feat: Add wait_for_instances and configurable timeout support for mig (#96)

BREAKING CHANGE: instance_redistribution_type must now be specified for update policies.

Author: seanson

autogen/main.tf.tmpl

@@ -71,6 +71,8 @@ resource "google_compute_region_instance_group_manager" "{{ module_name }}" {
   target_pools = var.target_pools
   target_size  = var.autoscaling_enabled ? null : var.target_size
 
+  wait_for_instances = var.wait_for_instances
+
   dynamic "auto_healing_policies" {
     for_each = local.healthchecks
     content {
@@ -91,20 +93,27 @@ resource "google_compute_region_instance_group_manager" "{{ module_name }}" {
   dynamic "update_policy" {
     for_each = var.update_policy
     content {
-      max_surge_fixed         = lookup(update_policy.value, "max_surge_fixed", null)
-      max_surge_percent       = lookup(update_policy.value, "max_surge_percent", null)
-      max_unavailable_fixed   = lookup(update_policy.value, "max_unavailable_fixed", null)
-      max_unavailable_percent = lookup(update_policy.value, "max_unavailable_percent", null)
-      min_ready_sec           = lookup(update_policy.value, "min_ready_sec", null)
-      minimal_action          = update_policy.value.minimal_action
-      type                    = update_policy.value.type
+      instance_redistribution_type = lookup(update_policy.value, "instance_redistribution_type", null)
+      max_surge_fixed              = lookup(update_policy.value, "max_surge_fixed", null)
+      max_surge_percent            = lookup(update_policy.value, "max_surge_percent", null)
+      max_unavailable_fixed        = lookup(update_policy.value, "max_unavailable_fixed", null)
+      max_unavailable_percent      = lookup(update_policy.value, "max_unavailable_percent", null)
+      min_ready_sec                = lookup(update_policy.value, "min_ready_sec", null)
+      minimal_action               = update_policy.value.minimal_action
+      type                         = update_policy.value.type
     }
   }
 
   lifecycle {
     create_before_destroy = true
     ignore_changes        = [distribution_policy_zones]
   }
+
+  timeouts {
+    create = var.mig_timeouts.create
+    update = var.mig_timeouts.update
+    delete = var.mig_timeouts.delete
+  }
 }
 
 resource "google_compute_region_autoscaler" "autoscaler" {
@@ -143,7 +152,7 @@ resource "google_compute_region_autoscaler" "autoscaler" {
   }
   {% if mig_with_percent %}
 
-  depends_on = ["google_compute_region_instance_group_manager.mig_with_percent"]
+  depends_on = [google_compute_region_instance_group_manager.mig_with_percent]
   {% endif %}
 }
 

autogen/variables.tf.tmpl

@@ -86,13 +86,14 @@ variable "stateful_disks" {
 variable "update_policy" {
   description = "The rolling update policy. https://www.terraform.io/docs/providers/google/r/compute_region_instance_group_manager.html#rolling_update_policy"
   type = list(object({
-    max_surge_fixed         = number
-    max_surge_percent       = number
-    max_unavailable_fixed   = number
-    max_unavailable_percent = number
-    min_ready_sec           = number
-    minimal_action          = string
-    type                    = string
+    max_surge_fixed              = number
+    instance_redistribution_type = string
+    max_surge_percent            = number
+    max_unavailable_fixed        = number
+    max_unavailable_percent      = number
+    min_ready_sec                = number
+    minimal_action               = string
+    type                         = string
   }))
   default = []
 }
@@ -203,3 +204,22 @@ variable "named_ports" {
   }))
   default = []
 }
+
+variable "wait_for_instances" {
+  description = "Whether to wait for all instances to be created/updated before returning. Note that if this is set to true and the operation does not succeed, Terraform will continue trying until it times out."
+  default     = "false"
+}
+
+variable "mig_timeouts" {
+  description = "Times for creation, deleting and updating the MIG resources. Can be helpful when using wait_for_instances to allow a longer VM startup time. "
+  type = object({
+    create = string
+    update = string
+    delete = string
+  })
+  default = {
+    create = "5m"
+    update = "5m"
+    delete = "15m"
+  }
+}

examples/mig/full/variables.tf

@@ -179,13 +179,14 @@ variable "distribution_policy_zones" {
 variable "update_policy" {
   description = "The rolling update policy. https://www.terraform.io/docs/providers/google/r/compute_region_instance_group_manager.html#rolling_update_policy"
   type = list(object({
-    max_surge_fixed         = number
-    max_surge_percent       = number
-    max_unavailable_fixed   = number
-    max_unavailable_percent = number
-    min_ready_sec           = number
-    minimal_action          = string
-    type                    = string
+    max_surge_fixed              = number
+    instance_redistribution_type = string
+    max_surge_percent            = number
+    max_unavailable_fixed        = number
+    max_unavailable_percent      = number
+    min_ready_sec                = number
+    minimal_action               = string
+    type                         = string
   }))
   default = []
 }
@@ -268,4 +269,3 @@ variable "autoscaling_enabled" {
   type        = bool
   default     = false
 }
-

modules/mig/README.md

@@ -28,16 +28,19 @@ The current version is 2.X. The following guides are available to assist with up
 | hostname | Hostname prefix for instances | string | `"default"` | no |
 | instance\_template | Instance template self_link used to create compute instances | string | n/a | yes |
 | max\_replicas | The maximum number of instances that the autoscaler can scale up to. This is required when creating or updating an autoscaler. The maximum number of replicas should not be lower than minimal number of replicas. | string | `"10"` | no |
+| mig\_timeouts | Times for creation, deleting and updating the MIG resources. Can be helpful when using wait_for_instances to allow a longer VM startup time. | object | `<map>` | no |
 | min\_replicas | The minimum number of replicas that the autoscaler can scale down to. This cannot be less than 0. | string | `"2"` | no |
 | named\_ports | Named name and named port. https://cloud.google.com/load-balancing/docs/backend-service#named_ports | object | `<list>` | no |
 | network | Network to deploy to. Only one of network or subnetwork should be specified. | string | `""` | no |
 | project\_id | The GCP project ID | string | `"null"` | no |
 | region | The GCP region where the managed instance group resides. | string | n/a | yes |
+| stateful\_disks | Disks created on the instances that will be preserved on instance delete. https://cloud.google.com/compute/docs/instance-groups/configuring-stateful-disks-in-migs | object | `<list>` | no |
 | subnetwork | Subnet to deploy to. Only one of network or subnetwork should be specified. | string | `""` | no |
 | subnetwork\_project | The project that subnetwork belongs to | string | `""` | no |
 | target\_pools | The target load balancing pools to assign this group to. | list(string) | `<list>` | no |
 | target\_size | The target number of running instances for this managed instance group. This value should always be explicitly set unless this resource is attached to an autoscaler, in which case it should never be set. | string | `"1"` | no |
 | update\_policy | The rolling update policy. https://www.terraform.io/docs/providers/google/r/compute_region_instance_group_manager.html#rolling_update_policy | object | `<list>` | no |
+| wait\_for\_instances | Whether to wait for all instances to be created/updated before returning. Note that if this is set to true and the operation does not succeed, Terraform will continue trying until it times out. | string | `"false"` | no |
 
 ## Outputs
 

modules/mig/main.tf

@@ -55,6 +55,8 @@ resource "google_compute_region_instance_group_manager" "mig" {
   target_pools = var.target_pools
   target_size  = var.autoscaling_enabled ? null : var.target_size
 
