CAP*: expose kube-scheduler KUBE_MAX_PD_VOLS
environment variable
#3431
Labels
capi/migration
component/k8s-initiator
provider/cluster-api-aws
Cluster API based running on AWS
team/phoenix
Team Phoenix
team/turtles
Team Turtles
topic/capa-migration
topic/capi
There are cases in which we saw
kube-scheduler
fail to compute the maximum number of EBS volumes attachable to an instance. This is possibly due to volumes we currently attach to VMs "outside of Kubernetes". Basically,kube-scheduler
knows that an EC2 instance hasX
slots for EBS volumes, but indeed onlyX-n
are available. This leads to pods hanging while waiting for a volume to be attached.The environment variable
KUBE_MAX_PD_VOLS
allows for tellingkube-scheduler
the maximum number of volumes attachable to each node in a cluster. As of now, we are setting it in Vintage via thek8s-initiator-app
. It is currently only used on AWS, but would probably be helpful to expose such env variable (among others, I guess) in all providers.As part of the migration to CAPI-native functionalities from the
k8s-initiator
App, we need this capability. I am assigning Phoenix and Turtles for visibility but if we want to make it available on all providers, then I assume it's Turtles.The text was updated successfully, but these errors were encountered: