More info -- this is not a Hadoop issue. The network performance issue can be replicated with SSH only on the links where Hadoop has a problem, and only in the direction with a problem.
HDFS is slow to transfer data in certain directions from certain machines. So, for example, copying from node C to D may be slow, but not the other direction from C to D. Likewise, although only 3 of 8 nodes have this problem, it is not universal. For example, node C might have trouble copying data to 5 of the 7 other nodes, and node G might have trouble with all 7 other nodes. No idea what it is yet, but SSH exhibits the same issue -- only in those specific point-to-point links in one specific direction. -Scott On Apr 16, 2010, at 7:10 PM, Scott Carey wrote: > Ok, so here is a ... fun result. > > I have dfs.replication.min set to 2, so I can't just do > hsdoop fs -Ddfs.replication=1 put someFile someFile > Since that will fail. > > So here are two results that are fascinating: > > $ time hadoop fs -Ddfs.replication=3 -put test.tar test.tar > real 1m53.237s > user 0m1.952s > sys 0m0.308s > > $ time hadoop fs -Ddfs.replication=2 -put test.tar test.tar > real 0m1.689s > user 0m1.763s > sys 0m0.315s > > > > The file is 77MB and so is two blocks. > The test with replication level 3 is slow about 9 out of 10 times. When it > is slow it sometimes is 28 seconds, sometimes 2 minutes. It was fast one > time... > The test with replication level 2 is fast in 40 out of 40 tests. > > This is a development cluster with 8 nodes. > > It looks like the replication level of 3 or more causes trouble. Looking > more closely at the logs, it seems that certain datanodes (but not all) cause > large delays if they are in the middle of an HDFS write chain. So, a write > that goes from A > B > C is fast if B is a good node and C a bad node. If > its A > C > B then its slow. > > So, I can say that some nodes but not all are doing something wrong. when in > the middle of a write chain. If I do a replication = 2 write on one of these > bad nodes, its always slow. > > So the good news is I can identify the bad nodes, and decomission them. The > bad news is this still doesn't make a lot of sense, and 40% of the nodes have > the issue. Worse, on a couple nodes the behavior in the replication = 2 case > is not consistent -- sometimes the first block is fast. So it may be > dependent on not just the source, but the source <> target combination in the > chain. > > > At this point, I suspect something completely broken at the network level, > perhaps even routing. Why it would show up after an upgrade is yet to be > determined, but the upgrade did include some config changes and OS updates. > > Thanks Todd! > > -Scott > > > On Apr 16, 2010, at 5:34 PM, Todd Lipcon wrote: > >> Hey Scott, >> >> This is indeed really strange... if you do a straight "hadoop fs -put" with >> dfs.replication set to 1 from one of the DNs, does it upload slow? That >> would cut out the network from the equation. >> >> -Todd >> >> On Fri, Apr 16, 2010 at 5:29 PM, Scott Carey <[email protected]>wrote: >> >>> I have two clusters upgraded to CDH2. One is performing fine, and the >>> other is