Merged in MON-286-host-unreachable-monitor-on-syst (pull request #155)

MON-286 remove notify no data on metrics monitors

Approved-by: Quentin Manfroi <quentin.manfroi@yahoo.fr>
Approved-by: Jean-Philippe LAINÉ <jean-philippe.laine@fr.clara.net>
Approved-by: Rafael Romero Carmona <rafael.romero.carmona@fr.clara.net>

README.md

@@ -116,3 +116,4 @@ The `//` is very important, it's a terraform specific syntax used to separate gi
 	- [php-fpm](https://bitbucket.org/morea/terraform.feature.datadog/src/master/middleware/php-fpm/)
 - [system](https://bitbucket.org/morea/terraform.feature.datadog/src/master/system/)
 	- [generic](https://bitbucket.org/morea/terraform.feature.datadog/src/master/system/generic/)
+	- [unreachable](https://bitbucket.org/morea/terraform.feature.datadog/src/master/system/unreachable/)

system/generic/README.md

@@ -22,7 +22,6 @@ Creates DataDog monitors with the following checks:
 - Free disk inodes
 - Free disk space
 - Free memory
-- Host unreachable
 
 ## Inputs
 
@@ -88,11 +87,6 @@ Creates DataDog monitors with the following checks:
 | free_memory_timeframe | Monitor timeframe for Free memory [available values: `last_#m` (1, 5, 10, 15, or 30), `last_#h` (1, 2, or 4), or `last_1d`] | string | `last_5m` | no |
 | message | Message sent when an alert is triggered | string | - | yes |
 | new_host_delay | Delay in seconds before monitor new resource | string | `300` | no |
-| unreachable_enabled | Flag to enable Host unreachable monitor | string | `true` | no |
-| unreachable_extra_tags | Extra tags for Host unreachable monitor | list | `<list>` | no |
-| unreachable_message | Custom message for Host unreachable monitor | string | `` | no |
-| unreachable_no_data_timeframe | Timeframe for Host unreachable monitor to alert on no data | string | `2` | no |
-| unreachable_silenced | Groups to mute for Host unreachable monitor | map | `<map>` | no |
 
 ## Outputs
 
@@ -103,7 +97,6 @@ Creates DataDog monitors with the following checks:
 | datadog_free_disk_space_inodes_too_low_id | id for monitor datadog_free_disk_space_inodes_too_low |
 | datadog_free_disk_space_too_low_id | id for monitor datadog_free_disk_space_too_low |
 | datadog_free_memory_id | id for monitor datadog_free_memory |
-| datadog_host_unreachable_id | id for monitor datadog_host_unreachable |
 | datadog_load_too_high_id | id for monitor datadog_load_too_high |
 
 ## Related documentation

system/generic/inputs.tf

@@ -31,36 +31,6 @@ variable "filter_tags_custom" {
 
 # System generic specific
 
-variable "unreachable_silenced" {
-  description = "Groups to mute for Host unreachable monitor"
-  type        = "map"
-  default     = {}
-}
-
-variable "unreachable_enabled" {
-  description = "Flag to enable Host unreachable monitor"
-  type        = "string"
-  default     = "true"
-}
-
-variable "unreachable_extra_tags" {
-  description = "Extra tags for Host unreachable monitor"
-  type        = "list"
-  default     = []
-}
-
-variable "unreachable_message" {
-  description = "Custom message for Host unreachable monitor"
-  type        = "string"
-  default     = ""
-}
-
-variable "unreachable_no_data_timeframe" {
-  description = "Timeframe for Host unreachable monitor to alert on no data"
-  type        = "string"
-  default     = 2
-}
-
 variable "cpu_high_silenced" {
   description = "Groups to mute for CPU high monitor"
   type        = "map"

