Why suddenly the slave node lost connection to the master node in hadoop? - hadoop

Question

I have set up hadoop 2.7.2 upon a cluster with a master(ubuntu 15.10) and two slave(slave2,3) hosted in the master by virtualbox.
I have run several examples like wordcount,it all works ok. But when i try to run my own job,say Myjob, it runs well at first, but after a while, it will definitely be interrupted by this error:
INFO ipc.Client: Retrying connect to server: slave3/xxx.216.227.176(the ip of slave):38046. Already tried 2 time(s); retry policy is RetryUpToMaximumCountWithFixedSleep(maxRetries=3, sleepTime=1000 MILLISECONDS)
Sometimes it would be slave2 ,sometimes would be slave3. And my connection to that slave by ssh shows that the connection is closed by remote.
But the virtualbox shows that slave runs well, and i can ressh to that slave, However all the hadoop process are been killed.
Need to mention that my own job runs longer than the example job.
At first, i think it maybe some error caused by my config file, so, i reinstall hadoop at master and slaves. But the error stills.
So, i think that maybe it is caused by my network config in the slave nodes. So, i changed the last filed of the slave's ip like xxx.xxx.xxx.183 to xxx.xxx.xxx.176 and reinstall hadoop.
And I rerun the job, at this time the job runs longer than usual. But, at the last, when the map stage mostly finished (map 86% reduce 28%), it failed due to the same error!
INFO ipc.Client: Retrying connect to server: slave3/125.xxx.227.xxx(the ip of slave):38046. Already tried 2 time(s); retry policy is RetryUpToMaximumCountWithFixedSleep(maxRetries=3, sleepTime=1000 MILLISECONDS)
Also there is something log in yarn-user-resourcemanager-Master.log :
java.net.ConnectException: Call From Master/xxx.216.227.186 to slave2:44592 failed on connection exception: java.net.ConnectException: refuse to connect; For more details see: http://wiki.apache.org/hadoop/ConnectionRefused
It seems like that the longer time the app runs , the bigger chance it will fail.
Here is my hosts file:
127.0.0.1 localhost
#127.0.1.1 Master
xxx.216.227.186 Master
xxx.216.227.185 slave1# the slave1 has some problem thus do not connect to the cluster
xxx.216.227.176 slave2
xxx.216.227.166 slave3
# The following lines are desirable for IPv6 capable hosts
::1 ip6-localhost ip6-loopback
fe00::0 ip6-localnet
ff00::0 ip6-mcastprefix
ff02::1 ip6-allnodes
ff02::2 ip6-allrouters
Why? How to fix it? Thanks!