Network retries #559
Network retries #559
Conversation
There was a problem hiding this comment.
stopping review for now, thought I'd share these comments
ptal @paulasjes-stripe
There was a problem hiding this comment.
This is looking great! Few further nits on tests etc, pretty sure this should be ready to ship thereafter!
ptal @paulasjes-stripe
| }); | ||
|
|
||
| describe('_getSleepTimeInMS', function() { | ||
| it('should not exceed the maximum or minimum values', function() { |
There was a problem hiding this comment.
Call me a cranky fellow, but having randomness/fuzzing in tests like this makes me uncomfortable (even though I expect it to pass in all cases). Imagine we had an exceedingly rare edge-case bug, such as failing when Math.random() produces exactly 0.0 - it'd occasionally fail in CI, which is frustrating. Not the end of the world in stripe-node (we can just hit rebuild, and it doesn't block other things from getting released).
To be honest, I'm not sure this test is that important; thoughts on just deleting it?
There was a problem hiding this comment.
Well I think it's still useful to ensure that the sleep time is a value we'd expect.
I'd prefer a test that very rarely fails over a edge-case bug over never finding the edge-case bug at all. Especially since the Math.random() call is in our production code and not in the test. Also it's nice to have more code coverage.
|
@ob-stripe Done! We now mock |
paulasjes-stripe
assigned paulasjes-stripe and ob-stripe and unassigned paulasjes-stripe
Nice! Thanks @paulasjes-stripe. Reassigning to @brandur-stripe and @rattrayalex-stripe for the final decision regarding the exact retry conditions, but this looks nearly ready to 🚢 |
|
Okay, Alex and I had a long talk over lunch following a long talk yesterday, and we've decided to not retry on 429s for the time being, but allowing for a system that eventually does pending a few server-side improvements that we'd like to make. I just pushed a commit to take 429s out of the retry equation. r? @rattrayalex-stripe Mind taking a quick look at that? Thanks! |
|
(Also, just wanted to note that I'm really happy to see this come into stripe-node. Thanks for the implementation @paulasjes-stripe!) |
stripe-ci
assigned paulasjes-stripe and unassigned brandur-stripe and rattrayalex-stripe
|
@rattrayalex-stripe re-assigning to you to release! |
|
Can we squash the commits before merging please? :) |
|
Thanks for the reminder - will squash and merge before releasing sometime later today. |
r? @rattrayalex-stripe
cc @stripe/api-libraries
Fixes #558
Adds network retry functionality. The default retry amount is 0 but can be changed via
stripe.setMaxNetworkRetries(n). I modeled this after a similar change instripe-php: https://github.com/stripe/stripe-php/pull/428/filesIn the case of a
POSTrequest, an idempotency key is generated if not already provided.This also adds the
uuidandnocklibraries. The former is for generating idempotency keys, there already is a similar function used in testing however according to our docs [0]uuid/v4should ideally be used for real requests. I usenockto mock API responses and failures.[0]: https://stripe.com/docs/api/idempotent_requests