[12:56:18] <cdanis>	 do we have any PriorityClasses already in use in any of our clusters?
[13:00:12] * inflatador was looking for that in deployment-charts repo the other day
[13:01:11] <inflatador>	 I didn't find anything...so I'm guessing no
[13:02:09] <cdanis>	 could be one potential answer re: GPU crowding for klausman 
[13:04:28] <klausman>	 ahyes, good idea
[13:05:23] <inflatador>	 the docs on preemption are pretty good. Probably still wanna test, but it's clear they tried to explain exactly what happens: https://kubernetes.io/docs/concepts/scheduling-eviction/pod-priority-preemption/
[13:06:11] <klausman>	 Why do the number 119, 200, 360, 450 and 500 suddenly pop into my mind...
[13:07:40] <cdanis>	 klausman: I ran a production service that heavily (ab)used the batch scheduler :)
[13:10:37] <klausman>	 I also rember doing the numbers regarding Exacycle and BTC
[13:11:16] <klausman>	 at the time, it would have been a net positive, but it likely would also have completely wrecked the market
[13:11:27] <cdanis>	 hodl