test: Google Kubernetes Engine - Prover Agent Spot Node Support

spartan/aztec-network/templates/prover-agent.yaml

@@ -17,12 +17,20 @@ spec:
         {{- include "aztec-network.selectorLabels" . | nindent 8 }}
         app: prover-agent
     spec:
+      {{- if .Values.proverAgent.gke.spotEnabled }}
+      affinity:
+        nodeAffinity:
+          requiredDuringSchedulingIgnoredDuringExecution:
+            nodeSelectorTerms:
+            - matchExpressions:
+              - key: cloud.google.com/gke-spot
+                operator: Exists
+      tolerations:
+      - key: "cloud.google.com/gke-spot"
+        operator: "Equal"
+        value: "true"
+        effect: "NoSchedule"
       serviceAccountName: {{ include "aztec-network.fullname" . }}-node
-      {{- if .Values.proverAgent.nodeSelector }}
-      nodeSelector:
-        {{- toYaml .Values.proverAgent.nodeSelector | nindent 8 }}
-      {{- end }}
-
       {{- if .Values.network.public }}
       hostNetwork: true
       {{- end }}

spartan/aztec-network/values.yaml

@@ -205,6 +205,8 @@ proverAgent:
     nodePort: 8083
   enabled: true
   replicas: 1
+  gke:
+    spotEnabled: false
   logLevel: "debug"
   debug: "aztec:*,-aztec:avm_simulator*,-aztec:libp2p_service*,-aztec:circuits:artifact_hash,-json-rpc*,-aztec:world-state:database,-aztec:l2_block_stream*"
   realProofs: false

spartan/aztec-network/values/gcp-proving-test.yaml

@@ -45,9 +45,6 @@ proverAgent:
   realProofs: true
   bb:
     hardwareConcurrency: 31
-  nodeSelector:
-    cloud.google.com/compute-class: "Performance"
-    cloud.google.com/machine-family: "t2d"
   resources:
     requests:
       memory: "116Gi"

spartan/terraform/multicloud-deploy/main.tf

@@ -58,12 +58,28 @@ resource "helm_release" "aztec-eks-cluster" {
   # base values file
   values = [file("../../aztec-network/values/${var.values-file}")]
 
-  # removing prover nodes
+  # network customization
+  set {
+    name  = "images.aztec.image"
+    value = var.image
+  }
+
+  set {
+    name  = "telemetry.enabled"
+    value = var.telemetry
+  }
+
   set {
     name  = "network.public"
     value = true
   }
 
+  set {
+    name  = "validator.replicas"
+    value = 16
+  }
+
+  # removing prover nodes
   set {
     name  = "proverNode.replicas"
     value = "0"
@@ -75,7 +91,7 @@ resource "helm_release" "aztec-eks-cluster" {
   }
 
   # Setting timeout and wait conditions
-  timeout       = 600 # 10 minutes in seconds
+  timeout       = 1200 # 20 minutes in seconds
   wait          = true
   wait_for_jobs = true
 }
@@ -92,17 +108,43 @@ resource "helm_release" "aztec-gke-cluster" {
   # base values file
   values = [file("../../aztec-network/values/${var.values-file}")]
 
-  # disabling all nodes except provers
+  # network customization
   set {
-    name  = "network.setupL2Contracts"
-    value = false
+    name  = "images.aztec.image"
+    value = var.image
+  }
+
+  set {
+    name  = "telemetry.enabled"
+    value = var.telemetry
   }
 
   set {
     name  = "proverNode.public"
     value = true
   }
 
+  set {
+    name  = "proverAgent.replicas"
+    value = 32
+  }
+
+  set {
+    name  = "proverAgent.gke.spotEnabled"
+    value = true
+  }
+
+  set {
+    name  = "network.setupL2Contracts"
+    value = false
+  }
+
+  set {
+    name  = "proverAgent.bb.hardwareConcurrency"
+    value = 16
+  }
+
+  # disabling all nodes except provers
   set {
     name  = "bootNode.replicas"
     value = "0"
@@ -160,7 +202,7 @@ resource "helm_release" "aztec-gke-cluster" {
   }
 
   # Setting timeout and wait conditions
-  timeout       = 600 # 10 minutes in seconds
+  timeout       = 1200 # 20 minutes in seconds
   wait          = true
   wait_for_jobs = true
 

spartan/terraform/multicloud-deploy/variables.tf

@@ -19,5 +19,17 @@ variable "testnet_name" {
 variable "values-file" {
   description = "Name of the values file to use for deployment"
   type        = string
-  default     = "multicloud-demo.yaml"
+  default     = "16-validators.yaml"
+}
+
+variable "image" {
+  description = "Aztec node image"
+  type        = string
+  default     = "aztecprotocol/aztec:master"
+}
+
+variable "telemetry" {
+  description = "Toogle telemetry on/off"
+  type        = bool
+  default     = false
 }