[HADOOP-9107] Hadoop IPC client eats InterruptedException and sets interrupt on the thread which is not documented

• Type: Bug
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Affects Version/s: 1.1.0, 2.0.2-alpha
• Component/s: ipc
• Assignee: Unassigned
• Reporter: Hari Shreedharan
• Created: 29/Nov/12 21:02
• Updated: 02/May/13 02:30

Description

This code in Client.java looks fishy:

  public Writable call(RPC.RpcKind rpcKind, Writable rpcRequest,
      ConnectionId remoteId) throws InterruptedException, IOException {
    Call call = new Call(rpcKind, rpcRequest);
    Connection connection = getConnection(remoteId, call);
    connection.sendParam(call);                 // send the parameter
    boolean interrupted = false;
    synchronized (call) {
      while (!call.done) {
        try {
          call.wait();                           // wait for the result
        } catch (InterruptedException ie) {
          // save the fact that we were interrupted
          interrupted = true;
        }
      }

      if (interrupted) {
        // set the interrupt flag now that we are done waiting
        Thread.currentThread().interrupt();
      }

      if (call.error != null) {
        if (call.error instanceof RemoteException) {
          call.error.fillInStackTrace();
          throw call.error;
        } else { // local exception
          InetSocketAddress address = connection.getRemoteAddress();
          throw NetUtils.wrapException(address.getHostName(),
                  address.getPort(),
                  NetUtils.getHostname(),
                  0,
                  call.error);
        }
      } else {
        return call.getRpcResult();
      }
    }
  }

Blocking calls are expected to throw InterruptedException if that is interrupted. Also it seems like this method waits on the call objects even if it is interrupted. Currently, this method does not throw an InterruptedException, nor is it documented that this method interrupts the thread calling it. If it is interrupted, this method should still throw InterruptedException, it should not matter if the call was successful or not.

This is a major issue for clients which do not call this directly, but call HDFS client API methods to write to HDFS, which may be interrupted by the client due to timeouts, but does not throw InterruptedException. Any HDFS client calls can interrupt the thread but it is not documented anywhere.

Issue Links

• is related to: HADOOP-6762 exception while doing RPC I/O closes channel
• relates to: FLUME-1748 HDFS Sink should check if the thread is interrupted before performing any HDFS operations
• relates to: HADOOP-6221 RPC Client operations cannot be interrupted