use thread.start() to start an actual thread

[tjstansell]

thread.run() just runs the code, doesn't spawn off a new thread first.

[wolph]

lol, that error is hilarious actually :P

Frankly I would be a bit scared of merging this code though... if the code has run like this for a long time it means that it was probably never tested for thread safety and race conditions

[Rockstar04]

Thanks for the PR! I would have to agree with @wolph though, I think this should be something to definitely cover in #2.

[tjstansell]

Well, in theory the only things that might be interacting between threads is the logging and the queue where the results are written to. The logging module is designed to be thread-safe and so is the queue.

Do you guys have any suggestions on how to test for thread safety beyond just "trying it"? FWIW, I'll be running with this enabled myself once I finish my changes to support dynamic configs and get this rolled out to monitor our RDS instances.

[wolph]

I suppose just a bit of trying combined with a quick code check. I agree that both queue and logging should be safe (with logging that's actually debatable but we can ignore that).

Beyond that... I don't really see any obvious problems either but I'll check the code a bit more thoroughly just to be sure

[wolph]

I've checked the code as far as I can and I don't see any obvious problems. It might cause issue in some logging configurations but in general it will probably work just fine :)

[Rockstar04]

We can tag the next release as a major version bump if we are worried about the backwards compatibility this introduces.

[tjstansell]

FWIW, I haven't seen any issues with this in all of my testing I've done so far... thanks for merging!