Add secrets support and instruction for budget notifications (#444)

* Move to secrets

* Add billing actions

terraform/README.md

@@ -1,53 +1,94 @@
-#ToDo:
-Move secrets to gcp secrets
-Format readme in md
-Cloud build for terraform
-[?]Wait for kubernetes plugin to finish apply
-Add readme if the budget is exceeded
-
-Be aware, the actions you execute on your gcp project will generate some cost.
-
-#Permissions
+# Permissions
 TODO
 
-#1st run (bootstrap)
+# Step 1: Bootstrap
 
 Copy `variables.tfvars` from `variables.tfvars_example`
 Replace the placeholders for `project-id` and `billing-account` in `variables.tfvars`
 Insert secret values in the `variables.tfvars` file or insert values on runtime when using terraform plan or apply
 
-Initialise terraform
+### Initialise terraform with local backend
 Comment out everything in `backend.tf` file to use local state for the first run as the bucket for storing the state is not created.
-```terraform init```
+```
+terraform -chdir=terraform init
+```
 
-Create the state bucket
-```terraform apply -var-file=variables.tfvars -target="google_storage_bucket.terraform_state"```
+### Create the state bucket
+```
+terraform -chdir=terraform apply -var-file=variables.tfvars -target="google_storage_bucket.terraform_state"
+```
 To disable the conformation use `--auto-aprove` flag
 
-##Move the state to the bucket.
+## Migrate the state to the bucket.
 Uncomment everything in `backend.tf` file to use remote state with newly created bucket.
-```export PROJECT_ID="<PROJECT_ID>"```
-```terraform init -backend-config="bucket=terraform-state-${PROJECT_ID}" -backend-config="prefix=terraform/state"```
+```
+export PROJECT_ID="<PROJECT_ID>"
+```
+```
+terraform -chdir=terraform init -backend-config="bucket=terraform-state-${PROJECT_ID}" -backend-config="prefix=terraform/state"
+```
+## Create the secrets.
+```
+terraform -chdir=terraform apply -var-file=variables.tfvars -target="google_secret_manager_secret.secret"
+```
 
-Create the cluster. Due to the problem described [here](https://github.com/hashicorp/terraform-provider-kubernetes/issues/1775) terraform kubernetes provider requires kubernetes cluster to be created first. So to create the cluster without applying kubernetes resources we will do the apply in 2 runs using the `-target` flag.
-```terraform apply -var-file=variables.tfvars -target="google_container_cluster.llvm_premerge_checks_cluster"```
+## Create the cluster. 
+Due to the problem described [here](https://github.com/hashicorp/terraform-provider-kubernetes/issues/1775) terraform kubernetes provider requires kubernetes cluster to be created first. So to create the cluster without applying kubernetes resources we will do the apply in 2 runs using the `-target` flag.
+```
+terraform -chdir=terraform apply -var-file=variables.tfvars -target="google_container_cluster.llvm_premerge_checks_cluster"
+```
 
-##Creating worker images
+## Build the builders
 To deploy build workers you need the worker docker images in your project.
 TODO cloud build SA permissions
 
-###Linux worker image
+### Linux worker image
 Execute cloud build to build Linux worker:
-```gcloud builds submit --config=containers/buildkite-premerge-debian/cloudbuild.yaml containers/buildkite-premerge-debian/ --project=${PROJECT_ID}```
+```
+gcloud builds submit --config=containers/buildkite-premerge-debian/cloudbuild.yaml containers/buildkite-premerge-debian/ --project=${PROJECT_ID}
+```
 
-###Windows worker image
+### Windows worker image
 Build windows cloud builder. Follow the steps described here: [link](https://github.com/GoogleCloudPlatform/cloud-builders-community/tree/master/windows-builder)
 
 Execute cloud build to build Windows worker:
-```gcloud builds submit --config=containers/buildkite-premerge-windows/cloudbuild.yaml containers/buildkite-premerge-windows/ --project=${PROJECT_ID}```
+```
+gcloud builds submit --config=containers/buildkite-premerge-windows/cloudbuild.yaml containers/buildkite-premerge-windows/ --project=${PROJECT_ID}
+```
 
