Fork me on GitHub
#onyx
<
2015-07-07
>
noisesmith18:07:05

@michaeldrogalis: ok, I didn't realize this was here

michaeldrogalis18:07:06

@noisesmith: What did you mean by "send requests via ZooKeeper?"

noisesmith18:07:15

I mean, I have an http server that will want work done on its behalf - should it use onyx and have a mutually required lib in order to get the catalog definition, or is there a simpler route to use the messaging system instead?

michaeldrogalis18:07:28

Kind of application level concern, can't give you bullet proof advice there.

noisesmith18:07:53

OK. So at least there is a sensible way to make either one work.

noisesmith18:07:56

I was looking for an answer like "it would be crazy to send a request to an onyx peer group without using the onyx lib" or "it would totally not make sense to have two different jars from different codebases in the same peer group". But if it's flexible enough, I'll just focus on my own design criteria.

michaeldrogalis18:07:32

So when you use onyx.api/submit-job, it takes a Clojure data structure and serializes it to ZooKeeper, if thats not clear

michaeldrogalis18:07:39

Peers fetch the catalog etc from ZK.

michaeldrogalis18:07:54

So however you want to manage the catalog before submit-job is your perogative

noisesmith18:07:21

OK, so maybe all I need is onyx.api, and the catalog, in my client app, and then my onyx peers would define the rest. That seems sensible.

michaeldrogalis18:07:55

That sounds straightforward.

michaeldrogalis18:07:04

S'all just Clojure data 😉

noisesmith18:07:09

right, thanks