Windows agent connection issue


I have a weird problem connecting Windows agent agent to a linux controller server. The weird part is that another similar Windows server connects with same settings in configuration xml file. Only differences are secret key and node name.

Jenkins version: 2.440.2
Agent versions on both windows servers:
Remoting version: 3206.vb_15dcf73f6a_9

But then I found that besides fixed port the server getting connection also has https port open in firewall to the controller Jenkins and the non-working doesn’t. Fixed port seems to work on both servers when tested.

I understood from documentation that only JNLP port is needed for connection but could the agent still try https port for request?

Forcing fixed port to connection string in configuration didn’t help for the issue although then netstat -a command showed CONNECTION ESTABLISHED.

Both local Windows firewalls and linux local firewall has similar settings for both slaves. Enterprise firewall has still that https-port closed on non-working server and is “slow” to get open.

Logs on controller says errors like:
Apr 11, 2024 3:32:46 PM INFO hudson.TcpSlaveAgentListener$ConnectionHandler run
Connection #34 from /...:50154 failed: null

Afaik the agent will first contact the controller on the https port to fetch the jnlp settings. You can see that already from the url that you pass to the command which doesn’t contain the jnlp port

1 Like

Instead you can also try to use websockets, then you only need the https port and no longer the jnlp port.

mawinter69, thanks for replies.

I also tried later winsocket but didn’t get it connect either. Now the https-port is opened and when I went to see the node status it was full operating already. I was sure it is about it but saw not a mention anywhere about it.