forgejo/terraform/alarms.tf

resource "aws_cloudwatch_metric_alarm" "verification_errors" {
  count = var.enable_schedules ? 1 : 0

  alarm_name          = "forgejo-backup-verification-errors"
  alarm_description   = "Backup verification Lambda is failing. Slack notifications may not be firing."
  comparison_operator = "GreaterThanOrEqualToThreshold"
  evaluation_periods  = 1
  metric_name         = "Errors"
  namespace           = "AWS/Lambda"
  period              = 3600
  statistic           = "Sum"
  threshold           = 1
  treat_missing_data  = "notBreaching"

  dimensions = {
    FunctionName = aws_lambda_function.verification.function_name
  }

  alarm_actions = [local.site_alerts_arn]
}

resource "aws_cloudwatch_metric_alarm" "verification_not_running" {
  count = var.enable_schedules ? 1 : 0

  alarm_name          = "forgejo-backup-verification-not-running"
  alarm_description   = "Backup verification Lambda has not run in the last 24h."
  comparison_operator = "LessThanThreshold"
  evaluation_periods  = 1
  metric_name         = "Invocations"
  namespace           = "AWS/Lambda"
  period              = 86400
  statistic           = "Sum"
  threshold           = 1
  treat_missing_data  = "breaching"

  dimensions = {
    FunctionName = aws_lambda_function.verification.function_name
  }

  alarm_actions = [local.site_alerts_arn]
}