add validation rules to monitoring variables

.github/scripts/check.sh

@@ -48,7 +48,7 @@ update_component_status() {
 
 # Function to create an incident
 create_incident() {
-    local incident_name="Testing Instatus"
+    local incident_name="Degraded Service"
     local message="The following modules are experiencing issues:\n"
     for i in "${!failures[@]}"; do
         message+="$((i + 1)). ${failures[$i]}\n"
@@ -59,7 +59,7 @@ create_incident() {
         component_status="MAJOROUTAGE"
     fi
     # see https://instatus.com/help/api/incidents
-    response=$(curl -s -X POST "https://api.instatus.com/v1/$INSTATUS_PAGE_ID/incidents" \
+    incident_id=$(curl -s -X POST "https://api.instatus.com/v1/$INSTATUS_PAGE_ID/incidents" \
         -H "Authorization: Bearer $INSTATUS_API_KEY" \
         -H "Content-Type: application/json" \
         -d "{
@@ -74,10 +74,25 @@ create_incident() {
                     \"status\": \"PARTIALOUTAGE\"
                 }
             ]
-        }")
+        }" | jq -r '.id')
 
-    incident_id=$(echo "$response" | jq -r '.id')
-    echo "$incident_id"
+    echo "Created incident with ID: $incident_id"
+}
+
+# Function to check for existing unresolved incidents
+check_existing_incident() {
+    # Fetch the latest incidents with status not equal to "RESOLVED"
+    local unresolved_incidents=$(curl -s -X GET "https://api.instatus.com/v1/$INSTATUS_PAGE_ID/incidents" \
+        -H "Authorization: Bearer $INSTATUS_API_KEY" \
+        -H "Content-Type: application/json" | jq -r '.incidents[] | select(.status != "RESOLVED") | .id')
+
+    if [[ -n "$unresolved_incidents" ]]; then
+        echo "Unresolved incidents found: $unresolved_incidents"
+        return 0  # Indicate that there are unresolved incidents
+    else
+        echo "No unresolved incidents found."
+        return 1  # Indicate that no unresolved incidents exist
+    fi
 }
 
 force_redeploy_registry () {
@@ -174,9 +189,10 @@ else
         update_component_status "PARTIALOUTAGE"
     fi
 
-    # Create a new incident
-    incident_id=$(create_incident)
-    echo "Created incident with ID: $incident_id"
+    # Check if there is an existing incident before creating a new one
+    if ! check_existing_incident; then
+        create_incident
+    fi
 
     # If a module is down, force a reployment to try getting things back online
     # ASAP

monitoring/README.md

@@ -0,0 +1,72 @@
+---
+display_name: Monitoring
+description: Monitoring of workspace resources
+maintainer_github: coder
+verified: true
+tags: [monitoring]
+---
+
+# Monitoring
+
+This module adds monitoring of workspace resources.
+
+```tf
+module "monitoring" {
+  source  = "registry.coder.com/modules/monitoring/coder"
+  version = "1.0.0"
+  agent_id = coder_agent.dev.id
+}
+```
+
+## Examples
+
+```tf
+module "monitoring" {
+  source  = "registry.coder.com/modules/monitoring/coder"
+  version = "1.0.0"
+  agent_id = coder_agent.dev.id
+}
+```
+
+### Enable/Disable
+
+You can customize the monitoring by setting the `enabled`, `memory_enabled`, and `disk_enabled` variables.
+
+```tf
+module "monitoring" {
+  source  = "registry.coder.com/modules/monitoring/coder"
+  version = "1.0.0"
+  agent_id = coder_agent.dev.id
+  enabled = false
+  memory_enabled = true
+  disk_enabled = false
+}
+```
+
+### Customize Thresholds
+
+You can customize the thresholds by setting the `threshold`, `memory_threshold`, and `disk_threshold` variables.
+
+```tf
+module "monitoring" {
+  source  = "registry.coder.com/modules/monitoring/coder"
+  version = "1.0.0"
+  agent_id = coder_agent.dev.id
+  threshold = 90
+  memory_threshold = 95
+  disk_threshold = 90
+}
+```
+
+### Customize Disks
+
+You can customize the disks by setting the `disks` variable.
+
+```tf
+module "monitoring" {
+  source  = "registry.coder.com/modules/monitoring/coder"
+  version = "1.0.0"
+  agent_id = coder_agent.dev.id
+  disks = ["/"]
+}
+```

monitoring/main.tf

@@ -0,0 +1,92 @@
+terraform {
+  required_version = ">= 1.0.25"
+
+  required_providers {
+    coder = {
+      source  = "coder/coder"
+      version = ">= 2.0.2"
+    }
+  }
+}
+
+variable "threshold" {
+  type        = number
+  description = "The threshold for the monitoring, used for all resources unless overridden by *_threshold - expressed as a percentage."
+  default     = 90
+  validation {
+    condition     = var.threshold >= 0 && var.threshold <= 100
+    error_message = "The threshold must be between 0 and 100."
+  }
+}
+
+variable "memory_threshold" {
+  type        = number
+  description = "The threshold for the memory monitoring - expressed as a percentage."
+  default     = 90
+  validation {
+    condition     = var.memory_threshold >= 0 && var.memory_threshold <= 100
+    error_message = "The memory_threshold must be between 0 and 100."
+  }
+}
+
+variable "disk_threshold" {
+  type        = number
+  description = "The threshold for the disk monitoring - expressed as a percentage."
+  default     = 90
+  validation {
+    condition     = var.disk_threshold >= 0 && var.disk_threshold <= 100
+    error_message = "The disk_threshold must be between 0 and 100."
+  }
+}
+
+variable "disks" {
+  type        = list(string)
+  description = "The disks to monitor. e.g. ['/', '/home']"
+  default     = ["/"]
+}
+
+variable "enabled" {
+  type        = bool
+  description = "Whether the monitoring is enabled."
+  default     = true
+  validation {
+    condition     = var.enabled == true || var.enabled == false
+    error_message = "The enabled must be true or false."
+  }
+}
+
+variable "memory_enabled" {
+  type        = bool
+  description = "Whether the memory monitoring is enabled."
+  default     = true
+  validation {
+    condition     = var.memory_enabled == true || var.memory_enabled == false
+    error_message = "The memory_enabled must be true or false."
+  }
+}
+
+variable "disk_enabled" {
+  type        = bool
+  description = "Whether the disk monitoring is enabled."
+  default     = true
+  validation {
+    condition     = var.disk_enabled == true || var.disk_enabled == false
+    error_message = "The disk_enabled must be true or false."
+  }
+}
+
+variable "agent_id" {
+  type        = string
+  description = "The ID of the agent to monitor."
+}
+
+data "coder_monitoring" "monitoring" {
+  threshold    = var.threshold
+  memory_threshold = var.memory_threshold
+  disk_threshold = var.disk_threshold
+  disks = var.disks
+  enabled = var.enabled
+  memory_enabled = var.memory_enabled
+  disk_enabled = var.disk_enabled
+  agent_id = var.agent_id
+}