Choco is really neat. I played with it awhile back to pick fantasy basketball teams for a friend (using constrained knapsack): https://gist.github.com/hadronzoo/7916c32c0d434c9e7dc5#file-core-clj

Probabilistic models are much better for the fantasy basketball problem, but it was still fun.

@joe.lane: > any tips on getting started in distributed systems Maybe start with the CAP theorem and its critique by Martin Kleppmann. I was super lucky to have my first distributed systems talk be by Kleppmann talking about the CAP theory at Strangeloop 2015.

Watching it right now 🙂 Thanks @drewverlee !

We're putting together a self-help guide that has a series of questions that you can answer to diagnose common problems with your Onyx program. Does anyone have thoughts on whether this would be best as a static page with links to jump around between, or something more visual like a flow chart?

Static page. Scannable and searchable.

Been leaning in that direction. I imagine anyone using it will eventually want to read everything in its entirety and not need to click around to navigate.

Oh, I should mention that the reason a flow chart might be interesting is that some of the questions to determine what the problem is are several deep. e.g. "Is X true?", yes -> "Is Y true?"

Sounds possibly interesting but hard to reason about unless I considered some of the actual flows. I wonder if you could encode some of that in a doc.

I think in a doc it would look like:

What's the state of X?
- [It is A](link-1)
- [It is B](link-2)
...

Where link 1/2 are anchors into the same doc

👍

Looking at the onyx dashboard (very cool!) - one of my jobs shows blank under "Host Allocation". Does that mean there are no peers available to do the work?

@devn: I think so. Can I see a screenshot?

I think I had submitted that job with NPEERS=4

It says its running, so perhaps its a bug in the dash. o-o

I went back and re-submitted with NPEERS=3 and the job ran, but the old one continues to spin

You should kill off your old one.

(don't fully understand what the significance of NPEERS is, and why more than enough would casue it to not run)

It wouldn't ^

Ok. Don't see a way to kill jobs in the dashboard.

I need to ssh in and kill from a local repl or something?

Check out the scheduler chapter of the user guide to learn about that, btw.

Thanks will do

onyx.api/kill-job on the job ID

I thought we left kill-job in the dashboard, and just ditched restart-job. It’d be good to add kill-job back in at least. Anyway, yes onyx.api/kill-job

Can I do that from a local repl if I point it at the prod zookeeper? Or what's the normal workflow for submitting/restarting/killing jobs?

Yes to the first question. Latter part depends on your set up, we've done it a few different ways. Depends on the customer

This might be helpful, too. See the Clojure part and HTTP server below: https://github.com/onyx-platform/lib-onyx#replica-queries

Just another way to see what's going on.

Nice

It would be nice to have a cohesive command line tool to do submission/restarts/job killing. I think that needs to come out of the community organically though.

Got it. I'm assuming a CLI would hit the endpoints provided by

(require '[lib-onyx.lib-onyx.replica-query-server :as rqs])

(def server-port 3000)

(def server (rqs/start-replica-query-server peer-config server-port)

Yep yep 🙂

@devth: I’m working on a kubernetes example for onyx-twitter-sample and just remembered how to handle the shm issue,

- name: aeronshm
  emptyDir:
    medium: “Memory”

gardnervickers: nice, thanks!