You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
In #406, a new launch method for inbound agents on Linux was introduced.
The documentation states that it requires the TCP port for agent protocol to be set up, static and reachable.
I would like to be able to specify "Websockets" like I can do with permanent agents or Kubernetes agents, so the agent would connect to the controller through HTTP+Websockets to avoid setting up and opening this TCP port (which cannot be load balanced).
Thanks I'm going to try the solution with init script, and if it works I'll send a documentation PR.
Same as #422 (comment): it would we a nice UX if the end user could set up the agent through the UI and have it preconfigured (I guess passing userdata to the VM would help for this one, I'll give it a try as well).
What do you mean by cannot be loadbalanced? You can't load balance Jenkins anyway?
The controller itself: no. But the route from the agent process location to a given controller can have many moving paths and my experience with the JNLP over TCP is such setup is not the most positive one, while websockets has embeded retries mechanisms (also, easier to expose as HTTP(s) already is and less configuration).
What feature do you want to see added?
In #406, a new launch method for inbound agents on Linux was introduced.
The documentation states that it requires the TCP port for agent protocol to be set up, static and reachable.
I would like to be able to specify "Websockets" like I can do with permanent agents or Kubernetes agents, so the agent would connect to the controller through HTTP+Websockets to avoid setting up and opening this TCP port (which cannot be load balanced).
Upstream changes
#406
The text was updated successfully, but these errors were encountered: