fixing #1012, cluster autoscaler requres a valid LT intance type

[alexei-led]

Kubernetes cluster autoscaler (AWS) requres a valid LT intance type to scale up/down ASG. See #1012

Description

Adding first instance type from distribution instance types as a default LaunchTemplate instance type.

Checklist

• Code compiles correctly (i.e make build)
• Updated tests that cover your change
• All unit tests passing (i.e. make test)
• Manually tested

[errordeveloper]

@alexei-led thanks for this PR! This looks good overall, but I'm just not sure if there are any implications on functionality, would you mind to provide a reference to AWS docs that suggests what does it mean when mixed instances are set along with traditional instance type?

[alexei-led]

@errordeveloper Ilya take a look at AWS LaunchTemplate API docs

The overrides are used to override the instance type specified by the launch template with multiple instance types that can be used to launch On-Demand Instances and Spot Instances.

i.e. when LT has Overrides section the instanceType in LT is ignored. I've also checked this.

[martina-if]

👍 Thanks @alexei-led . I tested this creating 3 nodegroups: all spot instances, all on demand and mixed instances and it seems to work fine.

[errordeveloper]

Let's merge it for 0.2.0. We have 0.1.40 underway, and I'd prefer that to fix exactly one bug (#1003).

[whereisaaron]

@alexei-led @errordeveloper is m5 a good choice for this? Remember that some AZ's don't have m5 instances like ap-southeast-2b. Will that affect being able to use this change in that zone, or will the override prevent any errors.

[alexei-led]

@whereisaaron any non-empty string is a good choice - cluster autoscaler validates only that instanceType is not empty, no actual validation is done :)