front-integrations/terraform/alarms.tf

resource "aws_cloudwatch_metric_alarm" "sla_monitor_errors" {
  alarm_name          = "front-sla-monitor-errors"
  alarm_description   = "front-sla-monitor invocation errors"
  namespace           = "AWS/Lambda"
  metric_name         = "Errors"
  statistic           = "Sum"
  period              = 300
  evaluation_periods  = 1
  threshold           = 0
  comparison_operator = "GreaterThanThreshold"
  treat_missing_data  = "notBreaching"
  alarm_actions       = [var.alarm_sns_topic_arn]

  dimensions = {
    FunctionName = aws_lambda_function.sla_monitor.function_name
  }
}

resource "aws_cloudwatch_metric_alarm" "sla_monitor_throttles" {
  alarm_name          = "front-sla-monitor-throttles"
  alarm_description   = "front-sla-monitor invocations throttled"
  namespace           = "AWS/Lambda"
  metric_name         = "Throttles"
  statistic           = "Sum"
  period              = 300
  evaluation_periods  = 1
  threshold           = 0
  comparison_operator = "GreaterThanThreshold"
  treat_missing_data  = "notBreaching"
  alarm_actions       = [var.alarm_sns_topic_arn]

  dimensions = {
    FunctionName = aws_lambda_function.sla_monitor.function_name
  }
}

resource "aws_cloudwatch_metric_alarm" "sla_monitor_duration" {
  alarm_name          = "front-sla-monitor-duration"
  alarm_description   = "front-sla-monitor approaching its 300s timeout (>90%)"
  namespace           = "AWS/Lambda"
  metric_name         = "Duration"
  statistic           = "Maximum"
  period              = 300
  evaluation_periods  = 1
  threshold           = 270000
  comparison_operator = "GreaterThanThreshold"
  treat_missing_data  = "notBreaching"
  alarm_actions       = [var.alarm_sns_topic_arn]

  dimensions = {
    FunctionName = aws_lambda_function.sla_monitor.function_name
  }
}

resource "aws_cloudwatch_metric_alarm" "user_sync_errors" {
  alarm_name          = "front-user-sync-errors"
  alarm_description   = "front-user-sync invocation errors"
  namespace           = "AWS/Lambda"
  metric_name         = "Errors"
  statistic           = "Sum"
  period              = 300
  evaluation_periods  = 1
  threshold           = 0
  comparison_operator = "GreaterThanThreshold"
  treat_missing_data  = "notBreaching"
  alarm_actions       = [var.alarm_sns_topic_arn]

  dimensions = {
    FunctionName = aws_lambda_function.user_sync.function_name
  }
}

resource "aws_cloudwatch_metric_alarm" "user_sync_throttles" {
  alarm_name          = "front-user-sync-throttles"
  alarm_description   = "front-user-sync invocations throttled"
  namespace           = "AWS/Lambda"
  metric_name         = "Throttles"
  statistic           = "Sum"
  period              = 300
  evaluation_periods  = 1
  threshold           = 0
  comparison_operator = "GreaterThanThreshold"
  treat_missing_data  = "notBreaching"
  alarm_actions       = [var.alarm_sns_topic_arn]

  dimensions = {
    FunctionName = aws_lambda_function.user_sync.function_name
  }
}

resource "aws_cloudwatch_metric_alarm" "user_sync_duration" {
  alarm_name          = "front-user-sync-duration"
  alarm_description   = "front-user-sync approaching its 300s timeout (>90%)"
  namespace           = "AWS/Lambda"
  metric_name         = "Duration"
  statistic           = "Maximum"
  period              = 300
  evaluation_periods  = 1
  threshold           = 270000
  comparison_operator = "GreaterThanThreshold"
  treat_missing_data  = "notBreaching"
  alarm_actions       = [var.alarm_sns_topic_arn]

  dimensions = {
    FunctionName = aws_lambda_function.user_sync.function_name
  }
}

resource "aws_cloudwatch_metric_alarm" "alerts_read_throttle" {
  alarm_name          = "front-sla-alerts-read-throttle"
  alarm_description   = "front-sla-alerts DynamoDB table had one or more read throttle events"
  namespace           = "AWS/DynamoDB"
  metric_name         = "ReadThrottleEvents"
  statistic           = "Sum"
  period              = 300
  evaluation_periods  = 1
  threshold           = 0
  comparison_operator = "GreaterThanThreshold"
  treat_missing_data  = "notBreaching"
  alarm_actions       = [var.alarm_sns_topic_arn]

  dimensions = {
    TableName = aws_dynamodb_table.alerts.name
  }
}

resource "aws_cloudwatch_metric_alarm" "alerts_write_throttle" {
  alarm_name          = "front-sla-alerts-write-throttle"
  alarm_description   = "front-sla-alerts DynamoDB table had one or more write throttle events"
  namespace           = "AWS/DynamoDB"
  metric_name         = "WriteThrottleEvents"
  statistic           = "Sum"
  period              = 300
  evaluation_periods  = 1
  threshold           = 0
  comparison_operator = "GreaterThanThreshold"
  treat_missing_data  = "notBreaching"
  alarm_actions       = [var.alarm_sns_topic_arn]

  dimensions = {
    TableName = aws_dynamodb_table.alerts.name
  }
}