[tf/helm] defaults for scaling config

Closes: aptos-labs#800
vldslvpolakova · May 5, 2022 · 23996a1 · 23996a1
1 parent 768f544
commit 23996a1
Show file tree

Hide file tree

Showing 3 changed files with 11 additions and 1 deletion.
diff --git a/terraform/helm/k8s-metrics/templates/autoscaler.yaml b/terraform/helm/k8s-metrics/templates/autoscaler.yaml
@@ -164,6 +164,8 @@ spec:
             - --node-group-auto-discovery=asg:tag=k8s.io/cluster-autoscaler/enabled,k8s.io/cluster-autoscaler/{{ .Values.autoscaler.clusterName }}
             - --balance-similar-node-groups
             - --skip-nodes-with-system-pods=false
+            - --scale-down-unneeded-time={{ .Values.autoscaler.scaleDownUnneededTime }}
+            - --scale-down-delay-after-add={{ .Values.autoscaler.scaleDownDelayAfterAdd }}
           volumeMounts:
             - name: ssl-certs
               mountPath: /etc/ssl/certs/ca-certificates.crt #/etc/ssl/certs/ca-bundle.crt for Amazon Linux Worker Nodes

diff --git a/terraform/helm/k8s-metrics/values.yaml b/terraform/helm/k8s-metrics/values.yaml
@@ -5,6 +5,8 @@ coredns:
 autoscaler:
   enabled:
   clusterName:
+  scaleDownUnneededTime: 15m
+  scaleDownDelayAfterAdd: 5m
   image:
     repo: k8s.gcr.io/autoscaling/cluster-autoscaler
     tag: v1.21.0

diff --git a/terraform/validator/aws/cluster.tf b/terraform/validator/aws/cluster.tf
@@ -48,7 +48,7 @@ resource "aws_launch_template" "nodes" {
   for_each      = local.pools
   name          = "aptos-${local.workspace_name}/${each.key}"
   instance_type = each.value.instance_type
-  user_data     = base64encode(
+  user_data = base64encode(
     templatefile("${path.module}/templates/eks_user_data.sh", {
       taints = each.value.taint ? "aptos.org/nodepool=${each.key}:NoExecute" : ""
     })
@@ -71,6 +71,12 @@ resource "aws_eks_node_group" "nodes" {
   subnet_ids      = [aws_subnet.private[0].id]
   tags            = local.default_tags
 
+  lifecycle {
+    ignore_changes = [
+      scaling_config[0].desired_size
+    ]
+  }
+
   launch_template {
     id      = aws_launch_template.nodes[each.key].id
     version = aws_launch_template.nodes[each.key].latest_version