weightedroundrobin: Implements Slow Start (gRFC A100) [WiP]

[davinci26]

Implements slow start functionality as described in gRFC A100

I still need to do some work on tests and do an e2e experiment with xDS to ensure that things are working as expected, but the core implementation is in a state that can be reviewed.

At this phase I'm mainly looking for feedback on the overall approach and design and in the test code because I am worried that I might be missing some test utility that would make testing with time easier.

Pending Items before this is ready to be merged:

• Add more unit tests for slow start config cases
• Manually verify with an xDS experiment that slow start is working as expected

RELEASE NOTES:

• weightedroundrobin: Implements slow start functionality as described in gRFC A100

[codecov]

Codecov Report

❌ Patch coverage is 64.28571% with 15 lines in your changes missing coverage. Please review.
✅ Project coverage is 83.32%. Comparing base (363018c) to head (e8884a6).
⚠️ Report is 7 commits behind head on master.

Files with missing lines	Patch %	Lines
balancer/weightedroundrobin/balancer.go	72.97%	5 Missing and 5 partials ⚠️
...xds/xdsclient/xdslbregistry/converter/converter.go	0.00%	4 Missing and 1 partial ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #8689      +/-   ##
==========================================
- Coverage   83.43%   83.32%   -0.12%     
==========================================
  Files         415      416       +1     
  Lines       32195    32342     +147     
==========================================
+ Hits        26863    26949      +86     
- Misses       3980     4016      +36     
- Partials     1352     1377      +25     

Files with missing lines	Coverage Δ
...xds/xdsclient/xdslbregistry/converter/converter.go	70.00% <0.00%> (-3.34%)	⬇️
balancer/weightedroundrobin/balancer.go	83.94% <72.97%> (-1.61%)	⬇️

... and 34 files with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[davinci26]

imports a commit because the proto is not in a release yet, validate this is the correct version for all deps

[davinci26]

Don't want to commit this code because it introduces a new dependencies but wanted to make sure the aggression math is reasonable so I plotted them over time (https://gist.github.com/davinci26/d29bf9633fe9afd3b4164656a8013eb4)

(Note the flatline at start is due to math.Max(timeSinceReady.Seconds(), 1.0))

[davinci26]

Wanted to call some attention here since this is a bit of subtle point, where I made a decision but maybe its worth clarifying in the gRFC

[davinci26]

The implementation should be adjusted based on the convo grpc/proposal#498 (comment)

[easwars]

Apologies for the delay in reviewing this PR. I'm starting on this today and hopefully will have comments soon. Thanks.

[easwars]

@davinci26 : I finally got around to reading the gRFC and I'm ready to start reviewing this. But could you please split this up into smaller PRs. The minimum I would like to see is the following split

• WRR changes alone
• Add metrics to WRR
• xDS integration

I'm going to start reviewing the WRR changes alone here. So, if you could move the other pieces out, we can keep the changes small and have a more meaningful review. Thanks again for your contribution.