Details
-
Bug
-
Status: Open
-
Critical
-
Resolution: Unresolved
-
2.5.0
-
None
-
None
Description
We found that DFSInputStream#read does not return when hbase handlers read files from hdfs, and all handlers are in the org.apache.hadoop.net.SocketIOWithTimeout$SelectorPool.select(). jstack is as follows:
"RS_PARALLEL_SEEK-hadoop474:60020-9" prio=10 tid=0x00007f7350be0000 nid=0x1572 runnable [0x000000005a9de000]
java.lang.Thread.State: RUNNABLE
at sun.nio.ch.EPollArrayWrapper.epollWait(Native Method)
at sun.nio.ch.EPollArrayWrapper.poll(EPollArrayWrapper.java:269)
at sun.nio.ch.EPollSelectorImpl.doSelect(EPollSelectorImpl.java:79)
at sun.nio.ch.SelectorImpl.lockAndDoSelect(SelectorImpl.java:87)
- locked <0x000000039ad6e730> (a sun.nio.ch.Util$2)
- locked <0x000000039ad6e320> (a java.util.Collections$UnmodifiableSet)
- locked <0x00000002bf480738> (a sun.nio.ch.EPollSelectorImpl)
at sun.nio.ch.SelectorImpl.select(SelectorImpl.java:98)
at org.apache.hadoop.net.SocketIOWithTimeout$SelectorPool.select(SocketIOWithTimeout.java:335)
at org.apache.hadoop.net.SocketIOWithTimeout.doIO(SocketIOWithTimeout.java:157)
at org.apache.hadoop.net.SocketInputStream.read(SocketInputStream.java:161)
at org.apache.hadoop.net.SocketInputStream.read(SocketInputStream.java:131)
at org.apache.hadoop.net.SocketInputStream.read(SocketInputStream.java:118)
at java.io.FilterInputStream.read(FilterInputStream.java:83)
at org.apache.hadoop.hdfs.protocolPB.PBHelper.vintPrefixed(PBHelper.java:1986)
at org.apache.hadoop.hdfs.RemoteBlockReader2.newBlockReader(RemoteBlockReader2.java:395)
at org.apache.hadoop.hdfs.BlockReaderFactory.getRemoteBlockReader(BlockReaderFactory.java:786)
at org.apache.hadoop.hdfs.BlockReaderFactory.getRemoteBlockReaderFromTcp(BlockReaderFactory.java:665)
at org.apache.hadoop.hdfs.BlockReaderFactory.build(BlockReaderFactory.java:325)
at org.apache.hadoop.hdfs.DFSInputStream.actualGetFromOneDataNode(DFSInputStream.java:1023)
at org.apache.hadoop.hdfs.DFSInputStream.fetchBlockByteRange(DFSInputStream.java:966)
at org.apache.hadoop.hdfs.DFSInputStream.read(DFSInputStream.java:1293)
at org.apache.hadoop.fs.FSDataInputStream.read(FSDataInputStream.java:90)
at org.apache.hadoop.hbase.io.hfile.HFileBlock$AbstractFSReader.readAtOffset(HFileBlock.java:1223)
at org.apache.hadoop.hbase.io.hfile.HFileBlock$FSReaderV2.readBlockDataInternal(HFileBlock.java:1430)
at org.apache.hadoop.hbase.io.hfile.HFileBlock$FSReaderV2.readBlockData(HFileBlock.java:1312)
at org.apache.hadoop.hbase.io.hfile.HFileReaderV2.readBlock(HFileReaderV2.java:392)
at org.apache.hadoop.hbase.io.hfile.HFileBlockIndex$BlockIndexReader.loadDataBlockWithScanInfo(HFileBlockIndex.java:253)
at org.apache.hadoop.hbase.io.hfile.HFileReaderV2$AbstractScannerV2.seekTo(HFileReaderV2.java:532)
at org.apache.hadoop.hbase.io.hfile.HFileReaderV2$AbstractScannerV2.seekTo(HFileReaderV2.java:553)
at org.apache.hadoop.hbase.regionserver.StoreFileScanner.seekAtOrAfter(StoreFileScanner.java:237)
at org.apache.hadoop.hbase.regionserver.StoreFileScanner.seek(StoreFileScanner.java:152)
at org.apache.hadoop.hbase.regionserver.handler.ParallelSeekHandler.process(ParallelSeekHandler.java:57)
at org.apache.hadoop.hbase.executor.EventHandler.run(EventHandler.java:128)
at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1145)
at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:615)
at java.lang.Thread.run(Thread.java:744)
I read HDFS source code and discover:
1. NioInetPeer#in and NioInetPeer#out default timeout value is 0
NioInetPeer(Socket socket) throws IOException { this.socket = socket; this.in = new SocketInputStream(socket.getChannel(), 0); this.out = new SocketOutputStream(socket.getChannel(), 0); this.isLocal = socket.getInetAddress().equals(socket.getLocalAddress()); } public SocketInputStream(ReadableByteChannel channel, long timeout) throws IOException { SocketIOWithTimeout.checkChannelValidity(channel); reader = new Reader(channel, timeout); } Reader(ReadableByteChannel channel, long timeout) throws IOException { super((SelectableChannel)channel, timeout); this.channel = channel; } SocketIOWithTimeout(SelectableChannel channel, long timeout) throws IOException { checkChannelValidity(channel); this.channel = channel; this.timeout = timeout; // Set non-blocking channel.configureBlocking(false); }
and result in SocketIOWithTimeout#timeout=0
2. BlockReaderPeer#peer does not set ReadTimeout and WriteTimeout
which lead to org.apache.hadoop.net.SocketIOWithTimeout$SelectorPool.select(timeout=0) and does not return.
We can solve this problem by setting NioInetPeer's timeout in BlockReaderFactory#nextTcpPeer. Details is in patch file