EXTREMELY slow. >>> >>> Some jobs that formerly took 90 seconds, take 20 to 50 minutes. >>> >>> It is an HDFS issue from what I can tell. >>> >>> The simple DFS benchmark with one map task shows the problem clearly. I >>> have looked at every difference I can find and am wondering where else to >>> look to track this down. >>> The disks on all nodes in the cluster check out -- capable of 75MB/sec >>> minimum with a 'dd' write test. >>> top / iostat do not show any significant CPU usage or iowait times on any >>> machines in the cluster during the test. >>> ifconfig does not report any dropped packets or other errors on any machine >>> in the cluster. dmesg has nothing interesting. >>> The poorly performing cluster is on a slightly newer CentOS version: >>> Poor: 2.6.18-164.15.1.el5 #1 SMP Wed Mar 17 11:30:06 EDT 2010 x86_64 x86_64 >>> x86_64 GNU/Linux (CentOS 5.4, recent patches) >>> Good: 2.6.18-128.el5 #1 SMP Wed Jan 21 10:41:14 EST 2009 x86_64 x86_64 >>> x86_64 GNU/Linux (CentOS 5.3, I think) >>> The performance is always poor, not sporadically poor. It is poor with M/R >>> tasks as well as non-M/R HDFS clients (i.e. sqoop). >>> >>> Poor performance cluster (no other jobs active during the test): >>> --------------------------------------------------- >>> $ hadoop jar /usr/lib/hadoop/hadoop-0.20.1+169.68-test.jar TestDFSIO -write >>> -nrFiles 1 -fileSize 2000 >>> 10/04/16 12:53:13 INFO mapred.FileInputFormat: nrFiles = 1 >>> 10/04/16 12:53:13 INFO mapred.FileInputFormat: fileSize (MB) = 2000 >>> 10/04/16 12:53:13 INFO mapred.FileInputFormat: bufferSize = 1000000 >>> 10/04/16 12:53:14 INFO mapred.FileInputFormat: creating control file: 2000 >>> mega bytes, 1 files >>> 10/04/16 12:53:14 INFO mapred.FileInputFormat: created control files for: 1 >>> files >>> 10/04/16 12:53:14 WARN mapred.JobClient: Use GenericOptionsParser for >>> parsing the arguments. Applications should implement Tool for the same. >>> 10/04/16 12:53:15 INFO mapred.FileInputFormat: Total input paths to process >>> : 1 >>> 10/04/16 12:53:15 INFO mapred.JobClient: Running job: job_201004091928_0391 >>> 10/04/16 12:53:16 INFO mapred.JobClient: map 0% reduce 0% >>> 10/04/16 13:42:30 INFO mapred.JobClient: map 100% reduce 0% >>> 10/04/16 13:43:06 INFO mapred.JobClient: map 100% reduce 100% >>> 10/04/16 13:43:07 INFO mapred.JobClient: Job complete: >>> job_201004091928_0391 >>> [snip] >>> 10/04/16 13:43:07 INFO mapred.FileInputFormat: ----- TestDFSIO ----- : >>> write >>> 10/04/16 13:43:07 INFO mapred.FileInputFormat: Date & time: Fri >>> Apr 16 13:43:07 PDT 2010 >>> 10/04/16 13:43:07 INFO mapred.FileInputFormat: Number of files: 1 >>> 10/04/16 13:43:07 INFO mapred.FileInputFormat: Total MBytes processed: 2000 >>> 10/04/16 13:43:07 INFO mapred.FileInputFormat: Throughput mb/sec: >>> 0.678296742615553 >>> 10/04/16 13:43:07 INFO mapred.FileInputFormat: Average IO rate mb/sec: >>> 0.6782967448234558 >>> 10/04/16 13:43:07 INFO mapred.FileInputFormat: IO rate std deviation: >>> 9.568803140552889E-5 >>> 10/04/16 13:43:07 INFO mapred.FileInputFormat: Test exec time sec: >>> 2992.913 >>> ------------------------------------ >>> >>> Good performance cluster (other jobs active during the test): >>> --------------------------------- >>> hadoop jar /usr/lib/hadoop/hadoop-0.20.1+169.68-test.jar TestDFSIO -write >>> -nrFiles 1 -fileSize 2000 >>> 10/04/16 12:50:52 WARN conf.Configuration: DEPRECATED: hadoop-site.xml >>> found in the classpath. Usage of hadoop-site.xml is deprecated. Instead use >>> core-site.xml, mapred-site.xml and hdfs-site.xml to override properties of >>> core-default.xml, mapred-default.xml and hdfs-default.xml respectively >>> TestFDSIO.0.0.4 >>> 10/04/16 12:50:52 INFO mapred.FileInputFormat: nrFiles = 1 >>> 10/04/16 12:50:52 INFO mapred.FileInputFormat: fileSize (MB) = 2000 >>> 10/04/16 12:50:52 INFO mapred.FileInputFormat: bufferSize = 1000000 >>> 10/04/16 12:50:52 INFO mapred.FileInputFormat: creating control file: 2000 >>> mega bytes, 1 files >>> 10/04/16 12:50:52 INFO mapred.FileInputFormat: created control files for: 1 >>> files >>> 10/04/16 12:50:52 WARN mapred.JobClient: Use GenericOptionsParser for >>> parsing the arguments. Applications should implement Tool for the same. >>> 10/04/16 12:50:53 INFO mapred.FileInputFormat: Total input paths to process >>> : 1 >>> 10/04/16 12:50:54 INFO mapred.JobClient: Running job: job_201003311607_4098 >>> 10/04/16 12:50:55 INFO mapred.JobClient: map 0% reduce 0% >>> 10/04/16 12:51:22 INFO mapred.JobClient: map 100% reduce 0% >>> 10/04/16 12:51:32 INFO mapred.JobClient: map 100% reduce 100% >>> 10/04/16 12:51:32 INFO mapred.JobClient: Job complete: >>> job_201003311607_4098 >>> [snip] >>> 10/04/16 12:51:32 INFO mapred.FileInputFormat: ----- TestDFSIO ----- : >>> write >>> 10/04/16 12:51:32 INFO mapred.FileInputFormat: Date & time: Fri >>> Apr 16 12:51:32 PDT 2010 >>> 10/04/16 12:51:32 INFO mapred.FileInputFormat: Number of files: 1 >>> 10/04/16 12:51:32 INFO mapred.FileInputFormat: Total MBytes processed: 2000 >>> 10/04/16 12:51:32 INFO mapred.FileInputFormat: Throughput mb/sec: >>> 92.47699634715865 >>> 10/04/16 12:51:32 INFO mapred.FileInputFormat: Average IO rate mb/sec: >>> 92.47699737548828 >>> 10/04/16 12:51:32 INFO mapred.FileInputFormat: IO rate std deviation: >>> 0.008579127784432736 >>> 10/04/16 12:51:32 INFO mapred.FileInputFormat: Test exec time sec: >>> 39.522 >>> ------------------------------- >>> >>> On the namenode logs during the write, there are about 2 minute intervals >>> between block writes (64MB) >>> 2010-04-16 12:56:41,081 INFO org.apache.hadoop.hdfs.StateChange: BLOCK* >>> NameSystem.allocateBlock: /benchmarks/TestDFSIO/io_data/test_io_0. >>> blk_5304238373340657846_291044 >>> 2010-04-16 12:58:15,144 INFO org.apache.hadoop.hdfs.StateChange: BLOCK* >>> NameSystem.addStoredBlock: blockMap updated: 10.3.0.142:50010 is added to >>> blk_5304238373340657846_291044 size 67108864 >>> 2010-04-16 12:58:15,145 INFO org.apache.hadoop.hdfs.StateChange: BLOCK* >>> NameSystem.addStoredBlock: blockMap updated: 10.3.0.140:50010 is added to >>> blk_5304238373340657846_291044 size 67108864 >>> 2010-04-16 12:58:15,147 INFO org.apache.hadoop.hdfs.StateChange: BLOCK* >>> NameSystem.addStoredBlock: blockMap updated: 10.3.0.146:50010 is added to >>> blk_5304238373340657846_291044 size 67108864 >>> 