Fork me on GitHub
#onyx
<
2017-08-18
>
wildermuthn15:08:08

@michaeldrogalis Is your answer here still correct? https://groups.google.com/d/msg/onyx-user/FWUS7r6C4ww/7bA7cXnVAwAJ I was using emit yesterday, but if the original task function returns nothing, then the window never gets populated, and so the trigger emit has no data to send downstream. But if I use ‘identity’ (like in the twitter example), then I have both the original segment and the emitted segment passed downstream. My workaround was to have emit wrap the segment and then use a flow condition to prevent the original task fn from sending segments. But… seems like I’m doing something wrong.

michaeldrogalis15:08:48

@wildermuthn We’re actually about to fix this - that is the workaround currently.

michaeldrogalis15:08:04

I just didn’t make the right call when we originally implemented this.

michaeldrogalis15:08:13

Expecting to have a patch out for this in the next week or 2.

wildermuthn15:08:22

I feel less stupid now. 🙂

michaeldrogalis15:08:53

Haha good, not your fault.

michaeldrogalis15:08:22

@wildermuthn Proposed solution is by default disallow segments flowing to downstream tasks when trigger/emit is on and extend flow conditions to to handle trigger/emit segments first class.

devth16:08:07

i assume org.onyxplatform/onyx-peer-http-query "0.10.0.0" won't work with onyx 0.9.x, correct? curious because i'm wanting to get the prometheus /metrics endpoint but haven't upgraded to 0.10 yet

michaeldrogalis16:08:35

@devth Much to Rich’s dismay, 0.9 and 0.10 are incompatible, yes

devth16:08:01

haha. ok thanks. hopefully we'll make the upgrade to 0.10 soon. can't wait to see prom metrics 😄

devth16:08:50

yep i am running that currently. but as far as i could tell, /metrics was added in 0.10

michaeldrogalis16:08:14

Ah, I recalled wrong. Thought we had that for 0.9 too. Yeah, gotta be on 0.10 then

wildermuthn16:08:13

@michaeldrogalis That solution sounds great. I have an additional flow condition downstream from the emitter that would fit more correctly for the emitter itself.

michaeldrogalis17:08:58

@wildermuthn Cool, we’ll run the specifics we’re thinking by you to make sure it’ll work for your use case before we get going.