AWS: cache instance requirements

[alexanderConstantinescu]

What type of PR is this?

We can discuss this, but I am going to go ahead and label this:

/kind bug

since it forces users of the CA to re-configure the CA with a work-around that is sub-optimal, see below for more information.

What this PR does / why we need it:

The cluster-autoscaler on AWS currently builds a representation of the Node object during each scale up/down computation. This computation is performed depending on the scan-interval defined. Parts of this computation requires setting the Node capacity. The capacity is built by fetching instance requirements from the AWS API. For clusters which use mixed instance policy and which specify a launch template, the API call made is DescribeLaunchTemplateVersions. This information does not need to be fetched during each scale up/down computation, but can instead be cached along with all the other information required by the cluster-autoscaler running on AWS. By doing this the cluster-autoscaler can reduce the amount of API calls performed. If the scan-interval is 10s; the amount of API calls to this endpoint will be: 6 calls per minute, or 8640 calls per day. This obviously incurs unnecessary costs for people operating AWS clusters at scale and moreover: only leaves them with the possibility of increasing the scan-interal as to reduce the amount of API calls. This is not great since that has secondary impacts on how quickly the CA can react to changes on the cluster. The current refreshInterval is set to one minute and defines the interval at which the CA refreshes its cache, so by caching the instance requirements we can make sure that this API endpoint is only called once per minute, regardless of the scan-interval chosen.

This PR therefore caches the instance requirements as done with all the other MixedInstancesPolicy data, and uses the cached data when executing buildNodeFromTemplate.

Which issue(s) this PR fixes:

Fixes #

Special notes for your reviewer:

Does this PR introduce a user-facing change?

None

Additional documentation e.g., KEPs (Kubernetes Enhancement Proposals), usage docs, etc.:

[gjtempleton]

/assign @gjtempleton

[gjtempleton]

Thanks!
/lgtm
/approve

[k8s-ci-robot]

[APPROVALNOTIFIER] This PR is APPROVED

This pull-request has been approved by: alexanderConstantinescu, gjtempleton

The full list of commands accepted by this bot can be found here.

The pull request process is described here

Needs approval from an approver in each of these files:

• cluster-autoscaler/cloudprovider/aws/OWNERS [gjtempleton]

Approvers can indicate their approval by writing /approve in a comment
Approvers can cancel approval by writing /approve cancel in a comment