Get capi targetsize from cache #5025
Conversation
k8s-ci-robot
added
area/provider/cluster-api
k8s-ci-robot
added
the
size/S
k8s-ci-robot
added
the
approved
There was a problem hiding this comment.
this makes sense to me, thanks for the update @enxebre
i just have a quick question about
Also note that the scale down implementation has changed https://github.com/kubernetes/autoscaler/commits/master/cluster-autoscaler/core/scaledown.
did you have something specific in mind?
| return 0, errors.Wrap(err, "error getting replica count") | ||
| } | ||
| if !found { | ||
| replicas = 0 |
There was a problem hiding this comment.
should we set replicas = ng.scalableResource.MinSize() here?
There was a problem hiding this comment.
that seems reasonable to me, i guess we still need to default to 0 if there is an issue with minsize though.
There was a problem hiding this comment.
actually I made it an error, as I don't think it is a valid case for it to not be found.
|
[APPROVALNOTIFIER] This PR is APPROVED This pull-request has been approved by: arunmk, enxebre The full list of commands accepted by this bot can be found here. The pull request process is described here
Needs approval from an approver in each of these files:
Approvers can indicate their approval by writing |
This ensured that access to replicas during scale down operations were never stale by accessing the API server kubernetes#3104. This honoured that behaviour while moving to unstructured client kubernetes#3312. This regressed that behaviour while trying to reduce the API server load kubernetes#4443. This put back the never stale replicas behaviour at the cost of loading back the API server kubernetes#4634. Currently on e.g a 48 minutes cluster it does 1.4k get request to the scale subresource. This PR tries to satisfy both non stale replicas during scale down and prevent the API server from being overloaded. To achieve that it lets targetSize which is called on every autoscaling cluster state loop from come from cache. Also note that the scale down implementation has changed https://github.com/kubernetes/autoscaler/commits/master/cluster-autoscaler/core/scaledown.
enxebre
force-pushed
the
get-targetSize-fromcache
branch
from
a80af8e to
b2f1823
Compare
|
LGTM, i'll let @arunmk add the label |
|
/lgtm /approve |
|
/lgtm |
k8s-ci-robot
added
the
lgtm
Get capi targetsize from cache
Which component this PR applies to?
Cluster autoscaler - cluster api.
What type of PR is this?
This ensured that access to replicas during scale down operations were never stale by accessing the API server #3104.
This honoured that behaviour while moving to unstructured client #3312.
This regressed that behaviour while trying to reduce the API server load #4443.
This put back the never stale replicas behaviour at the cost of loading back the API server #4634.
Currently on e.g a 48 minutes cluster with no scaling activity it does 1.4k get request to the scale subresource.
This PR tries to satisfy both non stale replicas during scale down and prevent the API server from being overloaded. To achieve that it lets TargetSize which is called on every autoscaling cluster state loop from come from cache while getting fresh replicas at the time of perform the operations.
Also note that the scale down implementation has changed https://github.com/kubernetes/autoscaler/commits/master/cluster-autoscaler/core/scaledown.
/area provider/cluster-api
/kind bug
/kind cleanup