accept max_scale_up and max_scale_down for target-value

[brmzkw]

Hey,

The two optional parameters max_scale_up and max_scale_down determine the maximum and minimum number of scaling differences between two evaluations.

It is useful to avoid scaling down or scaling up a job too quickly.

The code is not tested yet. For now, this is a draft pull request to see if you might be interested to include the feature in the target-value plugin.

[hashicorp-cla]

All committers have signed the CLA.

[brmzkw]

I've force-pushed the unit tests. Feel free to reach out if there are any queries!

[lgfa29]

Thank you for the contribution @brmzkw! I really like the idea 🙂

It provides a stepped scaling instead of potentially abrupt changes. I was thinking if we could adopt this configuration more broadly such that it could be used with any strategy:

policy {
  # ...
  max_scale_up   = 10
  max_scale_down = 3
  # ...
}

But looking at the other strategies we have so far they seem more controlled than target-value because they usually define static changes. We can always expand later if we find the need.

Would be able to write some docs for these new config in https://github.com/hashicorp/nomad/blob/main/website/content/tools/autoscaling/plugins/strategy/target-value.mdx?

[brmzkw]

But looking at the other strategies we have so far they seem more controlled than target-value because they usually define static changes. We can always expand later if we find the need.

This is why I implemented it in the target-value plugin. For example, I'm not sure it would make sense to have these parameters for the pass-through or for the fixed-value strategies.

Thanks for the merge, I'll update the documentation today.

Do you know when you intend to make a release?

[lgfa29]

No specific date yet, but we will probably release a new version soon.