172)
  at org.apache.hadoop.hdfs.protocol.datatransfer.Receiver.processOp(Receiver.java:92)
  at org.apache.hadoop.hdfs.server.datanode.DataXceiver.run(DataXceiver.java:232)
  at java.lang.Thread.run(Thread.java:745)

"Thread-163852" daemon prio=10 tid=0x00007f55c811c800 nid=0x6757 runnable [0x00007f55aef6e000]
java.lang.Thread.State: RUNNABLE
at org.apache.hadoop.net.unix.DomainSocketWatcher.doPoll0(Native Method)
at org.apache.hadoop.net.unix.DomainSocketWatcher.access$800(DomainSocketWatcher.java:52)
at org.apache.hadoop.net.unix.DomainSocketWatcher$1.run(DomainSocketWatcher.java:457)
at java.lang.Thread.run(Thread.java:745)

Issue Links

• is related to: HADOOP-11604 Prevent ConcurrentModificationException while closing domain sockets during shutdown of DomainSocketWatcher thread.
• relates to: HDFS-7212 FsDatasetImpl#createTemporary sometimes holds the FSDatasetImpl lock for a very long time