code
code
> (e.g. using visit pattern) as well as assure as that this is the same or better. I have tried and it make performance a little worse.
Previous: ``` BM_AllCountersJson/per_endpoint_stats_disabled_mean 592 ms 592 ms 10 output per iteration: 135789011 BM_AllCountersJson/per_endpoint_stats_enabled_mean 967 ms 967 ms 10 output per iteration: 180082411 BM_UsedCountersJson/per_endpoint_stats_disabled_mean 13.2 ms 13.2 ms 10 output per...
> THe benchmark is the inlined version? Can you show it against the current version? @jmarantz updated.
> Was your machine totally idle during both tests? Or do you think there's a perf improvement in your template version? > > /wait > Yeah, I am sure it's...
> Looks like you still have CI issues; ping me and I'll reapprove when that's sorted. rbac is flaky. Help it's fine this time.
Three weeks past. This revert will also introduce a unexpected/break API change. cc @envoyproxy/api-shepherds Or could we disable the test temporarily if that is not an actual bug? cc @yangminzhu
Seems the @yangminzhu has no activity on GitHub recently. Then, let's revert it directly if no explicit objection from @envoyproxy/api-shepherds
I will insist on my previous point: use as less as possible header to do this work. The `endpoint-load-metrics-bin` is supported to keep compatibility with gRPC. But for new format,...