Merge branch 'MON-377-azure-sql-server-status' into 'master'

MON-377 Azure SQL Server status

Closes MON-377

See merge request claranet/cloudnative/projects/datadog/terraform/monitors!10

cloud/azure/sql-database/README.md

@@ -19,6 +19,7 @@ Creates DataDog monitors with the following checks:
 - SQL Database CPU too high
 - SQL Database Deadlocks too high
 - SQL Database DTU Consumption too high
+- SQL Database is down
 - SQL Database low free space
 
 ## Inputs
@@ -62,6 +63,12 @@ Creates DataDog monitors with the following checks:
 | 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 |
+| status_enabled | Flag to enable Redis status monitor | string | `true` | no |
+| status_extra_tags | Extra tags for Redis status monitor | list | `[]` | no |
+| status_message | Custom message for Redis status monitor | string | `` | no |
+| status_silenced | Groups to mute for Redis status monitor | map | `{}` | no |
+| status_time_aggregator | Monitor aggregator for Redis status [available values: min, max or avg] | string | `max` | no |
+| status_timeframe | Monitor timeframe for Redis status [available values: `last_#m` (1, 5, 10, 15, or 30), `last_#h` (1, 2, or 4), or `last_1d`] | string | `last_5m` | no |
 
 ## Outputs
 
@@ -71,6 +78,7 @@ Creates DataDog monitors with the following checks:
 | sql-database_deadlocks_count_id | id for monitor sql-database_deadlocks_count |
 | sql-database_dtu_consumption_high_id | id for monitor sql-database_dtu_consumption_high |
 | sql-database_free_space_low_id | id for monitor sql-database_free_space_low |
+| status_id | id for monitor status |
 
 ## Related documentation
 

cloud/azure/sql-database/inputs.tf

@@ -35,6 +35,42 @@ variable "filter_tags_custom_excluded" {
 }
 
 # Azure SQL Database specific variables
+variable "status_silenced" {
+  description = "Groups to mute for Redis status monitor"
+  type        = "map"
+  default     = {}
+}
+
+variable "status_enabled" {
+  description = "Flag to enable Redis status monitor"
+  type        = "string"
+  default     = "true"
+}
+
+variable "status_extra_tags" {
+  description = "Extra tags for Redis status monitor"
+  type        = "list"
+  default     = []
+}
+
+variable "status_message" {
+  description = "Custom message for Redis status monitor"
+  type        = "string"
+  default     = ""
+}
+
+variable "status_time_aggregator" {
+  description = "Monitor aggregator for Redis status [available values: min, max or avg]"
+  type        = "string"
+  default     = "max"
+}
+
+variable "status_timeframe" {
+  description = "Monitor timeframe for Redis status [available values: `last_#m` (1, 5, 10, 15, or 30), `last_#h` (1, 2, or 4), or `last_1d`]"
+  type        = "string"
+  default     = "last_5m"
+}
+
 variable "cpu_silenced" {
   description = "Groups to mute for SQL CPU monitor"
   type        = "map"

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

@@ -1,3 +1,31 @@
+resource "datadog_monitor" "status" {
+  count   = "${var.status_enabled ? 1 : 0}"
+  name    = "[${var.environment}] SQL Database is down"
+  message = "${coalesce(var.status_message, var.message)}"
+
+  query = <<EOF
+    ${var.status_time_aggregator}(${var.status_timeframe}): (
+      avg:azure.sql_servers_databases.status${module.filter-tags.query_alert} by {resource_group,region,server_name,name}
+    ) != 1
+  EOF
+
+  type = "metric alert"
+
+  silenced = "${var.status_silenced}"
+
+  notify_no_data      = true
+  evaluation_delay    = "${var.evaluation_delay}"
+  renotify_interval   = 0
+  notify_audit        = false
+  timeout_h           = 0
+  include_tags        = true
+  locked              = false
+  require_full_window = false
+  new_host_delay      = "${var.new_host_delay}"
+
+  tags = ["env:${var.environment}", "type:cloud", "provider:azure", "resource:sql-database", "team:claranet", "created-by:terraform", "${var.status_extra_tags}"]
+}
+
 resource "datadog_monitor" "sql-database_cpu_90_15min" {
   count   = "${var.cpu_enabled ? 1 : 0}"
   name    = "[${var.environment}] SQL Database CPU too high {{#is_alert}}{{{comparator}}} {{threshold}}% ({{value}}%){{/is_alert}}{{#is_warning}}{{{comparator}}} {{warn_threshold}}% ({{value}}%){{/is_warning}}"

cloud/azure/sql-database/outputs.tf

@@ -1,3 +1,8 @@
+output "status_id" {
+  description = "id for monitor status"
+  value       = "${datadog_monitor.status.*.id}"
+}
+
 output "sql-database_cpu_90_15min_id" {
   description = "id for monitor sql-database_cpu_90_15min"
   value       = "${datadog_monitor.sql-database_cpu_90_15min.*.id}"