[HDFS-9659] EditLogTailerThread to Active Namenode RPC should timeout

• Type: Bug
• Status: Resolved
• Priority: Critical
• Resolution: Fixed
• Affects Version/s: 3.0.0-alpha1
• Fix Version/s: 3.0.0-alpha1
• Component/s: ha, namenode
• Hadoop Flags: Reviewed
• Assignee: Surendra Singh Lilhore
• Reporter: Surendra Singh Lilhore
• Created: 18/Jan/16 12:32
• Updated: 27/Mar/19 17:34
• Resolved: 01/Feb/16 08:42

Description

EditLogTailerThread to Active Namenode RPC doesn't have timeout and it’s removed in HDFS-6440.

When inject the disk slow and consume system IO to the active name node, the nameservice can't switch and this is because SNN not able to stop EditLogTailerThread.

Thread dump from SNN

"IPC Server handler 33 on 25000" #118 daemon prio=5 os_prio=0 tid=0x00007f2384409800 nid=0x26c89 in Object.wait() [0x00007f2376ac7000]
   java.lang.Thread.State: WAITING (on object monitor)
	at java.lang.Object.wait(Native Method)
	at java.lang.Thread.join(Thread.java:1245)
	- locked <0x00000006d517f538> (a org.apache.hadoop.hdfs.server.namenode.ha.EditLogTailer$EditLogTailerThread)
	at java.lang.Thread.join(Thread.java:1319)
	at org.apache.hadoop.hdfs.server.namenode.ha.EditLogTailer.stop(EditLogTailer.java:183)
	at org.apache.hadoop.hdfs.server.namenode.FSNamesystem.stopStandbyServices(FSNamesystem.java:1284)
	at org.apache.hadoop.hdfs.server.namenode.NameNode$NameNodeHAContext.stopStandbyServices(NameNode.java:1852)
	at org.apache.hadoop.hdfs.server.namenode.ha.StandbyState.exitState(StandbyState.java:72)
	at org.apache.hadoop.hdfs.server.namenode.ha.HAState.setStateInternal(HAState.java:62)
	at org.apache.hadoop.hdfs.server.namenode.ha.StandbyState.setState(StandbyState.java:49)
	at org.apache.hadoop.hdfs.server.namenode.NameNode.transitionToActive(NameNode.java:1684)

Thread dump for EditLogTailerThread, it is stuck in NamenodeProtocolTranslatorPB.rollEditLog() rpc call.

"Edit log tailer" #150 prio=5 os_prio=0 tid=0x00007f2395569800 nid=0x26cac in Object.wait() [0x00007f2374aa7000]
   java.lang.Thread.State: WAITING (on object monitor)
	at java.lang.Object.wait(Native Method)
	at java.lang.Object.wait(Object.java:502)
	at org.apache.hadoop.ipc.Client.call(Client.java:1503)
	- locked <0x00000006d581bb90> (a org.apache.hadoop.ipc.Client$Call)
	at org.apache.hadoop.ipc.Client.call(Client.java:1448)
	at org.apache.hadoop.ipc.ProtobufRpcEngine$Invoker.invoke(ProtobufRpcEngine.java:229)
	at com.sun.proxy.$Proxy16.rollEditLog(Unknown Source)
	at org.apache.hadoop.hdfs.protocolPB.NamenodeProtocolTranslatorPB.rollEditLog(NamenodeProtocolTranslatorPB.java:148)
	at org.apache.hadoop.hdfs.server.namenode.ha.EditLogTailer$2.doWork(EditLogTailer.java:301)
	at org.apache.hadoop.hdfs.server.namenode.ha.EditLogTailer$2.doWork(EditLogTailer.java:298)
	at org.apache.hadoop.hdfs.server.namenode.ha.EditLogTailer$MultipleNameNodeProxy.call(EditLogTailer.java:420)

Issue Links

• is broken by: HDFS-6440 Support more than 2 NameNodes
• relates to: HDFS-14391 Backport HDFS-9659 to branch-2