Introduction and Question

[Zachary Musgrave]

Hi everyone,

My name's Zach and I'm an infrastructure engineer at Yelp. We're looking into a new metrics pipeline solution and have been pretty impressed with Metrics 2.0 so far! Nice work, it's definitely a huge improvement over vanilla statsd. Plus, Go!

One quick question, if anyone would be so kind: Etsy's statsd includes a UDP proxy that can sit in front of a cluster of statsd nodes and shard out metric names, so windowing/aggregation routines make sense. Has anyone built something similar for statsdaemon as yet? Or is there some other architectural pattern that I'm missing?

Best,

Zach

[Dieter Plaetinck]

pasting here for posterity
16:37:19 < Dieterbe> _ztm: hey
16:37:27 < _ztm> hi!
16:38:10 < _ztm> Dieterbe: i was just curious if a statsdaemon-compatible UDP proxy existed?
16:38:19 < _ztm> like the one statsd has
16:51:39 < _ztm> Dieterbe?
16:52:53 < Dieterbe> you could just use the same one, _ztm
16:53:11 < _ztm> Dieterbe, i was under the impression that the wire format was different?
16:55:05 < Dieterbe> statsdaemon supports interpreting the metric keys and acting differently for metrics2.0 metrics, but the input format of
                     the packets is the same
17:04:15 < _ztm> Dieterbe: i see. kind of confused then between statsd name:value|type and the space-separated version in metrics20.org/spec
17:04:25 < _ztm> tags  meta_tags value timestamp
17:06:46 < _ztm> oh, or i guess that's the format that gets sent to graphite
17:07:21 < Dieterbe> yep
17:15:59 < _ztm> oh ok. thanks!

--
You received this message because you are subscribed to the Google Groups "metrics2.0" group.
To unsubscribe from this group and stop receiving emails from it, send an email to metrics20+...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.