Jenkins Windows slave connection getting terminated with java.nio.channels.ClosedChannelException

I was experiencing a similar error as the OP where the connection to my slave was dropping. The root cause of the issue was not due to a mismatch in Java versions between Jenkins slave and master hosts.

Solution If you are running Jenkins in an EC2 instance on AWS behind an Elastic Load Balancer (ELB), increase the "idle timeout" value under the "attributes" section from the default 60 seconds. I set the new value to 600 and no longer experienced the error.

It appears that if a single command in your build process takes greater than 60 seconds with no log output, the ELB will terminate the session due to idle activity.

Source: https://issues.jenkins-ci.org/browse/JENKINS-44001?focusedCommentId=312412&page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#comment-312412

• I observed the same error after our jenkins master was updated. It is likely due to incompatibility between Java 7 (v80) and latest Java 8.
• Check the java version being used by your master, and the java version of your slave.

• In my case, I am running swarm-client-2.0-jar-with-dependencies.jar on a linux host, and it was using Java 7.

  java version "1.7.0_80" Java(TM) SE Runtime Environment (build 1.7.0_80-b15) Java HotSpot(TM) 64-Bit Server VM (build 24.80-b11, mixed mode)

• Our jenkins master was upgraded and is now running Java 8

  java version "1.8.0_121" Java(TM) SE Runtime Environment (build 1.8.0_121-b13) Java HotSpot(TM) 64-Bit Server VM (build 25.121-b13, mixed mode)

• When the java on the slave was updated to Java 8, the connection issues disappeared.

in addition to the error log in the post, I got also the error log under the jenkins directory in the slave (for me it was C:\jenkins\jenkins-slave.err.log):

JNLP file http://jenkins.domain.com/computer/my_slave_name/slave-agent.jnlp?encrypt=true has invalid arguments: [#####################################, my_slave_name, -workDir, c:\jenkins, -internalDir, remoting, -url, http://jenkins.domain.com/, -headless, -jar-cache, C:\Users\Administrator.jenkins\cache\jars] Most likely a configuration error in the master "-workDir" is not a valid option

my solution:

1)windows slave level: close the services console in the GUI for all users - this is must. from some reason Microsoft is locking installation/removal of windows services

2)windows slave level: kill all java and jenkins-slave processes (if exist)

3)windows slave level: delete the jenkins slave service (if exist) from cmd: sc delete jenkinsslave-c__jenkins /force (in my case)

4)windows slave level: verify that you have java 8 installed: i'm using jdk1.8.0_151 . uninstall all old java version

5)jenkins master ui level: Change the way the Jenkins is connect to the slave under slave configure --> Launch method: Let Jenkins control this Windows slave as a Windows service (instead of Launch agent via Java Web Start)

6) aws level: Increase the aws elb Idle timeout to 600 (from 60) - like @njtman suggested

7)jenkins master ui level: relaunch the agent in jenkins and wait several minutes.

my environment:

jenkins: 2.89.2 , os: windows 2012 R2, java: jdk1.8.0_151