1120988 - [research] look into loop ratelimiting

Status: RESOLVED WONTFIX

(Input Graveyard :: Submission, defect, P3)

Description

[Will Kahn-Greene [:willkg] ET needinfo? me]

The Loop in-product feedback form sends the same description pretty often. We use the description and the ip address as a key for rate-limiting. Since Loop is sending the same description often, it's possible that users sending feedback for Loop are hitting ratelimiting and we're losing feedback we shouldn't be losing.

This bug covers investigating whether we're losing Loop feedback because of ratelimiting and Loop sending the same description.

If we are, this bug also covers figuring out what our options are and writing up a bug to fix it.

Making this a P1 for 2015q1.

Comment 1

[Mark Banner (:standard8)]

(In reply to Will Kahn-Greene [:willkg] from comment #0)
> The Loop in-product feedback form sends the same description pretty often.
> We use the description and the ip address as a key for rate-limiting. Since
> Loop is sending the same description often, it's possible that users sending
> feedback for Loop are hitting ratelimiting and we're losing feedback we
> shouldn't be losing.
> 
> This bug covers investigating whether we're losing Loop feedback because of
> ratelimiting and Loop sending the same description.

We're definitely seeing some form of rate limiting or some limiting happening - I've seen messages console messages for other bugs along the lines of:

"Error encountered while submitting feedback" Error: Error posting user feedback data: 429 TOO MANY REQUESTS; Request was throttled.Expected available in 600 seconds.

Comment 2

[Will Kahn-Greene [:willkg] ET needinfo? me]

When you say "for other bugs along the lines of", what does that mean? Is this in a QA scenario or in the real world?

If you're running tests over and over again from the same ip address, you'll definitely hit the rate-limiting. That's expected.

This bug is about whether real users are hitting the ratelimiting.

Comment 3

[Mark Banner (:standard8)]

Ah yes, it could have been from someone that was doing QA-like testing, so its possible that they would be limited.

Comment 4

[Will Kahn-Greene [:willkg] ET needinfo? me]

Dropping the priority on this. It's probably the case we're hitting rate-limiting, but with the current feedback it's unlikely to affect analysis so it's not crucial someone look into it right now.

Comment 5

[Will Kahn-Greene [:willkg] ET needinfo? me]

Pushing off to next quarter.

Comment 6

[Will Kahn-Greene [:willkg] ET needinfo? me]

The Firefox 36 release shows a huge spike in Loop feedback as well as a corresponding spike in double-submits from the API.

I'm pretty sure this isn't a coincidence and that loop is affected by rate-limiting. However, it'll only affect feedback where the description is set by the Loop in-product feedback form as opposed to feedback the user typed in.

Comment 7

[Will Kahn-Greene [:willkg] ET needinfo? me]

I'm going to bump this back to this quarter to mull over it.

Comment 8

[Will Kahn-Greene [:willkg] ET needinfo? me]

Ran out of time this quarter. Pushing it off.

Comment 9

[Will Kahn-Greene [:willkg] ET needinfo? me]

Hello (aka Loop) doesn't use Input anymore. They're the only project that was sending us duplicate descriptions, so I'm declaring this a non-issue again and marking it as WONTFIX.