Open JacobHenner opened 4 hours ago
This issue is currently awaiting triage.
If Karpenter contributors determines this is a relevant issue, they will accept it by applying the triage/accepted
label and provide further guidance.
The triage/accepted
label can be added by org members by writing /triage accepted
in a comment.
Description
What problem are you trying to solve?
Today, limits can only be specified on individual NodePools. While this is fine for simple situations, it is insufficient when multiple NodePools comprise a logical grouping of compute that should share a limit. This happens most often when there are important variations in a NodePool's properties beyond its requirements that mandate the use of multiple NodePools, but when they are otherwise related in some way relevant to limits (e.g. same department, team, application, budget line-item).
For example, an organization might group limits by team. A team might require nodes labelled in two distinct ways, necessitating the use of two NodePools. Splitting the team's limit in half for each NodePool might not be sufficient if the balance of nodes between the NodePools varies over time.
I propose a
NodePoolGroupLimit
CRD (or a similar appropriate name) that would allow a defined limit to apply to NodePools chosen by a selector. If multipleNodePoolGroupLimit
objects select the same NodePool, the most prohibitive limit should take precedence.It might look something like this:
How important is this feature to you?