Merged in MON-75-azure-sql-monitors (pull request #14)

MON-75: Azure SQL Databases monitors

Approved-by: Laurent Piroelle <laurent.piroelle@fr.clara.net>
Approved-by: Alexandre Gaillet <alexandre.gaillet@fr.clara.net>

cloud/azure/sql-database/README.md

@@ -0,0 +1,49 @@
+Azure SQL Database DataDog monitors
+===================================
+
+How to use this module
+----------------------
+
+```
+module "datadog-monitors-azure-storage" {
+  source = "git::ssh://git@bitbucket.org/morea/terraform.feature.datadog.git//cloud/azure/sql-database?ref={revision}"
+
+  message     = "${module.datadog-message-alerting.alerting-message}"
+  environment = "${var.environment}"
+}
+```
+
+Purpose
+-------
+Creates a DataDog monitors with the following checks :
+
+* CPU High
+* Free disk space low
+* DTU Consumption high
+* SQL deadlocks
+
+Inputs
+------
+
+| Name | Description | Type | Default | Required |
+|------|-------------|:----:|:-----:|:-----:|
+| cpu_threshold_critical | CPU usage in percent (critical threshold) | string | `90` | no |
+| cpu_threshold_warning | CPU usage in percent (warning threshold) | string | `` | no |
+| deadlock_threshold_critical | Amount of Deadlocks (critical threshold) | string | `1` | no |
+| delay | Delay in seconds for the metric evaluation | string | `600` | no |
+| diskspace_threshold_critical | Disk space used in percent (critical threshold) | string | `90` | no |
+| diskspace_threshold_warning | Disk space used in percent (warning threshold) | string | `80` | no |
+| dtu_threshold_critical | Amount of DTU used (critical threshold) | string | `90` | no |
+| dtu_threshold_warning | Amount of DTU used (warning threshold) | string | `85` | no |
+| environment | Architecture Environment | string | - | yes |
+| 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 |
+
+Related documentation
+---------------------
+
+DataDog documentation: [https://docs.datadoghq.com/integrations/azure_sql_database/](https://docs.datadoghq.com/integrations/azure_sql_database/)
+
+Azure SQL Database metrics documentation: [https://docs.microsoft.com/en-us/azure/sql-database/saas-dbpertenant-log-analytics](https://docs.microsoft.com/en-us/azure/sql-database/saas-dbpertenant-log-analytics)
+

cloud/azure/sql-database/inputs.tf

@@ -0,0 +1,62 @@
+# Global Terraform
+variable "environment" {
+  description = "Architecture Environment"
+  type        = "string"
+}
+
+# Global DataDog
+variable "delay" {
+  description = "Delay in seconds for the metric evaluation"
+  default     = 600
+}
+
+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     = "*"
+}
+
+# Azure SQL Database specific
+
+variable "cpu_threshold_warning" {
+  description = "CPU usage in percent (warning threshold)"
+  default     = ""
+}
+
+variable "cpu_threshold_critical" {
+  description = "CPU usage in percent (critical threshold)"
+  default     = "90"
+}
+
+variable "diskspace_threshold_warning" {
+  description = "Disk space used in percent (warning threshold)"
+  default     = "80"
+}
+
+variable "diskspace_threshold_critical" {
+  description = "Disk space used in percent (critical threshold)"
+  default     = "90"
+}
+
+variable "dtu_threshold_warning" {
+  description = "Amount of DTU used (warning threshold)"
+  default     = "85"
+}
+
+variable "dtu_threshold_critical" {
+  description = "Amount of DTU used (critical threshold)"
+  default     = "90"
+}
+
+variable "deadlock_threshold_critical" {
+  description = "Amount of Deadlocks (critical threshold)"
+  default     = "1"
+}

cloud/azure/sql-database/monitors-sql-database-basics.tf

@@ -0,0 +1,129 @@
+data "template_file" "filter" {
+  template = "$${filter}"
+
+  vars {
+    filter = "${var.filter_tags_use_defaults == "true" ? format("dd_monitoring:enabled,dd_azure_sqldatabase:enabled,env:%s", var.environment) : "${var.filter_tags_custom}"}"
+  }
+}
+
+resource "datadog_monitor" "sql-database_cpu_90_15min" {
+  name    = "[${var.environment}] SQL Database CPU high > ${var.cpu_threshold_critical}% on {{name}}"
+  message = "${var.message}"
+
+  query = <<EOF
+    avg(last_15m): (
+      avg:azure.sql_servers_databases.cpu_percent{${data.template_file.filter.rendered}} by {name,resource_group}
+    ) > ${var.cpu_threshold_critical}
+  EOF
+
+  type = "metric alert"
+
+  thresholds {
+    critical = "${var.cpu_threshold_critical}"
+  }
+
+  notify_no_data      = true
+  evaluation_delay    = "${var.delay}"
+  renotify_interval   = 0
+  notify_audit        = false
+  timeout_h           = 0
+  include_tags        = true
+  locked              = false
+  require_full_window = true
+  new_host_delay      = "${var.delay}"
+  no_data_timeframe   = 20
+
+  tags = ["env:${var.environment}", "resource:sqldatabase", "team:azure", "provider:azure"]
+}
+
+resource "datadog_monitor" "sql-database_free_space_low" {
+  name    = "[${var.environment}] SQL Database free space < ${var.diskspace_threshold_critical}% on {{name}}"
+  message = "${var.message}"
+
+  type = "metric alert"
+
+  query = <<EOF
+    avg(last_15m): (
+      avg:azure.sql_servers_databases.storage_percent{${data.template_file.filter.rendered}} by {name,resource_group}
+    ) > ${var.diskspace_threshold_critical}
+  EOF
+
+  thresholds {
+    warning  = "${var.diskspace_threshold_warning}"
+    critical = "${var.diskspace_threshold_critical}"
+  }
+
+  notify_no_data      = true
+  evaluation_delay    = "${var.delay}"
+  renotify_interval   = 0
+  notify_audit        = false
+  timeout_h           = 0
+  include_tags        = true
+  locked              = false
+  require_full_window = true
+  new_host_delay      = "${var.delay}"
+  no_data_timeframe   = 20
+
+  tags = ["env:${var.environment}", "resource:sqldatabase", "team:azure", "provider:azure"]
+}
+
+resource "datadog_monitor" "sql-database_dtu_consumption_high" {
+  name    = "[${var.environment}] SQL Database DTU Consumption on {{name}} > ${var.dtu_threshold_critical}"
+  message = "${var.message}"
+
+  type = "metric alert"
+
+  query = <<EOF
+    avg(last_15m): (
+      azure.sql_servers_databases.dtu_consumption_percent{${data.template_file.filter.rendered}} by {name,resource_group}
+    ) > ${var.dtu_threshold_critical}
+  EOF
+
+  thresholds {
+    warning  = "${var.dtu_threshold_warning}"
+    critical = "${var.dtu_threshold_critical}"
+  }
+
+  notify_no_data      = true
+  evaluation_delay    = "${var.delay}"
+  renotify_interval   = 0
+  notify_audit        = false
+  timeout_h           = 0
+  include_tags        = true
+  locked              = false
+  require_full_window = true
+  new_host_delay      = "${var.delay}"
+  no_data_timeframe   = 20
+
+  tags = ["env:${var.environment}", "resource:sqldatabase", "team:azure", "provider:azure"]
+}
+
+resource "datadog_monitor" "sql-database_deadlocks_count" {
+  name    = "[${var.environment}] SQL Database Deadlocks too high on {{name}}"
+  message = "${var.message}"
+
+  type = "metric alert"
+
+  query = <<EOF
+    sum(last_5m): (
+      avg:azure.sql_servers_databases.deadlock{${data.template_file.filter.rendered}} by {name,resource_group}.as_count()
+    ) > ${var.deadlock_threshold_critical}
+  EOF
+
+  thresholds {
+    critical = "${var.deadlock_threshold_critical}"
+  }
+
+  notify_no_data      = false
+  evaluation_delay    = "${var.delay}"
+  renotify_interval   = 0
+  notify_audit        = false
+  timeout_h           = 0
+  include_tags        = true
+  locked              = false
+  require_full_window = true
+  new_host_delay      = "${var.delay}"
+  no_data_timeframe   = 20
+
+  tags = ["env:${var.environment}", "resource:sqldatabase", "team:azure", "provider:azure"]
+}