system/generic/monitors-system.tf

@@ -1,32 +1,3 @@
-resource "datadog_monitor" "datadog_host_unreachable" {
-  count   = "${var.unreachable_enabled ? 1 : 0}"
-  name    = "[${var.environment}] Host unreachable"
-  message = "${coalesce(var.unreachable_message, var.message)}"
-
-  query = "\"datadog.agent.up\".over${module.filter-tags.service_check}.last(6).count_by_status()"
-
-  type = "service check"
-
-  thresholds {
-    ok       = 1
-    warning  = 1
-    critical = 5
-  }
-
-  notify_no_data      = true
-  new_host_delay      = "${var.new_host_delay}"
-  notify_audit        = false
-  timeout_h           = 0
-  include_tags        = true
-  locked              = false
-  require_full_window = true
-  no_data_timeframe   = "${var.unreachable_no_data_timeframe}"
-
-  silenced = "${var.unreachable_silenced}"
-
-  tags = ["env:${var.environment}", "type:system", "provider:host", "resource:generic", "team:claranet", "created-by:terraform", "${var.unreachable_extra_tags}"]
-}
-
 resource "datadog_monitor" "datadog_cpu_too_high" {
   count   = "${var.cpu_high_enabled ? 1 : 0}"
   name    = "[${var.environment}] CPU usage {{#is_alert}}{{{comparator}}} {{threshold}}% ({{value}}%){{/is_alert}}{{#is_warning}}{{{comparator}}} {{warn_threshold}}% ({{value}}%){{/is_warning}}"
@@ -45,7 +16,7 @@ resource "datadog_monitor" "datadog_cpu_too_high" {
     critical = "${var.cpu_high_threshold_critical}"
   }
 
-  notify_no_data      = true
+  notify_no_data      = false
   evaluation_delay    = "${var.evaluation_delay}"
   new_host_delay      = "${var.new_host_delay}"
   notify_audit        = false
@@ -78,7 +49,7 @@ resource "datadog_monitor" "datadog_load_too_high" {
     critical = "${var.cpu_load_threshold_critical}"
   }
 
-  notify_no_data      = true
+  notify_no_data      = false
   evaluation_delay    = "${var.evaluation_delay}"
   new_host_delay      = "${var.new_host_delay}"
   notify_audit        = false
@@ -111,7 +82,7 @@ resource "datadog_monitor" "datadog_free_disk_space_too_low" {
     critical = "${var.free_disk_space_threshold_critical}"
   }
 
-  notify_no_data      = true
+  notify_no_data      = false
   evaluation_delay    = "${var.evaluation_delay}"
   new_host_delay      = "${var.new_host_delay}"
   notify_audit        = false
@@ -184,7 +155,7 @@ resource "datadog_monitor" "datadog_free_disk_space_inodes_too_low" {
     critical = "${var.free_disk_inodes_threshold_critical}"
   }
 
-  notify_no_data      = true
+  notify_no_data      = false
   evaluation_delay    = "${var.evaluation_delay}"
   new_host_delay      = "${var.new_host_delay}"
   notify_audit        = false
@@ -217,7 +188,7 @@ resource "datadog_monitor" "datadog_free_memory" {
     critical = "${var.free_memory_threshold_critical}"
   }
 
-  notify_no_data      = true
+  notify_no_data      = false
   evaluation_delay    = "${var.evaluation_delay}"
   new_host_delay      = "${var.new_host_delay}"
   renotify_interval   = 0

system/generic/outputs.tf

@@ -1,8 +1,3 @@
-output "datadog_host_unreachable_id" {
-  description = "id for monitor datadog_host_unreachable"
-  value       = "${datadog_monitor.datadog_host_unreachable.*.id}"
-}
-
 output "datadog_cpu_too_high_id" {
   description = "id for monitor datadog_cpu_too_high"
   value       = "${datadog_monitor.datadog_cpu_too_high.*.id}"