+  wait_for_instances = var.wait_for_instances
+
   dynamic "auto_healing_policies" {
     for_each = local.healthchecks
     content {
@@ -90,6 +92,12 @@ resource "google_compute_region_instance_group_manager" "mig" {
     create_before_destroy = true
     ignore_changes        = [distribution_policy_zones]
   }
+
+  timeouts {
+    create = var.mig_timeouts.create
+    update = var.mig_timeouts.update
+    delete = var.mig_timeouts.delete
+  }
 }
 
 resource "google_compute_region_autoscaler" "autoscaler" {

modules/mig/variables.tf

@@ -63,6 +63,7 @@ variable "stateful_disks" {
   }))
   default = []
 }
+
 #################
 # Rolling Update
 #################
@@ -188,3 +189,22 @@ variable "named_ports" {
   }))
   default = []
 }
+
+variable "wait_for_instances" {
+  description = "Whether to wait for all instances to be created/updated before returning. Note that if this is set to true and the operation does not succeed, Terraform will continue trying until it times out."
+  default     = "false"
+}
+
+variable "mig_timeouts" {
+  description = "Times for creation, deleting and updating the MIG resources. Can be helpful when using wait_for_instances to allow a longer VM startup time. "
+  type = object({
+    create = string
+    update = string
+    delete = string
+  })
+  default = {
+    create = "5m"
+    update = "5m"
+    delete = "15m"
+  }
+}

modules/mig_with_percent/README.md

@@ -28,17 +28,20 @@ The current version is 2.X. The following guides are available to assist with up
 | instance\_template\_initial\_version | Instance template self_link used to create compute instances for the initial version | string | n/a | yes |
 | instance\_template\_next\_version | Instance template self_link used to create compute instances for the second version | string | n/a | yes |
 | max\_replicas | The maximum number of instances that the autoscaler can scale up to. This is required when creating or updating an autoscaler. The maximum number of replicas should not be lower than minimal number of replicas. | string | `"10"` | no |
+| mig\_timeouts | Times for creation, deleting and updating the MIG resources. Can be helpful when using wait_for_instances to allow a longer VM startup time. | object | `<map>` | no |
 | min\_replicas | The minimum number of replicas that the autoscaler can scale down to. This cannot be less than 0. | string | `"2"` | no |
 | named\_ports | Named name and named port. https://cloud.google.com/load-balancing/docs/backend-service#named_ports | object | `<list>` | no |
 | network | Network to deploy to. Only one of network or subnetwork should be specified. | string | `""` | no |
 | next\_version\_percent | Percentage of instances defined in the second version | string | n/a | yes |
 | project\_id | The GCP project ID | string | `"null"` | no |
 | region | The GCP region where the managed instance group resides. | string | n/a | yes |
+| stateful\_disks | Disks created on the instances that will be preserved on instance delete. https://cloud.google.com/compute/docs/instance-groups/configuring-stateful-disks-in-migs | object | `<list>` | no |
 | subnetwork | Subnet to deploy to. Only one of network or subnetwork should be specified. | string | `""` | no |
 | subnetwork\_project | The project that subnetwork belongs to | string | `""` | no |
 | target\_pools | The target load balancing pools to assign this group to. | list(string) | `<list>` | no |
 | target\_size | The target number of running instances for this managed instance group. This value should always be explicitly set unless this resource is attached to an autoscaler, in which case it should never be set. | string | `"1"` | no |
 | update\_policy | The rolling update policy. https://www.terraform.io/docs/providers/google/r/compute_region_instance_group_manager.html#rolling_update_policy | object | `<list>` | no |
+| wait\_for\_instances | Whether to wait for all instances to be created/updated before returning. Note that if this is set to true and the operation does not succeed, Terraform will continue trying until it times out. | string | `"false"` | no |
 
 ## Outputs
 

modules/mig_with_percent/main.tf

@@ -64,6 +64,8 @@ resource "google_compute_region_instance_group_manager" "mig_with_percent" {
   target_pools = var.target_pools
   target_size  = var.autoscaling_enabled ? null : var.target_size
 
+  wait_for_instances = var.wait_for_instances
+
   dynamic "auto_healing_policies" {
     for_each = local.healthchecks
     content {
@@ -72,24 +74,39 @@ resource "google_compute_region_instance_group_manager" "mig_with_percent" {
     }
   }
 
+  dynamic "stateful_disk" {
+    for_each = var.stateful_disks
+    content {
+      device_name = stateful_disk.value.device_name
+      delete_rule = lookup(stateful_disk.value, "delete_rule", null)
+    }
+  }
+
   distribution_policy_zones = local.distribution_policy_zones
   dynamic "update_policy" {
     for_each = var.update_policy
     content {
-      max_surge_fixed         = lookup(update_policy.value, "max_surge_fixed", null)
-      max_surge_percent       = lookup(update_policy.value, "max_surge_percent", null)
-      max_unavailable_fixed   = lookup(update_policy.value, "max_unavailable_fixed", null)
-      max_unavailable_percent = lookup(update_policy.value, "max_unavailable_percent", null)
-      min_ready_sec           = lookup(update_policy.value, "min_ready_sec", null)
-      minimal_action          = update_policy.value.minimal_action
-      type                    = update_policy.value.type
+      instance_redistribution_type = lookup(update_policy.value, "instance_redistribution_type", null)
+      max_surge_fixed              = lookup(update_policy.value, "max_surge_fixed", null)
+      max_surge_percent            = lookup(update_policy.value, "max_surge_percent", null)
+      max_unavailable_fixed        = lookup(update_policy.value, "max_unavailable_fixed", null)
+      max_unavailable_percent      = lookup(update_policy.value, "max_unavailable_percent", null)
+      min_ready_sec                = lookup(update_policy.value, "min_ready_sec", null)
+      minimal_action               = update_policy.value.minimal_action
+      type                         = update_policy.value.type
     }
   }
 
   lifecycle {
     create_before_destroy = true
     ignore_changes        = [distribution_policy_zones]
   }
+
+  timeouts {
+    create = var.mig_timeouts.create
+    update = var.mig_timeouts.update
+    delete = var.mig_timeouts.delete
+  }
 }
 
 resource "google_compute_region_autoscaler" "autoscaler" {
@@ -127,7 +144,7 @@ resource "google_compute_region_autoscaler" "autoscaler" {
     }
   }
 
-  depends_on = ["google_compute_region_instance_group_manager.mig_with_percent"]
+  depends_on = [google_compute_region_instance_group_manager.mig_with_percent]
 }
 
 resource "google_compute_health_check" "http" {

modules/mig_with_percent/variables.tf

@@ -60,20 +60,33 @@ variable "distribution_policy_zones" {
   default     = []
 }
 
+#################
+# Stateful disks
+#################
+variable "stateful_disks" {
+  description = "Disks created on the instances that will be preserved on instance delete. https://cloud.google.com/compute/docs/instance-groups/configuring-stateful-disks-in-migs"
+  type = list(object({
+    device_name = string
+    delete_rule = string
+  }))
+  default = []
+}
+
 #################
 # Rolling Update
 #################
 
 variable "update_policy" {
   description = "The rolling update policy. https://www.terraform.io/docs/providers/google/r/compute_region_instance_group_manager.html#rolling_update_policy"
   type = list(object({
-    max_surge_fixed         = number
-    max_surge_percent       = number
-    max_unavailable_fixed   = number
-    max_unavailable_percent = number
-    min_ready_sec           = number
-    minimal_action          = string
-    type                    = string
+    max_surge_fixed              = number
+    instance_redistribution_type = string
+    max_surge_percent            = number
+    max_unavailable_fixed        = number
+    max_unavailable_percent      = number
+    min_ready_sec                = number
+    minimal_action               = string
+    type                         = string
   }))
   default = []
 }
@@ -184,3 +197,22 @@ variable "named_ports" {
   }))
   default = []
 }
+
+variable "wait_for_instances" {
+  description = "Whether to wait for all instances to be created/updated before returning. Note that if this is set to true and the operation does not succeed, Terraform will continue trying until it times out."
+  default     = "false"
+}
+
+variable "mig_timeouts" {
+  description = "Times for creation, deleting and updating the MIG resources. Can be helpful when using wait_for_instances to allow a longer VM startup time. "
+  type = object({
+    create = string
+    update = string
+    delete = string
+  })
+  default = {
+    create = "5m"
+    update = "5m"
+    delete = "15m"
+  }
+}