Fork me on GitHub
#onyx
<
2016-03-03
>
acron10:03:54

@michaeldrogalis: beginners-guide is awesome :thumbsup:

michaeldrogalis15:03:12

@acron: Thanks! I guess I should actually finish it now that folks are unearthing it from nearly a year ago. simple_smile

lucasbradstreet16:03:13

We now have a newrelic onyx-metrics plugin

lucasbradstreet16:03:36

It was painful to write. The java api kinda sucks and then I found out that it just does REST calls anyway, so I switched to the REST API

lucasbradstreet16:03:04

@michaeldrogalis: it depends on the job-metadata functionality, so it’ll only get shipped with 0.9.0

michaeldrogalis16:03:46

@lucasbradstreet: Okay, sounds good. Glad we got that patch ready.

lucasbradstreet16:03:34

Jepsen post is getting there

gardnervickers16:03:22

blown away by those posts you recommended my last night @lucasbradstreet, had no idea how relatively weak ec2 instances were

lucasbradstreet16:03:52

Yeah, I couldn’t figure out why my 6 core Xeon v2 was besting a 8 core v2 on EC2

lucasbradstreet16:03:29

I don’t think it makes EC2 necessarily a bad deal, it’s just annoying that it’s easy to be fooled by their specs

gardnervickers16:03:25

For a while I was under the impression that they somehow just had absurdly low prices compared to buying hardware. Makes a lot more sense now.

lucasbradstreet16:03:10

Yeah, you're actually paying a lot for the flexibility they give

lucasbradstreet16:03:36

What's surprising is that there aren't any 36 core machines via dual socket 18 core CPUs

gardnervickers16:03:02

I had no idea there were even 18 core chips

gardnervickers16:03:22

last time I bought a cpu was during the core 2 quad days

lucasbradstreet16:03:11

I'm sure Intel will have 30+ core CPUs in a few years

lucasbradstreet16:03:42

How'd the bench look after those changes?

gardnervickers16:03:56

latency didnt drop much but throughput did

gardnervickers16:03:04

restarting kubernetes with 6 nodes

gardnervickers16:03:04

You can dynamically add more nodes, I’m just doing chores around the house and am too lazy to go through the setup process, rather just edit the numbers in the cloudformation scripts 😉

gardnervickers17:03:16

Is the only way to get an instance limit increase with AWS to use their support system?

pguillebert17:03:46

as far as I know, yes

pguillebert17:03:11

but usually they are OK with it and respond fairly quickly

pguillebert17:03:25

like, under 2 days

pguillebert17:03:44

(in the “I won’t pay a thing” support plan)

pguillebert17:03:12

can also be done in just a few hours. Just, there is no commitment if you don’t have support.

lucasbradstreet17:03:10

@gardnervickers: we should bump those every time we get a chance

lucasbradstreet17:03:23

I'd like to have a really high limit there

lvh20:03:42

I noticed that the sample-catalog from the template requires the fn it uses. Is that necessary? On the one hand, the catalog just refers to it as a keyword, but I could see how it needs to be required somewhere for that keyword to be resolvable to the actual fn. Or does that always work by default?

lucasbradstreet20:03:12

That’s not necessary

lucasbradstreet20:03:36

It only needs to be required by code that is running on the peers

lvh20:03:59

OK, so something does need to at least require that namespace somewhere?

gardnervickers20:03:35

I dont believe it’s included in the jar if you dont require it somewhere.

lvh20:03:05

right, either that or require-time side-effects like defmethod

lucasbradstreet20:03:36

Typically we do this in the code that launches the peers https://github.com/onyx-platform/onyx-template/blob/0.8.x/src/leiningen/new/onyx_app/src/onyx_app/launcher/launch_prod_peers.clj but you could do it in a separate namespace (so you could require it in your tests and such too)