-##Create the rest of the gcp resources including workers in kubernetes pods
-```terraform apply -var-file="variables.tfvars"```
+## Create the rest of the gcp resources including workers in kubernetes pods
+```
+terraform -chdir=terraform apply -var-file="variables.tfvars"
+```
 
-#Budget
-TODO
+## Terraform cloud build automation
+Manual trigger
+```
+gcloud builds submit --config=terraform/cloudbuild.yaml terraform --project=${PROJECT_ID} --substitutions=_GIT_REPO=${GIT_REPO}
+```
+
+Automatic trigger:
+```
+<TODO>
+```
+
+# Budget
+Budget alerts set on the monthly basis. Notification emails will be triggered on 50%, 90% and 100% of you budget.
+
+## Actions to reduce project costs
+Adjust the GKE cluster nodes: Set the node count to 1 per built platform (Linux and Windows). The node count can be controlled from the `variables.tfvars` file using the `linux-agents-count` and `windows-agents-count` parameters. If these parameters are not set in the `variables.tfvars` file, you can check the default parameters configured in the `variables.tf` file.
+
+## Break glass procedure
+Regarding the break glass procedure for emergency configuration of the Kubernetes build nodes, you can use the following set of gcloud commands. It is recommended to execute them from the  [Cloud Shell](https://cloud.google.com/shell/docs/using-cloud-shell) for simplicity. Please note that these commands assume you have the necessary permissions and authentication to access and modify the GKE cluster.
+```
+export PROJECT_ID=<your project id>
+export ZONE="europe-west3-c"
+
+gcloud container clusters get-credentials ${PROJECT_ID}-cluster --zone ${ZONE} --project ${PROJECT_ID}
+#gcloud container clusters update llvm-premerge-checks-cluster --node-pool linux-agents --zone ${ZONE} --project ${PROJECT_ID} --no-enable-autoscaling
+gcloud container clusters resize llvm-premerge-checks-cluster --node-pool linux-agents --num-nodes 1 --zone ${ZONE} --project ${PROJECT_ID}
+#gcloud container clusters update llvm-premerge-checks-cluster --node-pool windows-agents --zone ${ZONE} --project ${PROJECT_ID} --no-enable-autoscaling
+gcloud container clusters resize llvm-premerge-checks-cluster --node-pool windows-agents --num-nodes 1 --zone ${ZONE} --project ${PROJECT_ID}
+```
+
+These commands will scale down the deployments to have a single replica for both the Linux and Windows agents. Adjust the replica count as needed. Please note that making changes to your GKE cluster configuration or scaling down nodes may impact the availability and performance of the pipeline. 

terraform/billing.tf

@@ -1,10 +1,8 @@
 #todo fix billing alert creation
-data "google_billing_account" "account" {
-  billing_account = var.billing-account
-}
+#todo do not create billing if option in variables is off
 
 resource "google_billing_budget" "budget" {
-  billing_account = data.google_billing_account.account.id
+  billing_account = data.google_project.current_project.billing_account
   display_name    = "budget"
   amount {
     specified_amount {
@@ -13,9 +11,11 @@ resource "google_billing_budget" "budget" {
     }
   }
   
+
   budget_filter {
-    projects               = ["projects/${var.project-id}"]
+    projects               = ["projects/${data.google_project.current_project.number}"]
     credit_types_treatment = "EXCLUDE_ALL_CREDITS"
+    calendar_period        = "MONTH"
     #services = ["services/24E6-581D-38E5"] # Bigquery
   }
 
@@ -29,11 +29,12 @@ resource "google_billing_budget" "budget" {
     threshold_percent = 1.0
   }
 
