[RFC] Define Hard Eviction Limits on kubelet-extra-config

Bug #1921524 reported by Peter De Sousa
10
This bug affects 2 people
Affects Status Importance Assigned to Milestone
Kubernetes Worker Charm
Triaged
Wishlist
Unassigned

Bug Description

Sister bug of: https://bugs.launchpad.net/fce-templates/+bug/1921367

Hi,

When running Kubernetes in production there are cases where pods can aggressively consume memory and or CPU causing the node to become unreachable without a restart.

This can be solved on a per application basis with limits, but this relies on the user to define these limits rather than Kubernetes evicting when it's running out of memory.

This is a RFC to consider changing the charm setting the kubelet-extra-config arg on Charmed Kubernetes, preventing any application from consuming the node until failure.

Cheers,

Peter

Changed in charm-kubernetes-worker:
milestone: none → 1.21+ck1
George Kraft (cynerva)
Changed in charm-kubernetes-worker:
importance: Undecided → Wishlist
status: New → Triaged
George Kraft (cynerva)
Changed in charm-kubernetes-worker:
milestone: 1.21+ck1 → none
To post a comment you must log in.
This report contains Public information  
Everyone can see this information.

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.