Uploaded image for project: 'Hadoop HDFS'
  1. Hadoop HDFS
  2. HDFS-8070

Pre-HDFS-7915 DFSClient cannot use short circuit on post-HDFS-7915 DataNode

    XMLWordPrintableJSON

Details

    Description

      HDFS ShortCircuitShm layer keeps the task locked up during multi-threaded split-generation.

      I hit this immediately after I upgraded the data, so I wonder if the ShortCircuitShim wire protocol has trouble when 2.8.0 DN talks to a 2.7.0 Client?

      2015-04-06 00:04:30,780 INFO [ORC_GET_SPLITS #3] orc.OrcInputFormat: ORC pushdown predicate: leaf-0 = (IS_NULL ss_sold_date_sk)
      expr = (not leaf-0)
      2015-04-06 00:04:30,781 ERROR [ShortCircuitCache_SlotReleaser] shortcircuit.ShortCircuitCache: ShortCircuitCache(0x29e82045): failed to release short-circuit shared memory slot Slot(slotIdx=2, shm=DfsClientShm(a86ee34576d93c4964005d90b0d97c38)) by sending ReleaseShortCircuitAccessRequestProto to /grid/0/cluster/hdfs/dn_socket.  Closing shared memory segment.
      java.io.IOException: ERROR_INVALID: there is no shared memory segment registered with shmId a86ee34576d93c4964005d90b0d97c38
      	at org.apache.hadoop.hdfs.shortcircuit.ShortCircuitCache$SlotReleaser.run(ShortCircuitCache.java:208)
      	at java.util.concurrent.Executors$RunnableAdapter.call(Executors.java:511)
      	at java.util.concurrent.FutureTask.run(FutureTask.java:266)
      	at java.util.concurrent.ScheduledThreadPoolExecutor$ScheduledFutureTask.access$201(ScheduledThreadPoolExecutor.java:180)
      	at java.util.concurrent.ScheduledThreadPoolExecutor$ScheduledFutureTask.run(ScheduledThreadPoolExecutor.java:293)
      	at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1142)
      	at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:617)
      	at java.lang.Thread.run(Thread.java:745)
      2015-04-06 00:04:30,781 INFO [ORC_GET_SPLITS #5] orc.OrcInputFormat: ORC pushdown predicate: leaf-0 = (IS_NULL ss_sold_date_sk)
      expr = (not leaf-0)
      2015-04-06 00:04:30,781 WARN [ShortCircuitCache_SlotReleaser] shortcircuit.DfsClientShmManager: EndpointShmManager(172.19.128.60:50010, parent=ShortCircuitShmManager(5e763476)): error shutting down shm: got IOException calling shutdown(SHUT_RDWR)
      java.nio.channels.ClosedChannelException
      	at org.apache.hadoop.util.CloseableReferenceCount.reference(CloseableReferenceCount.java:57)
      	at org.apache.hadoop.net.unix.DomainSocket.shutdown(DomainSocket.java:387)
      	at org.apache.hadoop.hdfs.shortcircuit.DfsClientShmManager$EndpointShmManager.shutdown(DfsClientShmManager.java:378)
      	at org.apache.hadoop.hdfs.shortcircuit.ShortCircuitCache$SlotReleaser.run(ShortCircuitCache.java:223)
      	at java.util.concurrent.Executors$RunnableAdapter.call(Executors.java:511)
      	at java.util.concurrent.FutureTask.run(FutureTask.java:266)
      	at java.util.concurrent.ScheduledThreadPoolExecutor$ScheduledFutureTask.access$201(ScheduledThreadPoolExecutor.java:180)
      	at java.util.concurrent.ScheduledThreadPoolExecutor$ScheduledFutureTask.run(ScheduledThreadPoolExecutor.java:293)
      	at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1142)
      	at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:617)
      	at java.lang.Thread.run(Thread.java:745)
      2015-04-06 00:04:30,783 INFO [ORC_GET_SPLITS #7] orc.OrcInputFormat: ORC pushdown predicate: leaf-0 = (IS_NULL cs_sold_date_sk)
      expr = (not leaf-0)
      2015-04-06 00:04:30,785 ERROR [ShortCircuitCache_SlotReleaser] shortcircuit.ShortCircuitCache: ShortCircuitCache(0x29e82045): failed to release short-circuit shared memory slot Slot(slotIdx=4, shm=DfsClientShm(a86ee34576d93c4964005d90b0d97c38)) by sending ReleaseShortCircuitAccessRequestProto to /grid/0/cluster/hdfs/dn_socket.  Closing shared memory segment.
      java.io.IOException: ERROR_INVALID: there is no shared memory segment registered with shmId a86ee34576d93c4964005d90b0d97c38
      	at org.apache.hadoop.hdfs.shortcircuit.ShortCircuitCache$SlotReleaser.run(ShortCircuitCache.java:208)
      	at java.util.concurrent.Executors$RunnableAdapter.call(Executors.java:511)
      	at java.util.concurrent.FutureTask.run(FutureTask.java:266)
      	at java.util.concurrent.ScheduledThreadPoolExecutor$ScheduledFutureTask.access$201(ScheduledThreadPoolExecutor.java:180)
      	at java.util.concurrent.ScheduledThreadPoolExecutor$ScheduledFutureTask.run(ScheduledThreadPoolExecutor.java:293)
      	at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1142)
      	at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:617)
      	at java.lang.Thread.run(Thread.java:745)
      

