Hello Sandeep, As long as you have enabled short-circuit read as per the documentation [1], I expect any Hadoop process will take advantage of it while reading a local replica. However, short-circuit read will not completely eliminate TCP connection activity to the DataNode. There will still be a TCP connection from the client to the DataNode to perform a handshake and establish the Unix domain socket. This is a very small payload though compared to the transfer of block data over the Unix domain socket.
[1] http://hadoop.apache.org/docs/r2.7.1/hadoop-project-dist/hadoop-hdfs/ShortCircuitLocalReads.html --Chris Nauroth From: sandeep das <yarnhad...@gmail.com<mailto:yarnhad...@gmail.com>> Reply-To: "user@hadoop.apache.org<mailto:user@hadoop.apache.org>" <user@hadoop.apache.org<mailto:user@hadoop.apache.org>> Date: Wednesday, November 18, 2015 at 10:44 PM To: "user@hadoop.apache.org<mailto:user@hadoop.apache.org>" <user@hadoop.apache.org<mailto:user@hadoop.apache.org>> Subject: Yarn application reading from Data node using short-circuit. Hi, I was going through some benchmarking and realized that there are lots of TCP connections are initiated while running my PIG jobs over YARN(MR2). These TCP connections are related to data node. Although short-circuit is enabled in my data nodes but still a lot TCP connections are being created. I wanted to check that how can we enable YARN applicationMaster to read data from Data node using short-circuits i.e. unix domain sockets. I believe that will improve the performance of our jobs. Can someone please help to understand how can I make sure that MR2 jobs created by PIG scripts are reading data from Data node using short-circuit instead of TCP connections? Regards, Sandeep
