419 reduce ops costs

airflow/helm/values.tmpl.yaml

@@ -56,10 +56,10 @@ affinity:
           values: ["on-demand"]
         - key: "karpenter.k8s.aws/instance-family"
           operator: "In"
-          values: ["c6i", "c5"]
+          values: ["r5"]
         - key: "karpenter.k8s.aws/instance-cpu"
           operator: "In"
-          values: ["2", "4"]
+          values: ["8"]
 
 topologySpreadConstraints:
 - maxSkew: 1
@@ -248,7 +248,9 @@ workers:
   keda:
     enabled: true
     pollingInterval: 1
-    minReplicaCount: 1
+    # Minimum node available for celery workers
+    # Set to 1 to always have a worker node available to run tasks
+    minReplicaCount: 0
     maxReplicaCount: 128
     # Specify HPA related options
     # https://github.com/kubernetes/enhancements/blob/master/keps/sig-autoscaling/853-configurable-hpa-scale-velocity/README.md

terraform-unity/main.tf

@@ -50,6 +50,7 @@ module "unity-sps-database" {
   venue        = var.venue
   service_area = var.service_area
   release      = var.release
+  db_instance_class = var.db_instance_class
 }
 
 module "unity-sps-efs" {

terraform-unity/modules/terraform-unity-sps-database/main.tf

@@ -68,7 +68,7 @@ resource "aws_db_instance" "sps_db" {
   storage_type         = "gp3"
   engine               = "postgres"
   engine_version       = "16.8"
-  instance_class       = "db.m5d.2xlarge"
+  instance_class       = var.db_instance_class
   db_name              = "sps_db"
   username             = "db_user"
   password             = aws_secretsmanager_secret_version.db.secret_string

terraform-unity/modules/terraform-unity-sps-database/variables.tf

@@ -17,3 +17,8 @@ variable "release" {
   description = "The software release version."
   type        = string
 }
+
+variable "db_instance_class" {
+  description = "The SPS RDS database instance class"
+  type        = string
+}

terraform-unity/modules/terraform-unity-sps-ogc-processes-api/main.tf

@@ -17,6 +17,34 @@ resource "kubernetes_deployment" "redis" {
         }
       }
       spec {
+        affinity {
+          node_affinity {
+            required_during_scheduling_ignored_during_execution {
+              node_selector_term {
+                match_expressions {
+                  key      = "karpenter.sh/nodepool"
+                  operator = "In"
+                  values   = compact([for pool in var.karpenter_node_pools : pool if pool == "airflow-core-components"])
+                }
+                match_expressions {
+                  key      = "karpenter.sh/capacity-type"
+                  operator = "In"
+                  values   = ["on-demand"]
+                }
+                match_expressions {
+                  key      = "karpenter.k8s.aws/instance-family"
+                  operator = "In"
+                  values   = ["r5"]
+                }
+                match_expressions {
+                  key      = "karpenter.k8s.aws/instance-cpu"
+                  operator = "In"
+                  values   = ["8"]
+                }
+              }
+            }
+          }
+        }
         container {
           name  = "redis"
           image = "${var.docker_images.redis.name}:${var.docker_images.redis.tag}"
@@ -83,12 +111,12 @@ resource "kubernetes_deployment" "ogc_processes_api" {
                 match_expressions {
                   key      = "karpenter.k8s.aws/instance-family"
                   operator = "In"
-                  values   = ["c6i", "c5"]
+                  values   = ["r5"]
                 }
                 match_expressions {
                   key      = "karpenter.k8s.aws/instance-cpu"
                   operator = "In"
-                  values   = ["4"]
+                  values   = ["8"]
                 }
               }
             }

terraform-unity/variables.tf

@@ -283,7 +283,7 @@ variable "karpenter_node_pools" {
         memory = "320Gi"
       }
       disruption = {
-        consolidationPolicy = "WhenEmpty"
+        consolidationPolicy = "WhenEmptyOrUnderutilized"
         consolidateAfter    = "1m"
       }
     },
@@ -367,3 +367,9 @@ variable "installprefix" {
   type        = string
   default     = ""
 }
+
+variable "db_instance_class" {
+  description = "The SPS RDS database instance class"
+  type        = string
+  default     = "db.m5d.xlarge"
+}