Kubernetes Antipatterns: CPU Limits • Natan Yellin

No one lacks an opinion when it comes to CPU limits. Are they a mandatory requirement or a misunderstood evil? Most cloud providers tell you to set them, yet members of the original Kubernetes team at Google say not to use them! Why all the debate about something that seems so simple? Should you use CPU limits or not?
Natan will go through both sides of the argument. He’ll explain why there is so much controversy over something so simple, and help you determine what is right for your team.
__________
- Interested in more great talks? Check the upcoming ones here: platformengine...
- Join our Platform Engineering Slack community: platformengine...

[@SyedAshrafulla]

Great talk. Can't wait for the sibling talk on memory resource configuration!
It would be cool if memory was "compressible" meaning unused memory for a pod was reclaimable and pods that didn't have enough memory waited for that memory to free up from other pods terminating or otherwise handing cleared memory back. Kind of like a buy-nothing-sell-nothjng for memory. Probably bug-prone but a fun idea.

[@Viet100]

If I configured limites, so do I need to use VPA?

[@royolsen1576]

I would have thought it obvious that (cpu) limits are for controlling long running/batch workloads?

[@oscarpoels8727]

One of the best videos on the subject tbh, perfect level of abstraction, not too deep to get lost and not too on the surface to miss key concepts

[@rishiraj2548]

Thanks a lot

[@kilipsbyhades]

if we are not setting on limits then how can we configure HPA(Horizontal Pod Autoscalar)

[@mrbb88]

HPA operates on the requested resources and not the limits. From the docs: " the controller calculates the utilization value as a percentage of the equivalent resource request on the containers in each Pod"

[@youtux2]

One way I realised CPU limits were actually limiting our performance was to track the throttling metrics

[@rajuhossain7819]

Nice

[@RobustaDev]

thank you!