Cloud plugin, disconnecting from slave properly

[Stanislav Baiduzhyi]

Hi All,

Another question regarding custom cloud plugins:

I'm creating a custom plugin that creates a node from prototype, uses
the node, discards it (or reverts the state).

I'm starting slave.jar with tcp and connecting to it. My
ComputerLauncher also implements ExecutorListener, and disconnects on
taskCompleted, and terminates/discards the node in afterDisconnect.
But I see huge stack traces about socket being closed.

I've copied the closing procedure from here:
https://github.com/jenkinsci/vsphere-cloud-plugin/blob/master/src/main/java/org/jenkinsci/plugins/vSphereCloudSlave.java#L305

So, what is the correct way to disconnect the channel before node termination?

[Oleg Nenashev]

Where do you see these traces?
"slave.waitUntilOffline();" ensures that channel is null before existing, but I suppose that there could be exceptions from background monitoring threads and hanging jobs (good ones should be gracefully aborted).

воскресенье, 6 сентября 2015 г., 22:20:37 UTC+3 пользователь Stanislav Baiduzhyi написал:

[Kanstantsin Shautsou]

Please, provide logs. 
Stacktrace is expected behaviour because some things in Jenkins works asynchronously and jenkins prints stacktraces even for expected behaviours.