MON-105: Add postgresql monitors

databases/postgresql/README.md

@@ -0,0 +1,45 @@
+PostgreSQL DataDog monitors
+======================
+
+How to use this module
+----------------------
+
+```
+module "datadog-monitors-postgresql" {
+  source = "git::ssh://git@bitbucket.org/morea/terraform.feature.datadog.git//databases/postgresql?ref={revision}"
+
+  environment = "${var.environment}"
+  message = "${module.datadog-message-alerting.alerting-message}"
+}
+
+```
+Purpose
+-------
+Creates DataDog monitors with the following checks :
+
+* PostgreSQL not enough available connections
+* PostgreSQL too many locks
+
+Inputs
+------
+
+| Name | Description | Type | Default | Required |
+|------|-------------|:----:|:-----:|:-----:|
+| delay | Delay in seconds for the metric evaluation | string | `15` | no |
+| environment | Environment | string | - | yes |
+| filter_tags_use_defaults | Use default filter tags convention | string | `true` | no |
+| filter_tags_custom | Tags used for custom filtering when filter_tags_use_defaults is false | string | `*` | no |
+| postgresql_connection_threshold_critical | Maximum critical acceptable percent of connections | string | `80` | no |
+| postgresql_connection_threshold_warning | Maximum warning acceptable percent of connections | string | `70` | no |
+| postgresql_connection_silenced | Groups to mute PostgreSQL connection monitor | map | `<map>` | no |
+| postgresql_connection_message | Custom message for PostgreSQL connection monitor | string | `` | no |
+| postgresql_thread_threshold_critical | Maximum critical acceptable number of threads | string | `500` | no |
+| postgresql_thread_threshold_warning | Maximum warning acceptable number of threads | string | `400` | no |
+| postgresql_thread_silenced | Groups to mute PostgreSQL threads monitor | map | `<map>` | no |
+| postgresql_thread_message | Custom message for PostgreSQL thread monitor | string | `` | no |
+| message | Message sent when a monitor is triggered | string | - | yes |
+
+Related documentation
+---------------------
+
+DataDog documentation: [https://docs.datadoghq.com/integrations/postgres/](https://docs.datadoghq.com/integrations/postgres/)

databases/postgresql/inputs.tf

@@ -0,0 +1,77 @@
+variable "environment" {
+  description = "Environment"
+  type        = "string"
+}
+
+# Global DataDog
+variable "delay" {
+  description = "Delay in seconds for the metric evaluation"
+  default     = 15
+}
+
+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     = "*"
+}
+
+# PostgreSQL specific
+##################################
+###   PostgreSQL connections   ###
+##################################
+
+variable "postgresql_connection_threshold_critical" {
+  default     = 80
+  description = "Maximum critical acceptable percent of connections"
+}
+
+variable "postgresql_connection_threshold_warning" {
+  default     = 70
+  description = "Maximum warning acceptable percent of connections"
+}
+
+variable "postgresql_connection_silenced" {
+  description = "Groups to mute PostgreSQL connection monitor"
+  type        = "map"
+  default     = {}
+}
+
+variable "postgresql_connection_message" {
+  description = "Custom message for PostgreSQL connection monitor"
+  type        = "string"
+  default     = ""
+}
+
+############################
+###   PostgreSQL locks   ###
+############################
+
+variable "postgresql_lock_threshold_critical" {
+  default     = 99
+  description = "Maximum critical acceptable number of locks"
+}
+
+variable "postgresql_lock_threshold_warning" {
+  default     = 70
+  description = "Maximum warning acceptable number of locks"
+}
+
+variable "posgresql_lock_silenced" {
+  description = "Groups to mute PostgreSQL lock monitor"
+  type        = "map"
+  default     = {}
+}
+
+variable "postgresql_lock_message" {
+  description = "Custom message for PostgreSQL lock monitor"
+  type        = "string"
+  default     = ""
+}

databases/postgresql/mon-postgresql.tf

@@ -0,0 +1,67 @@
+data "template_file" "filter" {
+  template = "$${filter}"
+
+  vars {
+    filter = "${var.filter_tags_use_defaults == "true" ? format("dd_monitoring:enabled,dd_postgres:enabled,db_env:%s", var.environment) : "${var.filter_tags_custom}"}"
+  }
+}
+
+resource "datadog_monitor" "postgresql_connection_too_high" {
+  name    = "[${var.environment}] PostgreSQL Connections {{#is_alert}}{{{comparator}}} {{threshold}}% ({{value}}%){{/is_alert}}{{#is_warning}}{{{comparator}}} {{warn_threshold}}% ({{value}}%){{/is_warning}}"
+  message = "${coalesce(var.postgresql_connection_message, var.message)}"
+  type    = "metric alert"
+
+  query = <<EOF
+    avg(last_15m): (
+      avg:postgresql.percent_usage_connections{${data.template_file.filter.rendered}} by {db_host}
+    ) * 100 > ${var.postgresql_connection_threshold_critical}
+  EOF
+
+  evaluation_delay = "${var.delay}"
+  new_host_delay   = "${var.delay}"
+
+  thresholds {
+    warning  = "${var.postgresql_connection_threshold_warning}"
+    critical = "${var.postgresql_connection_threshold_critical}"
+  }
+
+  notify_no_data      = true
+  renotify_interval   = 0
+  require_full_window = true
+  timeout_h           = 0
+  include_tags        = true
+
+  silenced = "${var.postgresql_connection_silenced}"
+
+  tags = ["env:${var.environment}", "resource:postgresql"]
+}
+
+resource "datadog_monitor" "postgresql_too_many_locks" {
+  name    = "[${var.environment}] PostgreSQL too many locks {{#is_alert}}{{{comparator}}} {{threshold}} ({{value}}){{/is_alert}}{{#is_warning}}{{{comparator}}} {{warn_threshold}} ({{value}}){{/is_warning}}"
+  message = "${coalesce(var.postgresql_lock_message, var.message)}"
+  type    = "metric alert"
+
+  query = <<EOF
+    avg(last_5m): (
+      avg:postgresql.locks{${data.template_file.filter.rendered}} by {db_host}
+    ) > ${var.postgresql_lock_threshold_critical}
+  EOF
+
+  evaluation_delay = "${var.delay}"
+  new_host_delay   = "${var.delay}"
+
+  thresholds {
+    warning  = "${var.postgresql_lock_threshold_warning}"
+    critical = "${var.postgresql_lock_threshold_critical}"
+  }
+
+  notify_no_data      = true
+  renotify_interval   = 0
+  require_full_window = true
+  timeout_h           = 0
+  include_tags        = true
+
+  silenced = "${var.posgresql_lock_silenced}"
+
+  tags = ["env:${var.environment}", "resource:postgresql"]
+}