      Looks like a double free-fd condition?

      2015-04-02 18:58:47,653 [DataXceiver for client unix:/grid/0/cluster/hdfs/dn_socket [Passing file descriptors for block BP-942051088-172.18.0.41-1370508013893:blk_1076973408_1099515627985]] INFO org.apache.hadoop.hdfs.server.datanode.DataNode: Unregistering SlotId(3bd7fd9aed791e95acfb5034e6617d83:0) because the requestShortCircuitFdsForRead operation failed.
      2015-04-02 18:58:47,653 [DataXceiver for client unix:/grid/0/cluster/hdfs/dn_socket [Passing file descriptors for block BP-942051088-<ip>-1370508013893:blk_1076973408_1099515627985]] INFO org.apache.hadoop.hdfs.server.datanode.DataNode.clienttrace: src: 127.0.0.1, dest: 127.0.0.1, op: REQUEST_SHORT_CIRCUIT_FDS, blockid: 1076973408, srvID: ba7b6f19-47e0-4b86-af50-23981649318c, success: false
      2015-04-02 18:58:47,654 [DataXceiver for client unix:/grid/0/cluster/hdfs/dn_socket [Passing file descriptors for block BP-942051088-172.18.0.41-1370508013893:blk_1076973408_1099515627985]] ERROR org.apache.hadoop.hdfs.server.datanode.DataNode: cn060-10.l42scl.hortonworks.com:50010:DataXceiver error processing REQUEST_SHORT_CIRCUIT_FDS operation  src: unix:/grid/0/cluster/hdfs/dn_socket dst: <local>
      java.io.EOFException
              at org.apache.hadoop.hdfs.server.datanode.DataXceiver.requestShortCircuitFds(DataXceiver.java:352)
              at org.apache.hadoop.hdfs.protocol.datatransfer.Receiver.opRequestShortCircuitFds(Receiver.java:187)
              at org.apache.hadoop.hdfs.protocol.datatransfer.Receiver.processOp(Receiver.java:89)
              at org.apache.hadoop.hdfs.server.datanode.DataXceiver.run(DataXceiver.java:251)
              at java.lang.Thread.run(Thread.java:745)
      

      Investigating more, since the exact exception from the DataNode call is not logged.

      Attachments

        1. HDFS-8070.branch-2.6.patch
          5 kB
          Akira Ajisaka
        2. HDFS-8070.001.patch
          5 kB
          Colin McCabe

        Issue Links

          Activity

            People

              cmccabe Colin McCabe
              gopalv Gopal Vijayaraghavan
              Votes:
              1 Vote for this issue
              Watchers:
              16 Start watching this issue

              Dates

                Created:
                Updated:
                Resolved: