"allow_scheduling_on_control_plane" disabled if "using_klipper_lb" is enabled.

[tijnhendrikson]

Hi,

Thanks for the effort you've put into this project!

I have a question regarding the "using_klipper_lb" option. I don't fully understand what the relation is between the "using_klipper_lb" option and to allow scheduling on control_plane.

The behavior seems unexpected to me, I was in the impression this was only controlled by the "allow_scheduling_on_control_plane". Is there a specific reason behind it?

I'm referring to this part:

terraform-hcloud-kube-hetzner/locals.tf

Line 82 in b008bcf

allow_scheduling_on_control_plane = local.using_klipper_lb ? true : var.allow_scheduling_on_control_plane

Cheers!

[mysticaltech]

Thanks, @tijnhendrikson! Very good find... It is not correct indeed. What happened is that initially, we only activated Klipper for single-node clusters, so we had to allow scheduling on the control plane for obvious reasons, but then I added the option of using Klipper even for multi-node clusters, and the old logic stayed.

So in line 82 above, we just need to replace local.using_killer_lb with local.is_single_node_cluster; please do not hesitate to submit a PR; I will merge it immediately. Otherwise, I will do it as soon as I can.

[mysticaltech]

@tijnhendrikson Did the fix! And released in v1.2.3. Thanks again!

[tijnhendrikson]

No way, you're too quick! Thanks

[schlichtanders]

I would like to turn off scheduling on control plane but still use klipper lb (or no extra load balancer) - just like the title of this discussion states.

The current implementation makes this impossible indeed, always if klipper_lb is active, scheduling is allowed on the control plane notes:

terraform-hcloud-kube-hetzner/locals.tf

Line 109 in 209d48a

allow_scheduling_on_control_plane = (local.is_single_node_cluster || local.using_klipper_lb) ? true : var.allow_scheduling_on_control_plane

so the discussion is not solved @tijnhendrikson can you unmark the answer so that this discussion is again open?

[schlichtanders]

@mysticaltech you seem to be the one who selected the answer - can you please unselect it, as this discussion is NOT solved