Add Runner module

[simontabor]

Run attacks at defined random probabilities!

Git diff on this is awful, sorry - it's detected cross-file changes which shouldn't be there... should have used git mv instead of mv 😞

Closes #2

[loujaybee]

Any reason you wanted to expose a function here? No strong reasons against it other than it making me wonder what the purpose is (I'll read through the rest of the PR and it might become apparent!)

[simontabor]

I wanted to allow users to "configure once, run multiple". The runners and attacks are not re-usable - a new instance must be created for each run or attack...

Maybe I've overthought it, and passing the config each time is absolutely fine, but I was trying to support this kind of usage pattern...

const createRunner = Runner.configure({
  probability: 1,
  possibleAttacks: [
    {
      probability: 0.01,
      createAttack: CPUAttack.configure({ allowLoopEvery: 10 })
    },
    {
      probability: 0.01,
      createAttack: OpenFilesAttack.configure({ number: 1000 })
    }
  ]
});

// then, when a request comes in
const runner = createRunner();
runner.start();

// and when the request finishes...
runner.stop();

I think this will also help with the environment configuration (#9 etc), but happy to hear reasons against (complexity etc)

[loujaybee]

Ah, gotcha, so configure is just a factory function, so they don't need to do that manually.

[simontabor]

I considered calling the configure methods factory or something - do you think that'd be a better name?

[loujaybee]

Any reason to not keep the original (non-runner) example?

[simontabor]

Those examples are probably best kept within the individual attacks, WDYT?

[simontabor]

I want to do this ASAP but split into another issue to avoid this from becoming monolithic - #14

[loujaybee]

The "probability" that's mentioned here, what determines that? Is it each invocation of the runner is a roll of the dice?

(As on the other comments, I'll keep digging and I'm sure I'll find the answer 😛 )

Just wondering if we could make it more explicit.

[loujaybee]

Any reason this example uses a constructor, and the other uses the configure method? Is that intentional to show two different API's? Or an error?

[simontabor]

Just to show the differences, could use some better documentation

[loujaybee]

Yeah, might be good just to call that out. I did wonder which was the preferred.

[loujaybee]

Would be good to mention that this takes precedence over the individually configured probabilities.

[loujaybee]

Looks good, Simon. 🥳

A couple thoughts from looking through...

• I'm not sure of the benefit of the createAttack configuration being a function if it takes no arguments (unless you had future plans for it).
• I'm not totally clear on how the running works, should I invoke it once per request, or once per service? It'd be good to clarify this in the README if possible.
• It'd be nice to have an examples folder with the runner working (possibly with some console logs) to show whether it's working or not (that said, I appreciate you mentioned the issue of adding logs)!