Allow configuring a range for jitter to be within #17
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Previously, jitter would only be available within a single second of the desired interval - eg, if we had jitter enabled and our next interval was 4.5 seconds, the actual calculated interval would be uniformly distributed between 4.5s and 5.5s. The issue with this approach is that it doesn't let callers spread load out efficiently over long timescales - eg, if i have a fleet of millions of agents set to ping every ten seconds, spreading those pings out over one second is only a little better than having them hit all at once - we'd ideally like to be able to spread those pings out over a larger area of the ping interval. This commit enables us to do just that - we can configure a custom range for jitter. To use our previous example, if we set a jitter range of (-1s, 3s], the actual intervals would be uniformly distributed between 3.5s -and 7.5s.
CerealBoy
approved these changes
Jan 9, 2025
2 tasks
moskyb
added a commit
that referenced
this pull request
Jan 13, 2025
The `roko.Retrier`'s `String()` method returns a message about the state of the retry loop, generally something of the form `Attempt 3/19 Retrying in 6s`. Previously, when jitter was enabled for a retrier, the call to `NextInterval()` within the `String()` method and the actual retry loop itself would be totally independent, and so if random jitter was enabled, they'd return different results - ie, `String()` would say `Attempt 3/19 Retrying in 5.42038947s`, but the retry loop would calculate a totally different value for jitter, and wait a longer or shorter amount of time. This wasn't so bad when jitter was limited to one second - `String()` might say `nest attempt in 5.01s`, and then the retrier would actually sleep for `5.99s`, but what's a second or two between friends, right? Unfortunately, in [#17](#17), some jabroni (me) added the ability for users to customise the range that jitter falls within, so it's now possible for `String()` to lie fairly significantly about the amount of time it's going to wait till the next request. If we configured a jitter range of [-10s, 30s] for our jitter range, the next interval reported by `String()` could be off by as much as 40 seconds, which is pretty bad. This PR fixes this issue by refactoring `roko.Retrier` a little bit - now, at the start of the retry loop - before the callback is called - and stored in the state of the retrier. This means that any accesses to the next interval time will be consistent, even with jitter enabled.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Previously, jitter would only be available within a single second of the desired interval - eg, if we had jitter enabled and our next interval was 4.5 seconds, the actual calculated interval would be uniformly distributed between 4.5s and 5.5s.
The issue with this approach is that it doesn't let callers spread load out efficiently over long timescales - eg, if i have a fleet of millions of agents set to ping every ten seconds, spreading those pings out over one second is only a little better than having them hit all at once - we'd ideally like to be able to spread those pings out over a larger area of the ping interval.
This commit PR us to do just that - we can configure a custom range for jitter. To use our previous example, if we set a jitter range of (-1s, 3s], the actual intervals would be uniformly distributed between 3.5s -and 7.5s.