+  #TODO add if not empty billing admins var. Else use default admins
   all_updates_rule {
     monitoring_notification_channels = [
       for k, v in google_monitoring_notification_channel.notification_channel : google_monitoring_notification_channel.notification_channel[k].id
     ]
-    disable_default_iam_recipients = true
+    disable_default_iam_recipients = length(var.billing-admins) < 1 ? false : true
   }
 }
 

terraform/cloudbuild.yaml

@@ -0,0 +1,57 @@
+steps:
+ - name: gcr.io/cloud-builders/git
+   args:
+     - '-c'
+     - 'git clone ${_GIT_REPO} repo --depth 1'
+   entrypoint: bash
+ - name: hashicorp/terraform
+   args:
+     - init
+     - '-backend-config=bucket=${_TF_BACKEND_BUCKET}'
+     - '-backend-config=prefix=${_TF_BACKEND_PREFIX}'
+   dir: repo/terraform
+ - name: hashicorp/terraform
+   args:
+     - plan
+     - '-var=project-id=${PROJECT_ID}'
+     - '-var=buildkite-api-token-readonly=$$BUILDKITE_API_TOKEN_READONLY'
+     - '-var=buildkite-agent-token=$$BUILDKITE_AGENT_TOKEN'
+     - '-var=conduit-api-token=$$CONDUIT_API_TOKEN'
+     - '-var=git-id-rsa=$$GIT_ID_RSA'
+     - '-var=id-rsa-pub=$$ID_RSA_PUB'
+     - '-var=git-known-hosts=$$GIT_KNOWN_HOSTS'
+     - '-out=/workspace/tfplan-${BUILD_ID}'
+   secretEnv:
+    - 'BUILDKITE_API_TOKEN_READONLY'
+    - 'BUILDKITE_AGENT_TOKEN'
+    - 'CONDUIT_API_TOKEN'
+    - 'GIT_ID_RSA'
+    - 'ID_RSA_PUB'
+    - 'GIT_KNOWN_HOSTS'
+   dir: repo/terraform
+#  - name: hashicorp/terraform
+#    args:
+#      - apply
+#      - '-auto-approve'
+#      - /workspace/tfplan-${BUILD_ID}
+#    dir: repo/terraform
+substitutions:
+ _GIT_REPO: $(body.project.git_http_url)
+ _TF_BACKEND_BUCKET: 'terraform-state-${PROJECT_ID}'
+ _TF_BACKEND_PREFIX: terraform/state
+availableSecrets:
+ secretManager:
+   - versionName: 'projects/${PROJECT_ID}/secrets/buildkite-api-token-readonly/versions/latest'
+     env: 'BUILDKITE_API_TOKEN_READONLY'
+   - versionName: 'projects/${PROJECT_ID}/secrets/buildkite-agent-token/versions/latest'
+     env: 'BUILDKITE_AGENT_TOKEN'
+   - versionName: 'projects/${PROJECT_ID}/secrets/conduit-api-token/versions/latest'
+     env: 'CONDUIT_API_TOKEN'
+   - versionName: 'projects/${PROJECT_ID}/secrets/git-id-rsa/versions/latest'
+     env: 'GIT_ID_RSA'
+   - versionName: 'projects/${PROJECT_ID}/secrets/id-rsa-pub/versions/latest'
+     env: 'ID_RSA_PUB'
+   - versionName: 'projects/${PROJECT_ID}/secrets/git-known-hosts/versions/latest'
+     env: 'GIT_KNOWN_HOSTS'
+options:
+  dynamic_substitutions: true

terraform/cluster.tf

@@ -58,7 +58,7 @@ resource "google_container_cluster" "llvm_premerge_checks_cluster" {
     cluster_secondary_range_name  = "pods"
     services_secondary_range_name = "services"
   }
-  depends_on = [google_project_service.compute_api, google_project_service.container_api]
+  depends_on = [google_project_service.google_api]
 }
 
 resource "google_container_node_pool" "linux_agents_nodepool" {
@@ -81,6 +81,7 @@ resource "google_container_node_pool" "linux_agents_nodepool" {
   autoscaling {
     min_node_count  = 0
     max_node_count  = var.linux-agents-count
+    location_policy = "BALANCED"
   }
 }
 
