chore: use Grpc inproc

[markphelps]

Re: https://www.bwplotka.dev/2025/go-grpc-inprocess-iter/

Switches to an inprocess channel for routing GRPC gateway requests to the GRPC server

This should be much more performant since we don't need to go over the loopback network as described in above blog post ☝🏻

[codecov]

Codecov Report

Attention: Patch coverage is 1.26582% with 78 lines in your changes missing coverage. Please review.

Project coverage is 63.61%. Comparing base (9326370) to head (899e730).
Report is 1 commits behind head on main.

Files with missing lines	Patch %	Lines
internal/cmd/authn.go	0.00%	38 Missing ⚠️
internal/cmd/grpc.go	0.00%	27 Missing ⚠️
cmd/flipt/main.go	0.00%	7 Missing ⚠️
internal/cmd/http.go	0.00%	6 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #3856      +/-   ##
==========================================
- Coverage   64.85%   63.61%   -1.24%     
==========================================
  Files         171      171              
  Lines       17400    17380      -20     
==========================================
- Hits        11284    11057     -227     
- Misses       5418     5657     +239     
+ Partials      698      666      -32     

Flag	Coverage Δ
unittests	63.61% <1.26%> (-1.24%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[markphelps]

Well something doesn't work quite right it seems

[GeorgeMac]

Well something doesn't work quite right it seems

Im guessing something to do with the middleware as its not failing wholesale to perform operations.
Some things are getting through and I can see e.g. errors being reported back as unknown when they have expected concrete status types (e.g. invalid), where there error messages are still as expected. Makes me think for example that the error status middleware is not registered properly.

The ITs used to have the protocol used in the test reported, but I think that is missing from the test output now.
Would be good to know if these failures are e.g. scoped to HTTP through gateway vs just gRPC.

[erka]

Somehow etag breaks the response for snapshot

[markphelps]

Thank you @erka !

[erka]

From a performance perspective, the results are very promising. wrk benchmarking on my laptop indicates a doubling of speed for response time.

[markphelps]

Awesome! My next step was gonna be to benchmark and also do some more testing to make sure everything works correctly

[markphelps]

@erka would you mind sharing your benchmark setup and/or results here in a comment and I'll do the same?

[erka]

I just use go-wrk for testing. Probably it's possible to use wrk and with extra lua script.

go-wrk -M POST -H "Content-Type: application/json" -H "Accept: application/json" -body '{"flagKey":"color-scheme","entityId":"a","context":{"subscription":"active"},"namespaceKey":"default"}' -d 30 http://localhost:8080/evaluate/v1/variant

[markphelps]

Before (main branch)

129625 requests in 29.967058948s, 50.42MB read
Requests/sec:           4325.58
Transfer/sec:           1.68MB
Overall Requests/sec:   4320.24
Overall Transfer/sec:   1.68MB
Fastest Request:        631µs
Avg Req Time:           2.311ms
Slowest Request:        66.859ms
Number of Errors:       0
10%:                    841µs
50%:                    955µs
75%:                    992µs
99%:                    1.022ms
99.9%:                  1.023ms
99.9999%:               1.023ms
99.99999%:              1.023ms
stddev:                 1.096ms

After (this branch)

147780 requests in 29.967327337s, 51.00MB read
Requests/sec:           4931.37
Transfer/sec:           1.70MB
Overall Requests/sec:   4925.62
Overall Transfer/sec:   1.70MB
Fastest Request:        513µs
Avg Req Time:           2.027ms
Slowest Request:        22.795ms
Number of Errors:       0
10%:                    690µs
50%:                    796µs
75%:                    832µs
99%:                    857µs
99.9%:                  858µs
99.9999%:               858µs
99.99999%:              858µs
stddev:                 751µs

[GeorgeMac]

Very nice 👌