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
When using --balance-similar-node-groups=true to balance scaling across multiple node groups, each in a different AZ, scaling across node group ASGs is not uniform unless all node labels match.
I created 3 node groups using eksctl (each in a different AZ) and found that a node label eks.amazonaws.com/sourceLaunchTemplateId (and possibly eks.amazonaws.com/sourceLaunchTemplateVersion) is different across nodes of different node groups. Scaling was not uniform unless I passed --balancing-ignore-label=eks.amazonaws.com/sourceLaunchTemplateId. This labels is not covered in the list of aws ignored labels or BasicIgnoredLabels.
Think it will be useful to include a note in documentation so that user can compare the node labels and pass --balancing-ignore-label for each label that needs ignoring.
The text was updated successfully, but these errors were encountered:
You'll also need kubernetes/autoscaler#3673 to be fixed and the chart to be released (CA chart releases are a manual process after the PR has been merged and the chart has been built).
I've handled incorporating this information as part of incorporating content for PR. I'm waiting on an engineer to do full confirmation on my summary. But feel free to re-open if there's a need beyond what will be covered with PR: "When setting --balance-similar-node-groups to true, make sure that the node groups you want the Cluster Autoscaler to balance have matching labels (except for automatically added zone labels). You can pass a --balancing-ignore-label flag to nodes with different labels to balance them regardless, but this should only be done as needed."
This relates to issue kubernetes/autoscaler#3515
When using
--balance-similar-node-groups=true
to balance scaling across multiple node groups, each in a different AZ, scaling across node group ASGs is not uniform unless all node labels match.I created 3 node groups using eksctl (each in a different AZ) and found that a node label
eks.amazonaws.com/sourceLaunchTemplateId
(and possiblyeks.amazonaws.com/sourceLaunchTemplateVersion
) is different across nodes of different node groups. Scaling was not uniform unless I passed--balancing-ignore-label=eks.amazonaws.com/sourceLaunchTemplateId
. This labels is not covered in the list of aws ignored labels or BasicIgnoredLabels.Think it will be useful to include a note in documentation so that user can compare the node labels and pass
--balancing-ignore-label
for each label that needs ignoring.The text was updated successfully, but these errors were encountered: