Item — HN Reader

anyone got recommendation on how to properly set CPU limits?

I've set limits about 2x higher than the max cpu usage, but still seeing some throttling, what cause performance degradation. I've ended up setting deployments without limits to avoid throttling but that's definitely not nice if one pod goes crazy for a reason ...

1 comments

Not sure what you mean by throttling. I tell my customers to do 4 things

   1) Set a Request at about 50% of max CPU usage
   2) Set a Limit at about 90% of max CPU usage
   3) Configure a HorizontalPodAutoscaler that kicks in at 70% 
   4) Scale out, in advance, to 2 or 3 pod replicas