Hello !
I have the same problem, it's blocking all our devs since we're all working on IntelliJ.
Thanks for the help on this !

Thank you for the report! Are you accessing the IDE through the recents screen? I think there is an issue where the SSH config is not regenerated from there, so we should make it do that.

As a workaround, it should regenerate properly if you add the connection again by going through the "Connect to Coder" or launching from the Coder dashboard.

If that still does not work, let me know.

Hello, we already use the "Connect to coder" way. we tried resetting everything without success.

Is there anything we can enable to log what is going wrong? thankyou.

PS: BTW your plugin was working very good before, great job!

Thank you for the kind words.

If you could gather the logs from Gateway (using the cog icon in the lower left) and send them to me (either via email or you can upload them here) I will take a look and see what I can find.

It is strange though, if the blocks are getting replaced (you are talking about the # --- START CODER JETBRAINS blocks right?) then I am not sure how the --bg host is being skipped, there are no conditionals in that code as far as I can see. 🤔

Here is the issue:

2024-09-12 20:39:08,120 [1330103] INFO - CoderCLIManager - No workspaces and no existing config blocks to remove

related to this : https://github.com/coder/jetbrains-coder/blob/main/src/main/kotlin/com/coder/gateway/cli/CoderCLIManager.kt#L324

from my understanding .ssh/config is not written so the "fake" host cannot be reached.

N.B: we have several agents per workspace ( we use resources to nest multiples apps per workspace )

Fun fact: old workspaces with old coder agents (2.10) are still working. Is there any BC break on agent that could explain this issue ?

api version was 2.0 in 2.10 coder agent now it's 2.2

OMG. the issue is on our side,in workspace configuration. thank you very much for your time. sorry

Glad you got it figured out!

Hi, I figured out why! the issue is related to the fact that you are relying on /api/v2/templateversion/ID/resource to get workspace resources. We have conditional parameters in tf template that make some resources unavailable by default so they are not evaluated.
Why don't you get running resources informations from /api/v2/workspace api call ?

Oh interesting!

The reason we use the resource endpoint is that the API does not return resources for workspaces that are off, so when we configure SSH it would be missing all the agents from off workspaces, which could break connections in the recent connections view (say you try to connect to an agent from the recents view from a workspace that is off).

The CLI does the same thing when you run coder config-ssh, the plugin mimics that.

I thought we only did that if the workspace was off though, which I think is what you were suggesting.

Ahh apparently we just always fetch them, but we should use the resources in the workspaces call if the workspace is running, rather than trying to fetch them from resources since we already have them. I will try to get this issue on the roadmap next week.

#489