New docs: idioms for testing Onyx code / having a very smooth repl experience - https://github.com/onyx-platform/onyx/blob/develop/doc/user-guide/testing-onyx-jobs.md

This is awesome. It was taking too much time to reload everything.

hey @lucasbradstreet turns out DataDog’s agent has a statsd implementation, so we actually don’t need riemann at all… if we can ship from Onyx to statsd directly. are you aware of anyone who might already be doing this?

otherwise, we might be positioned to contribute something back, finally

Ah cool. I don’t know anyone who is doing so

I refactored the metrics plugin recently, so it should be very little work for you to do so

All you need to do is supply your own sender-fn: :metrics/sender-fn :onyx.lifecycle.metrics.timbre/timbre-sender

Which looks something like this: https://github.com/onyx-platform/onyx-metrics/blob/0.8.x/src/onyx/lifecycle/metrics/riemann.clj

You’ll probably have to transform the metrics-msg values you’re taking off the channel into the appropriate form

great. sounds like we need to upgrade to 0.8.1 first?

Yeah, you probably should

@lucasbradstreet: hey just a quick question about metrics, is there any reason there isn't a 'max' for the past 60s option for latency? would you accept a PR for one?

think I need you to clarify that a little, what do you mean by max?

For latency measurements we record each latency measurement for 10 seconds, then calculate all the percentiles of the recordings, then clear the measurer

instead of stopping at the 99.9th percentile of the last 60s, include the maximum value for the last 60s aswell

if you’re seeing a 60s option / maximums, etc, it might be a feature of datadog

Ah, I see what you mean

I think we only do it over 10s though?

graphing the max is more useful apparently (according to guys like Gil Tene)

https://www.youtube.com/watch?v=9MKY4KypBzg

Ah yes, I think I’ve read some of his other stuff.

should be pretty simple to just include the max in all of those right? i'd be happy to give you a PR for that

ahh yeah i see its over the last 10s

We use this to measure: https://github.com/aphyr/interval-metrics/blob/master/src/interval_metrics/core.clj

was mistaking the 60s for throughput

I think if you can get a max out of the data structure we put the measurements in, you’d be sweet

I’ll take a patch for that

https://github.com/aphyr/interval-metrics/blob/master/src/interval_metrics/core.clj#L234

yeah 😄

sweet thanks!

This is the data structure we use for latency https://github.com/aphyr/interval-metrics/blob/master/src/interval_metrics/core.clj#L194

@lucasbradstreet: would you take a PR for a statsd writer?

absolutely

yay. we get to contribute back!

😄 go team

hahaha

we’re starting to colour inside the lines finally 😁

@robert-stuttaford, @greywolve: 0.8.2 (and plugins 0.8.2.0) are all releasing now

metrics was released with the max latency improvement you PR’d. Thanks.