You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Please vote on this issue by adding a 👍 reaction to the original issue to help the community and maintainers prioritize this request
Please do not leave "+1" or "me too" comments, they generate extra noise for issue followers and do not help prioritize the request
If you are interested in working on this issue or have submitted a pull request, please leave a comment
Tell us about your request
The ability for a kube service object of type LoadBalancer to configure what nodes to add to target groups based upon a node selector. This would allow large shared clusters with multiple node groups/autoscaling groups to avoid having ALL nodes get added to a given target group.
Which service(s) is this request for?
EKS
Tell us about the problem you're trying to solve. What are you trying to do, and why is it hard?
As long as we are using node ports for the connectivity between a load balancer created by a service of type LoadBalancer (see #981 for an alternative approach to today's host port solution) and the applications, all nodes within the kube cluster are getting added to the target group. This presents issues if there are multiple node groups/autoscaling groups within a given cluster (e.g. for segmentation or different workload types) as a given application likely does not run on all nodes.
Being able to define a node selector on the service for the target group to use could significantly cut down unnecessary target group membership and hence simplify reasoning about what network connectivity is available between a cluster and load balancers.
Are you currently working around this issue?
No
Additional context
N/A
Attachments
N/A
The text was updated successfully, but these errors were encountered:
The recently released AWS Load Balancer Controller includes support NLB IP targeting #981. If you are using the VPC CNI plugin and can directly target pods, this can optimize the size of target groups in large clusters.
Additionally, a recent upstream feature merged adds support for filtering load balancer target nodes, and this will be available once EKS supports version 1.19.
Community Note
Tell us about your request
The ability for a kube service object of type
LoadBalancer
to configure what nodes to add to target groups based upon a node selector. This would allow large shared clusters with multiple node groups/autoscaling groups to avoid having ALL nodes get added to a given target group.Which service(s) is this request for?
EKS
Tell us about the problem you're trying to solve. What are you trying to do, and why is it hard?
As long as we are using node ports for the connectivity between a load balancer created by a service of type
LoadBalancer
(see #981 for an alternative approach to today's host port solution) and the applications, all nodes within the kube cluster are getting added to the target group. This presents issues if there are multiple node groups/autoscaling groups within a given cluster (e.g. for segmentation or different workload types) as a given application likely does not run on all nodes.Being able to define a node selector on the service for the target group to use could significantly cut down unnecessary target group membership and hence simplify reasoning about what network connectivity is available between a cluster and load balancers.
Are you currently working around this issue?
No
Additional context
N/A
Attachments
N/A
The text was updated successfully, but these errors were encountered: