ysawa0

+1 on this change. Printing the diff is not as useful as auto-correcting the styling.

Great idea, that makes sense to me.

I don't think there was a decision made to strictly use a certain algorithm. There was work being done in #243 and #193 to implement rolling window but it's gone...

@Pchelolo after testing the latest master with these changes I noticed a big performance degradation with V2 API. Under light load the p999 latency went from 15ms => 22ms which...

Thanks @Pchelolo I tried out the PR and that looks better in pprof. Let me do more long-term tests and get back to you with a definitive answer!

#155 did indeed fix a lot of the issues but after testing ratelimit before the v2/v3 api code was even merged at all, I still see performance degradation. I compared...

@caitong93 I tested with ``` - name: REDIS_PIPELINE_WINDOW value: "0" - name: REDIS_PIPELINE_LIMIT value: "0" ``` this would turn the feature completely off, correct?

Apologizes for delay in this update. Been trying to find the time to do this test the right way. Here's what I saw after trying 3 builds. For Redis, I...

@caitong93 Beautiful. I've run #163 vs pre #137 and they are reporting basically same p999 latency and timeouts! Thank you!

Hi @lkamenkovich, thanks for reporting this. Can you post here the contents of your shadowreader.yml and serverless.yml files? I suspect what is happening is that Serverless is trying (and failing)...