2010-04-16 12:58:15,148 INFO org.apache.hadoop.hdfs.StateChange: BLOCK* >>> NameSystem.allocateBlock: /benchmarks/TestDFSIO/io_data/test_io_0. >>> blk_-4700875175596279045_291044 >>> 2010-04-16 12:58:15,792 INFO org.apache.hadoop.hdfs.StateChange: BLOCK* >>> NameSystem.addStoredBlock: blockMap updated: 10.3.0.147:50010 is added to >>> blk_-4700875175596279045_291044 size 67108864 >>> 2010-04-16 12:58:15,792 INFO org.apache.hadoop.hdfs.StateChange: BLOCK* >>> NameSystem.addStoredBlock: blockMap updated: 10.3.0.149:50010 is added to >>> blk_-4700875175596279045_291044 size 67108864 >>> 2010-04-16 12:58:15,794 INFO org.apache.hadoop.hdfs.StateChange: BLOCK* >>> NameSystem.addStoredBlock: blockMap updated: 10.3.0.146:50010 is added to >>> blk_-4700875175596279045_291044 size 67108864 >>> 2010-04-16 12:58:15,795 INFO org.apache.hadoop.hdfs.StateChange: BLOCK* >>> NameSystem.allocateBlock: /benchmarks/TestDFSIO/io_data/test_io_0. >>> blk_1059681682487510599_291044 >>> >>> On the node where the map task ran, its datanode log has similar chunks of >>> time while it receives the block: >>> 2010-04-16 12:55:08,303 INFO >>> org.apache.hadoop.hdfs.server.datanode.DataNode: Receiving block >>> blk_-7985493313278868469_291044 src: /10.3.0.146:48168 dest: / >>> 10.3.0.146:50010 >>> 2010-04-16 12:56:42,281 INFO >>> org.apache.hadoop.hdfs.server.datanode.DataNode.clienttrace: src: / >>> 10.3.0.146:48168, dest: /10.3.0.146:50010, bytes: 67108864, op: >>> HDFS_WRITE, cliID: DFSClient_attempt_201004091928_0391_m_000000_0, offset: >>> 0, srvID: DS-350876050-10.3.0.146-50010-1265338485169, blockid: >>> blk_-7985493313278868469_291044, duration: 93972113000 >>> 2010-04-16 12:56:42,282 INFO >>> org.apache.hadoop.hdfs.server.datanode.DataNode: PacketResponder 2 for block >>> blk_-7985493313278868469_291044 terminating >>> 2010-04-16 12:56:42,285 INFO >>> org.apache.hadoop.hdfs.server.datanode.DataNode: Receiving block >>> blk_5304238373340657846_291044 src: /10.3.0.146:48175 dest: / >>> 10.3.0.146:50010 >>> 2010-04-16 12:58:16,350 INFO >>> org.apache.hadoop.hdfs.server.datanode.DataNode.clienttrace: src: / >>> 10.3.0.146:48175, dest: /10.3.0.146:50010, bytes: 67108864, op: >>> HDFS_WRITE, cliID: DFSClient_attempt_201004091928_0391_m_000000_0, offset: >>> 0, srvID: DS-350876050-10.3.0.146-50010-1265338485169, blockid: >>> blk_5304238373340657846_291044, duration: 94059819000 >>> 2010-04-16 12:58:16,350 INFO >>> org.apache.hadoop.hdfs.server.datanode.DataNode: PacketResponder 2 for block >>> blk_5304238373340657846_291044 terminating >>> 2010-04-16 12:58:16,353 INFO >>> org.apache.hadoop.hdfs.server.datanode.DataNode: Receiving block >>> blk_-4700875175596279045_291044 src: /10.3.0.146:48182 dest: / >>> 10.3.0.146:50010 >>> 2010-04-16 12:58:16,997 INFO >>> org.apache.hadoop.hdfs.server.datanode.DataNode.clienttrace: src: / >>> 10.3.0.146:48182, dest: /10.3.0.146:50010, bytes: 67108864, op: >>> HDFS_WRITE, cliID: DFSClient_attempt_201004091928_0391_m_000000_0, offset: >>> 0, srvID: DS-350876050-10.3.0.146-50010-1265338485169, blockid: >>> blk_-4700875175596279045_291044, duration: 637761000 >>> 2010-04-16 12:58:16,997 INFO >>> org.apache.hadoop.hdfs.server.datanode.DataNode: PacketResponder 2 for block >>> blk_-4700875175596279045_291044 terminating >>> >>> >>> On a different datanode it looks essentially the same but fewer blocks, >>> since all blocks are written to the local node once but only (replication >>> -1)/(cluster size -1) times as often elsewhere. >>> >>> I grabbed a few stack traces of the running map task, but I don't think >>> there is any surprises there -- its waiting on an HDFS write: >>> "DataStreamer for file /benchmarks/TestDFSIO/io_data/test_io_0 block >>> blk_6860895546192904948_291044" daemon prio=10 tid=0x00002aab08244000 >>> nid=0x1031 runnable [0x0000000040c7a000] >>> java.lang.Thread.State: RUNNABLE >>> at sun.nio.ch.EPollArrayWrapper.epollWait(Native Method) >>> at sun.nio.ch.EPollArrayWrapper.poll(EPollArrayWrapper.java:215) >>> at sun.nio.ch.EPollSelectorImpl.doSelect(EPollSelectorImpl.java:65) >>> at sun.nio.ch.SelectorImpl.lockAndDoSelect(SelectorImpl.java:69) >>> - locked <0x00002aaab363dc88> (a sun.nio.ch.Util$1) >>> - locked <0x00002aaab363dc78> (a >>> java.util.Collections$UnmodifiableSet) >>> - locked <0x00002aaab363da50> (a sun.nio.ch.EPollSelectorImpl) >>> at sun.nio.ch.SelectorImpl.select(SelectorImpl.java:80) >>> at >>> org.apache.hadoop.net.SocketIOWithTimeout$SelectorPool.select(SocketIOWithTimeout.java:332) >>> at >>> org.apache.hadoop.net.SocketIOWithTimeout.doIO(SocketIOWithTimeout.java:157) >>> at >>> org.apache.hadoop.net.SocketOutputStream.write(SocketOutputStream.java:146) >>> at >>> org.apache.hadoop.net.SocketOutputStream.write(SocketOutputStream.java:107) >>> at java.io.BufferedOutputStream.write(BufferedOutputStream.java:105) >>> - locked <0x00002aaaf89023c0> (a java.io.BufferedOutputStream) >>> at java.io.DataOutputStream.write(DataOutputStream.java:90) >>> >>> "main" prio=10 tid=0x0000000055fb6000 nid=0x1017 waiting for monitor entry >>> [0x000000004112c000] >>> java.lang.Thread.State: BLOCKED (on object monitor) >>> at >>> org.apache.hadoop.hdfs.DFSClient$DFSOutputStream.writeChunk(DFSClient.java:3006) >>> - waiting to lock <0x00002aaac36bb440> (a java.util.LinkedList) >>> - locked <0x00002aaac36bae60> (a >>> org.apache.hadoop.hdfs.DFSClient$DFSOutputStream) >>> at >>> org.apache.hadoop.fs.FSOutputSummer.writeChecksumChunk(FSOutputSummer.java:150) >>> >>> >>> >>> I can get some stack traces from the namenode and datanodes during this >>> too. Any other suggestions? Is there anything in a 0.19.2 to 0.20.1+ >>> upgrade that might end up improperly configured that would have this sort of >>> symptom? Its almost like HDFS just decides to sleep for a second between >>> each 64K network transfer. Non-hadoop network transfer (i.e. scp) seems >>> fine. >>> >>> Another bit of information -- its not just HDFS writes, the read test is >>> almost equally awful. (1.9MB/sec) >>> >>> The Map task running the read test is in this stack trace: >>> "main" prio=10 tid=0x00000000554b3000 nid=0x3376 runnable >>> [0x0000000040a1d000] >>> java.lang.Thread.State: RUNNABLE >>> at sun.nio.ch.EPollArrayWrapper.epollWait(Native Method) >>> at sun.nio.ch.EPollArrayWrapper.poll(EPollArrayWrapper.java:215) >>> at