How to fix Connection reset by peer message from apache-spark?

Question

I keep getting the the following exception very frequently and I wonder why this is happening? After researching I found I could do .set("spark.submit.deployMode", "nio"); but that did not work either and I am using spark 2.0.0

WARN TransportChannelHandler: Exception in connection from /172.31.3.245:46014
    java.io.IOException: Connection reset by peer
    at sun.nio.ch.FileDispatcherImpl.read0(Native Method)
    at sun.nio.ch.SocketDispatcher.read(SocketDispatcher.java:39)
    at sun.nio.ch.IOUtil.readIntoNativeBuffer(IOUtil.java:223)
    at sun.nio.ch.IOUtil.read(IOUtil.java:192)
    at sun.nio.ch.SocketChannelImpl.read(SocketChannelImpl.java:380)
    at io.netty.buffer.PooledUnsafeDirectByteBuf.setBytes(PooledUnsafeDirectByteBuf.java:221)
    at io.netty.buffer.AbstractByteBuf.writeBytes(AbstractByteBuf.java:898)
    at io.netty.channel.socket.nio.NioSocketChannel.doReadBytes(NioSocketChannel.java:242)
    at io.netty.channel.nio.AbstractNioByteChannel$NioByteUnsafe.read(AbstractNioByteChannel.java:119)
    at io.netty.channel.nio.NioEventLoop.processSelectedKey(NioEventLoop.java:511)
    at io.netty.channel.nio.NioEventLoop.processSelectedKeysOptimized(NioEventLoop.java:468)
    at io.netty.channel.nio.NioEventLoop.processSelectedKeys(NioEventLoop.java:382)
    at io.netty.channel.nio.NioEventLoop.run(NioEventLoop.java:354)
    at io.netty.util.concurrent.SingleThreadEventExecutor$2.run(SingleThreadEventExecutor.java:112)

Answer 1

I was getting the same error even if I tried many things.My job used to get stuck throwing this error after running a very long time. I tried few work around which helped me to resolve. Although, I still get the same error by at least my job runs fine.

1. one reason could be the executors kills themselves thinking that they lost the connection from the master. I added the below configurations in spark-defaults.conf file.

   spark.network.timeout 10000000 spark.executor.heartbeatInterval 10000000 basically,I have increased the network timeout and heartbeat interval

2. The particular step which used to get stuck, I just cached the dataframe that is used for processing (in the step which used to get stuck)

Note:- These are work arounds, I still see the same error in error logs but the my job does not get terminated.