@@ -104,6 +105,7 @@ resource "google_container_node_pool" "windows_agents_nodepool" {
   autoscaling {
     min_node_count  = 0
     max_node_count  = var.windows-agents-count
+    location_policy = "BALANCED"
   }
 }
 

terraform/main.tf

@@ -5,9 +5,19 @@ data "google_project" "current_project" {
 }
 
 locals {
-  cloud_build_sa_roles = ["roles/storage.objectAdmin", "roles/compute.instanceAdmin", "roles/compute.securityAdmin"]
+  cloud_build_sa_roles = ["roles/editor", "roles/storage.objectAdmin", "roles/secretmanager.secretAccessor","roles/secretmanager.viewer","roles/resourcemanager.projectIamAdmin"]
+  enabled_apis = [
+    "secretmanager.googleapis.com", 
+    "billingbudgets.googleapis.com", 
+    "cloudbuild.googleapis.com", 
+    "compute.googleapis.com", 
+    "container.googleapis.com", 
+    "cloudresourcemanager.googleapis.com", 
+    "cloudbilling.googleapis.com"
+  ]
 }
 
+#todo create separate sa for cloud build
 # data "google_iam_policy" "cloud_build_sa" {
 #   binding {
 #     role = "roles/iam.serviceAccountUser"
@@ -31,34 +41,16 @@ resource "google_project_iam_member" "cloudbuild_sa_roles" {
   member = "serviceAccount:${data.google_project.current_project.number}@cloudbuild.gserviceaccount.com"
 }
 
-resource "google_project_service" "cloudbuild_api" {
-  service = "cloudbuild.googleapis.com"
-}
-
-resource "google_project_service" "compute_api" {
-  service = "compute.googleapis.com"
-}
-
-resource "google_project_service" "container_api" {
-  service = "container.googleapis.com"
-}
-
-resource "google_project_service" "cloudresourcemanager_api" {
-  service = "cloudresourcemanager.googleapis.com"
-}
-
-resource "google_project_service" "cloudbilling_api" {
-  service = "cloudbilling.googleapis.com"
-}
-
-resource "google_project_service" "billingbudgets_api" {
-  service = "billingbudgets.googleapis.com"
+resource "google_project_service" "google_api" {
+  for_each = toset(local.enabled_apis)
+  service = each.value
 }
 
 resource "google_storage_bucket" "terraform_state" {
   name                        = "terraform-state-${var.project-id}"
   uniform_bucket_level_access = true
   location                    = "EU"
+  depends_on = [google_project_service.google_api]
 }
 
 resource "google_compute_network" "vpc_network" {

terraform/secrets.tf

@@ -0,0 +1,26 @@
+locals {
+    secrets = {
+        "buildkite-api-token-readonly": var.buildkite-api-token-readonly, 
+        "buildkite-agent-token": var.buildkite-agent-token, 
+        "conduit-api-token": var.conduit-api-token, 
+        "git-id-rsa": var.git-id-rsa, 
+        "id-rsa-pub": var.id-rsa-pub, 
+        "git-known-hosts": var.git-known-hosts
+    }
+}
+
+resource "google_secret_manager_secret" "secret" {
+  for_each = local.secrets
+  secret_id = each.key
+
+  replication {
+    automatic = true
+  }
+  depends_on = [google_project_service.google_api]
+}
+
+resource "google_secret_manager_secret_version" "secret_version" {
+  for_each = local.secrets
+  secret = google_secret_manager_secret.secret[each.key].id
+  secret_data = each.value
+}

terraform/variables.tf

@@ -2,16 +2,14 @@ variable "project-id" {
   type = string
 }
 
-variable "billing-account" {
-  type = string
-}
-
 variable "billing-budget" {
   type    = number
+  default = "25000"
 }
 
 variable "billing-admins" {
   type    = map(any)
+  default = {}
 }
 
 variable "region" {

terraform/variables.tfvars_example

@@ -1,7 +1,5 @@
 project-id                   = ""
-billing-account              = ""
-billing-budget               = 25000
-billing-admins               = {"test": "test@test.com"}
+#billing-admins               = {"test": "test@test.com"}
 
 #linux-agents-machine-type    = "e2-standard-8"
 #linux-agents-count           = 1