provider/aws: Beanstalk environments, bump the minimum timeout between API calls

[catsby]

This is a band-aid to #6870. Beanstalk environments can take a long time to setup, and if you are managing many of them you can hit API rate limits for your account.

A more permanent solution may be to expose a knob to set a customizable timeout / poll interval, but as I mention in #6870 the design of such a knob hasn't been planed out yet.

[tecnobrat]

Isn't it the delay you want to update?

[catsby]

Delay is the amount of time before we start. MinTimeout is the minimum amount of time between refreshes:

Delay          time.Duration    // Wait this time before starting checks
MinTimeout     time.Duration    // Smallest time to wait before refreshes

I'll be refactoring this PR to expose at least MinTimeout, stay tuned!

[catsby]

10s is our current minimum. 60s seemed like a reasonable max, but we can tune it higher if need be

[tecnobrat]

60s seems like a reasonable max, I'd agree. You don't want atlas jobs taking forever because of someone's value here. Its a fine line.

[tecnobrat]

Also I like poll_interval as the variable, seems to be generic that could be added in another providers / resources in the future.

[phinze]

I'm not sure if MinTimeout does what we want here:

terraform/helper/resource/state.go

Lines 77 to 83 in 2a45b1f

	wait := time.Duration(math.Pow(2, float64(tries))) *
	100 * time.Millisecond
	if wait < conf.MinTimeout {
	wait = conf.MinTimeout
	} else if wait > 10*time.Second {
	wait = 10 * time.Second
	}

Looks like it's capped at 10s based on that conditional.

That's all based on doing exponential backoff. Maybe we need to plumb through a fresh option for a static poll interval if the user requests it?

[catsby]

Paul and I discussed this and found that MinTimeout was in fact "working" as I expected it, however, we discovered that it would only do so for about 10 minutes, until the calculated back off time exceeded the 60s, in which case the polling would return to every 10 seconds 😵

I'm going to take another pass and try adding a separate config option that's preferable to the min timeout

[catsby]

@phinze I refactored this to use poll_interval as a separate thing in WaitForState.

[phinze]

LGTM

[ghost]

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.