diff --git a/cluster/config-defaults.yaml b/cluster/config-defaults.yaml index 9f76313153..a4c80730fc 100644 --- a/cluster/config-defaults.yaml +++ b/cluster/config-defaults.yaml @@ -51,6 +51,11 @@ karpenter_in_transit_support_required: "false" # t type instances have burstable CPU, which can be undesirable in production karpenter_instance_family_t_enabled: "false" +# configure whether we allow g6f instance family for Karpenter nodes +# g6f has fractional GPU count which is also mislabeled by Karpenter and causes +# scheduling issues, we need to properly test and support it later if needed +karpenter_instance_family_g6f_enabled: "false" + # configure whether spot instances should be enabled in Karpenter's capacity-types karpenter_enable_spot: "true" diff --git a/cluster/node-pools/worker-karpenter/provisioners.yaml b/cluster/node-pools/worker-karpenter/provisioners.yaml index 44488851bc..0cdd58099b 100644 --- a/cluster/node-pools/worker-karpenter/provisioners.yaml +++ b/cluster/node-pools/worker-karpenter/provisioners.yaml @@ -164,6 +164,10 @@ spec: - "t3" - "t2" #{{ end }} +#{{ if eq .NodePool.ConfigItems.karpenter_instance_family_g6f_enabled "false"}} + # g6f has fractional GPU count which is also mislabeled by Karpenter and causes scheduling issues + - "g6f" +#{{ end }} #{{ if eq .NodePool.ConfigItems.karpenter_in_transit_support_required "true" }} - key: karpenter.k8s.aws/instance-encryption-in-transit-supported operator: In