sun.nio.ch.EPollSelectorImpl.doSelect(EPollSelectorImpl.java:65) >>> at sun.nio.ch.SelectorImpl.lockAndDoSelect(SelectorImpl.java:69) >>> - locked <0x00002aaaf7641740> (a sun.nio.ch.Util$1) >>> - locked <0x00002aaaf7641730> (a >>> java.util.Collections$UnmodifiableSet) >>> - locked <0x00002aaaf7641508> (a sun.nio.ch.EPollSelectorImpl) >>> at sun.nio.ch.SelectorImpl.select(SelectorImpl.java:80) >>> at >>> org.apache.hadoop.net.SocketIOWithTimeout$SelectorPool.select(SocketIOWithTimeout.java:332) >>> at >>> org.apache.hadoop.net.SocketIOWithTimeout.doIO(SocketIOWithTimeout.java:157) >>> at >>> org.apache.hadoop.net.SocketInputStream.read(SocketInputStream.java:155) >>> at >>> org.apache.hadoop.net.SocketInputStream.read(SocketInputStream.java:128) >>> at java.io.BufferedInputStream.fill(BufferedInputStream.java:218) >>> at java.io.BufferedInputStream.read1(BufferedInputStream.java:258) >>> at java.io.BufferedInputStream.read(BufferedInputStream.java:317) >>> - locked <0x00002aaaf76400f8> (a java.io.BufferedInputStream) >>> at java.io.DataInputStream.read(DataInputStream.java:132) >>> at org.apache.hadoop.io.IOUtils.readFully(IOUtils.java:100) >>> at >>> org.apache.hadoop.hdfs.DFSClient$BlockReader.readChunk(DFSClient.java:1283) >>> >>> >>> The namenode is not doing anything as far as I can tell in the stack >>> traces. The datanode that is being accessed for data at any given time >>> seems to be in this stack trace: >>> "org.apache.hadoop.hdfs.server.datanode.dataxcei...@7b90a2d" daemon >>> prio=10 tid=0x00002aaaf43b8000 nid=0x79cc runnable [0x000000004433c000] >>> java.lang.Thread.State: RUNNABLE >>> at sun.nio.ch.EPollArrayWrapper.epollWait(Native Method) >>> at sun.nio.ch.EPollArrayWrapper.poll(EPollArrayWrapper.java:215) >>> at sun.nio.ch.EPollSelectorImpl.doSelect(EPollSelectorImpl.java:65) >>> at sun.nio.ch.SelectorImpl.lockAndDoSelect(SelectorImpl.java:69) >>> - locked <0x00002aaae7ee8440> (a sun.nio.ch.Util$1) >>> - locked <0x00002aaae7ee8428> (a >>> java.util.Collections$UnmodifiableSet) >>> - locked <0x00002aaae7ee8098> (a sun.nio.ch.EPollSelectorImpl) >>> at sun.nio.ch.SelectorImpl.select(SelectorImpl.java:80) >>> at >>> org.apache.hadoop.net.SocketIOWithTimeout$SelectorPool.select(SocketIOWithTimeout.java:332) >>> at >>> org.apache.hadoop.net.SocketIOWithTimeout.waitForIO(SocketIOWithTimeout.java:245) >>> at >>> org.apache.hadoop.net.SocketOutputStream.waitForWritable(SocketOutputStream.java:159) >>> at >>> org.apache.hadoop.net.SocketOutputStream.transferToFully(SocketOutputStream.java:198) >>> at >>> org.apache.hadoop.hdfs.server.datanode.BlockSender.sendChunks(BlockSender.java:313) >>> at >>> org.apache.hadoop.hdfs.server.datanode.BlockSender.sendBlock(BlockSender.java:401) >>> at >>> org.apache.hadoop.hdfs.server.datanode.DataXceiver.readBlock(DataXceiver.java:180) >>> at >>> org.apache.hadoop.hdfs.server.datanode.DataXceiver.run(DataXceiver.java:95) >>> at java.lang.Thread.run(Thread.java:619) >>> >>> >>> So the behavior is as if the network is extremely slow, but it seems to >>> only affect Hadoop. Any ideas? >>> >>> Thanks, >>> >>> -Scott >>> >>> >> >> >> -- >> Todd Lipcon >> Software Engineer, Cloudera >
