blumeops/fly/alloy.river

// Grafana Alloy configuration for flyio-proxy
// Collects nginx access logs → Loki, extracts metrics → Prometheus.
// Note: stub_status connection metrics are not collected — Alloy has no
// built-in nginx exporter. The log-derived metrics cover the key signals.

// ============== LOG COLLECTION ==============

// Tail the JSON access log written by nginx
local.file_match "nginx_access" {
  path_targets = [
    {__path__ = "/var/log/nginx/access.json.log", job = "flyio-nginx"},
  ]
}

loki.source.file "nginx_access" {
  targets    = local.file_match.nginx_access.targets
  forward_to = [loki.process.nginx.receiver]
}

// Parse JSON fields, extract labels, derive metrics
loki.process "nginx" {
  forward_to = [loki.relabel.instance.receiver]

  // Parse the JSON log line
  stage.json {
    expressions = {
      client_ip            = "client_ip",
      status               = "status",
      method               = "request_method",
      host                 = "http_host",
      cache_status         = "upstream_cache_status",
      request_time         = "request_time",
      body_bytes_sent      = "body_bytes_sent",
      upstream_response_time = "upstream_response_time",
    }
  }

  // Promote to labels for filtering in Loki
  stage.labels {
    values = {
      status       = "",
      method       = "",
      host         = "",
      cache_status = "",
    }
  }

  // --- Derived metrics (exposed on Alloy's /metrics endpoint) ---

  stage.metrics {
    metric.counter {
      name        = "flyio_nginx_http_requests_total"
      description = "Total HTTP requests by status, method, and host."
      match_all   = true
      action      = "inc"
    }
  }

  stage.metrics {
    metric.histogram {
      name        = "flyio_nginx_http_request_duration_seconds"
      description = "HTTP request latency in seconds."
      source      = "request_time"
      buckets     = [0.005, 0.01, 0.025, 0.05, 0.1, 0.25, 0.5, 1, 2.5, 5, 10, 15, 20, 30, 45, 60]
    }
  }

  stage.metrics {
    metric.histogram {
      name        = "flyio_nginx_upstream_response_time_seconds"
      description = "Upstream (Forgejo) response time in seconds, excluding proxy overhead."
      source      = "upstream_response_time"
      buckets     = [0.005, 0.01, 0.025, 0.05, 0.1, 0.25, 0.5, 1, 2.5, 5, 10, 15, 20, 30, 45, 60]
    }
  }

  stage.metrics {
    metric.counter {
      name        = "flyio_nginx_http_response_bytes_total"
      description = "Total bytes sent in HTTP responses."
      source      = "body_bytes_sent"
      action      = "add"
    }
  }

  stage.metrics {
    metric.counter {
      name        = "flyio_nginx_cache_requests_total"
      description = "Total cache lookups by cache status."
      source      = "cache_status"
      action      = "inc"
    }
  }
}

// Add instance label to logs
loki.relabel "instance" {
  forward_to = [loki.write.loki.receiver]

  rule {
    target_label = "instance"
    replacement  = "flyio-proxy"
  }
}

// Write logs to Loki via Tailscale Ingress (direct, bypasses Caddy)
// Uses direct Tailscale endpoint because flyio-proxy ACLs only allow
// tag:flyio-target — Caddy on indri (tag:homelab) is not reachable.
loki.write "loki" {
  endpoint {
    url = "https://loki.tail8d86e.ts.net/loki/api/v1/push"
  }
}

// ============== METRICS PIPELINE ==============

// Self-scrape to collect the log-derived metrics from /metrics
prometheus.scrape "self" {
  targets         = [{"__address__" = "127.0.0.1:12345"}]
  forward_to      = [prometheus.relabel.instance.receiver]
  scrape_interval = "15s"
}

// Strip the "loki_process_custom_" prefix that Alloy adds to stage.metrics,
// then add instance label. This keeps dashboard queries clean.
prometheus.relabel "instance" {
  forward_to = [prometheus.remote_write.prometheus.receiver]

  rule {
    source_labels = ["__name__"]
    regex         = "loki_process_custom_(.*)"
    target_label  = "__name__"
    replacement   = "$1"
  }

  // Drop internal labels added by the loki pipeline
  rule {
    regex  = "component_id|component_path|filename"
    action = "labeldrop"
  }

  rule {
    target_label = "instance"
    replacement  = "flyio-proxy"
  }
}

// Push metrics to Prometheus via Tailscale Ingress (direct, bypasses Caddy)
// Uses direct Tailscale endpoint because flyio-proxy ACLs only allow
// tag:flyio-target — Caddy on indri (tag:homelab) is not reachable.
prometheus.remote_write "prometheus" {
  endpoint {
    url = "https://prometheus.tail8d86e.ts.net/api/v1/write"
  }
}