area/perfarea/statshelp wanted
仓库指标
- Star
- (27,997 star)
- PR 合并指标
- (平均合并 8天) (30 天内合并 378 个 PR)
描述
I've been debugging a perf problem at Lyft and the dispatcher loop stats are indeed incredibly useful. We should make these stats usable with statsd. The best way I can see to do this is to potentially have the ability to tag a histogram as "high rate" so that it gets ignored by statsd for individual flushes, and then statsd could potentially flush high rate histograms as gauges with various percentiles dumped?