mixed_instances_policy for managed node groups

[v1nc3nt27]

If I understand correctly, currently it is only possible to specify mixed_instances_policy for self-managed worker groups. For managed node groups, it is only possible to set all instances to either ON_DEMAND or SPOT. As mixed_instances_policy is an option for managed node groups in AWS, it would be a nice feature to add.

[daroga0002]

for self managed workers you can check this example

terraform-aws-eks/examples/spot_instances/main.tf

Line 53 in e3216e3

override_instance_types = ["m5.large", "m5a.large", "m5d.large", "m5ad.large"]

[v1nc3nt27]

@daroga0002 I'm sorry, I was too unspecific and forgot to mention that I refer to instances_distribution in mixed_instances_policy of worker groups. If I understand correctly, only worker groups can specify this and I was wondering if there is a reason why this is not possible for managed node groups. Currently, I'd have to manually change it for every node group via AWS.

[daroga0002]

I think you can also make it over node groups as here is some code:

terraform-aws-eks/modules/node_groups/node_groups.tf

Line 19 in cd01601

instance_types = !each.value["set_instance_types_on_lt"] ? each.value["instance_types"] : null

I dont remember exactly so this require some testing

[v1nc3nt27]

@daroga0002 Yup, these are the instance types. They are working (at least they are listed on AWS in the node groups). I'm refering to the instances_distribution which is the percentage of SPOT and On-Demand instances you want to launch. If I see correctly, it's only capacity_type that can be chosen and there the valid variables are SPOT and ON_DEMAND but you cannot mix them. Not sure whether there is a reason why this is not possible, because on AWS you can definitely select it for managed node groups.

[daroga0002]

Ok, so mixed capacity type instances seems be not available under Node groups.

Fully AWS managed:

delivering own launch template:

If you think it is achievable then please feel free to experiment and create some PR.

[v1nc3nt27]

Yes, it seems it's not available if you add a node group via EKS, but you can do so in the automatically created auto scaling group. This cannot be influenced via terraform? I'm pretty new to this so please excuse my ignorance.

[daroga0002]

nope, AWS Node groups have own api which on backend is using a AWS autoscaling, but you cannot influence it.

Any changes directly to autoscaling is just a unsupported by AWS hack

There was some similar discussion under #1525

[v1nc3nt27]

Ah, that's too bad. At this point, I don't want to switch to worker groups so I'll probably manually edit the ASG, I guess. Well, thanks a lot for clarifying it!

[github-actions]

I'm going to lock this issue because it has been closed for 30 days ⏳. This helps our maintainers find and focus on the active issues. If you have found a problem that seems similar to this, please open a new issue and complete the issue template so we can capture all the details necessary to investigate further.