log something useful when dry-running cluster scaling policies

[erulabs]

When using dry-run for cluster autoscaling, it's easy to get a bit confused because we see no output whatsoever!

The change means that AWS/GKE/Azure cluster autoscaling will print an INFO log letting us know what would have happened.

[hashicorp-cla]

All committers have signed the CLA.

[erulabs]

Ah, it seems that I should have tried TRACE level logging first:

nomad-autoscaler/plugins/builtin/strategy/target-value/plugin/plugin.go

Lines 136 to 142 in 04c6cf6

	// Log at trace level the details of the strategy calculation. This is
	// helpful in ultra-debugging situations when there is a need to understand
	// all the calculations made.
	s.logger.Trace("calculated scaling strategy results",
	"check_name", eval.Check.Name, "current_count", count, "new_count", newCount,
	"metric_value", metric.Value, "metric_time", metric.Timestamp, "factor", factor,
	"direction", eval.Action.Direction)

I'd suggest that this is probably the most important thing nomad-autoscaler could log! Particularly in dry-run mode, it's very useful to know what would happen and why.

I'll change these to INFO level logs and that ought to be enough for me!

[lgfa29]

Hi @erulabs 👋

Thanks for the PR, but yeah, we have log a lot of information at TRACE level because the Autoscaler can be quite chatty depending on how many policies you have and how often they are evaluated.

But I agree that they are easy to miss, specially since the default level is INFO. I created #624 to describe the problem and also propose some options.

Feel free to add any more comments there or take on it if you are interested 🙂