Intermittent Agent Disconnection During Long-Running Jenkins Pipeline on Windows Agent

[Ravindra Verma]

Hello Jenkins Community,

I am encountering a recurring issue with a Jenkins pipeline running on a Windows agent. The pipeline typically takes around 4 hours to complete, but it intermittently disconnects midway with the following error:

FATAL: command execution failed
java.nio.channels.ClosedChannelException
at jenkins.agents.WebSocketAgents$Session.closed(WebSocketAgents.java:157)
...
Caused: java.io.IOException: Backing channel 'AWS-TEST-AUTOMATION' is disconnected.
at hudson.remoting.RemoteInvocationHandler.channelOrFail(RemoteInvocationHandler.java:215)

Key Observations:

• The issue occurs unpredictably but consistently for long-running builds (4+ hours).
• The agent is configured using a WebSocket connection, which seems to fail as indicated by the stack trace.

Steps Taken:

1. Verified network stability between the Jenkins controller and the Windows agent.
2. Recreated the agent and the pipeline in a test environment, but the issue persists intermittently.
3. Checked resource utilization (CPU, memory, etc.) on both the Jenkins controller and the agent during pipeline execution; no significant anomalies observed.

Request for Assistance:

1. Has anyone experienced similar disconnections during long-running builds?
2. Are there specific configurations or best practices for improving the stability of WebSocket-based agent connections, especially for pipelines with extended runtimes?
3. Could this be related to the agent timeout settings, or should we investigate other areas such as the Jenkins remoting layer or JVM configurations?
4. Are there tools or logs I can analyze further to identify the root cause?

Your guidance and expertise would be greatly appreciated. I am open to any suggestions for debugging or mitigating this issue.

Thank you for your support!