system/unreachable/README.md

@@ -0,0 +1,44 @@
+# SYSTEM UNREACHABLE DataDog monitors
+
+## How to use this module
+
+```
+module "datadog-monitors-system-unreachable" {
+  source = "git::ssh://git@bitbucket.org/morea/terraform.feature.datadog.git//system/unreachable?ref={revision}"
+
+  environment = "${var.environment}"
+  message     = "${module.datadog-message-alerting.alerting-message}"
+}
+
+```
+
+## Purpose
+
+Creates DataDog monitors with the following checks:
+
+- Host unreachable
+
+## Inputs
+
+| Name | Description | Type | Default | Required |
+|------|-------------|:----:|:-----:|:-----:|
+| environment | Architecture Environment | string | - | yes |
+| evaluation_delay | Delay in seconds for the metric evaluation | string | `15` | no |
+| filter_tags_custom | Tags used for custom filtering when filter_tags_use_defaults is false | string | `*` | no |
+| filter_tags_use_defaults | Use default filter tags convention | string | `true` | no |
+| message | Message sent when an alert is triggered | string | - | yes |
+| new_host_delay | Delay in seconds before monitor new resource | string | `300` | no |
+| unreachable_enabled | Flag to enable Host unreachable monitor | string | `true` | no |
+| unreachable_extra_tags | Extra tags for Host unreachable monitor | list | `<list>` | no |
+| unreachable_message | Custom message for Host unreachable monitor | string | `` | no |
+| unreachable_no_data_timeframe | Timeframe for Host unreachable monitor to alert on no data | string | `20` | no |
+| unreachable_silenced | Groups to mute for Host unreachable monitor | map | `<map>` | no |
+
+## Outputs
+
+| Name | Description |
+|------|-------------|
+| host_unreachable_id | id for monitor host_unreachable |
+
+## Related documentation
+

system/unreachable/inputs.tf

@@ -0,0 +1,62 @@
+# Global Terraform
+variable "environment" {
+  description = "Architecture Environment"
+  type        = "string"
+}
+
+# Global DataDog
+variable "evaluation_delay" {
+  description = "Delay in seconds for the metric evaluation"
+  default     = 15
+}
+
+variable "new_host_delay" {
+  description = "Delay in seconds before monitor new resource"
+  default     = 300
+}
+
+variable "message" {
+  description = "Message sent when an alert is triggered"
+}
+
+variable "filter_tags_use_defaults" {
+  description = "Use default filter tags convention"
+  default     = "true"
+}
+
+variable "filter_tags_custom" {
+  description = "Tags used for custom filtering when filter_tags_use_defaults is false"
+  default     = "*"
+}
+
+# Unreachable
+
+variable "unreachable_silenced" {
+  description = "Groups to mute for Host unreachable monitor"
+  type        = "map"
+  default     = {}
+}
+
+variable "unreachable_enabled" {
+  description = "Flag to enable Host unreachable monitor"
+  type        = "string"
+  default     = "true"
+}
+
+variable "unreachable_extra_tags" {
+  description = "Extra tags for Host unreachable monitor"
+  type        = "list"
+  default     = []
+}
+
+variable "unreachable_message" {
+  description = "Custom message for Host unreachable monitor"
+  type        = "string"
+  default     = ""
+}
+
+variable "unreachable_no_data_timeframe" {
+  description = "Timeframe for Host unreachable monitor to alert on no data"
+  type        = "string"
+  default     = 20
+}

system/unreachable/modules.tf

@@ -0,0 +1,8 @@
+module "filter-tags" {
+  source = "../../common/filter-tags"
+
+  environment              = "${var.environment}"
+  resource                 = "system"
+  filter_tags_use_defaults = "${var.filter_tags_use_defaults}"
+  filter_tags_custom       = "${var.filter_tags_custom}"
+}

system/unreachable/monitors-unreachable.tf

@@ -0,0 +1,28 @@
+resource "datadog_monitor" "host_unreachable" {
+  count   = "${var.unreachable_enabled ? 1 : 0}"
+  name    = "[${var.environment}] Host unreachable"
+  message = "${coalesce(var.unreachable_message, var.message)}"
+
+  query = "\"datadog.agent.up\".over${module.filter-tags.service_check}.last(6).count_by_status()"
+
+  type = "service check"
+
+  thresholds {
+    ok       = 1
+    warning  = 1
+    critical = 5
+  }
+
+  notify_no_data      = true
+  no_data_timeframe   = "${var.unreachable_no_data_timeframe}"
+  new_host_delay      = "${var.new_host_delay}"
+  notify_audit        = false
+  timeout_h           = 0
+  include_tags        = true
+  locked              = false
+  require_full_window = true
+
+  silenced = "${var.unreachable_silenced}"
+
+  tags = ["env:${var.environment}", "type:system", "provider:host", "resource:unreachable", "team:claranet", "created-by:terraform", "${var.unreachable_extra_tags}"]
+}

system/unreachable/outputs.tf

@@ -0,0 +1,4 @@
+output "host_unreachable_id" {
+  description = "id for monitor host_unreachable"
+  value       = "${datadog_monitor.host_unreachable.*.id}"
+}