[HADOOP-14262] rpcTimeOut is not set up correctly in Client thus client doesn't time out

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Duplicate
• Assignee: Yongjun Zhang
• Reporter: Yongjun Zhang
• Created: 31/Mar/17 08:09
• Updated: 13/Aug/17 05:12
• Resolved: 13/Aug/17 05:12

Description

NameNodeProxies.createNNProxyWithClientProtocol does

  ClientNamenodeProtocolPB proxy = RPC.getProtocolProxy(
        ClientNamenodeProtocolPB.class, version, address, ugi, conf,
        NetUtils.getDefaultSocketFactory(conf),
        org.apache.hadoop.ipc.Client.getTimeout(conf), defaultPolicy,
        fallbackToSimpleAuth).getProxy();

which calls Client.getTimeOut(conf) to get timeout value.

Client.getTimeOut(conf) doesn't consider IPC_CLIENT_RPC_TIMEOUT_KEY right now. Thus rpcTimeOut doesn't take effect for relevant RPC calls, and they hang!

For example, receiveRpcResponse blocked forever at:

Thread 16127: (state = BLOCKED)                                                                     
 - sun.nio.ch.SocketChannelImpl.readerCleanup() @bci=6, line=279 (Compiled frame)                   
 - sun.nio.ch.SocketChannelImpl.read(java.nio.ByteBuffer) @bci=205, line=390 (Compiled frame)       
 - org.apache.hadoop.net.SocketInputStream$Reader.performIO(java.nio.ByteBuffer) @bci=5, line=57 (Compiled frame)
 - org.apache.hadoop.net.SocketIOWithTimeout.doIO(java.nio.ByteBuffer, int) @bci=35, line=142 (Compiled frame)
 - org.apache.hadoop.net.SocketInputStream.read(java.nio.ByteBuffer) @bci=6, line=161 (Compiled frame)
 - org.apache.hadoop.net.SocketInputStream.read(byte[], int, int) @bci=7, line=131 (Compiled frame) 
 - java.io.FilterInputStream.read(byte[], int, int) @bci=7, line=133 (Compiled frame)               
 - java.io.FilterInputStream.read(byte[], int, int) @bci=7, line=133 (Compiled frame)               
 - org.apache.hadoop.ipc.Client$Connection$PingInputStream.read(byte[], int, int) @bci=4, line=521 (Compiled frame)
 - java.io.BufferedInputStream.fill() @bci=214, line=246 (Compiled frame)                           
 - java.io.BufferedInputStream.read() @bci=12, line=265 (Compiled frame)                            
 - java.io.DataInputStream.readInt() @bci=4, line=387 (Compiled frame)                              
 - org.apache.hadoop.ipc.Client$Connection.receiveRpcResponse() @bci=19, line=1081 (Compiled frame) 
 - org.apache.hadoop.ipc.Client$Connection.run() @bci=62, line=976 (Compiled frame)     

Filing this jira to fix it.

Issue Links

• duplicates: HADOOP-12672 RPC timeout should not override IPC ping interval
• is duplicated by: HADOOP-14198 Should have a way to let PingInputStream to abort