mirror of
https://github.com/fosrl/newt.git
synced 2026-03-26 20:46:41 +00:00
4.9 KiB
4.9 KiB
Newt Metrics: Recommendations, Gaps, and Roadmap
This document captures the current state of Newt metrics, prioritized fixes, and a pragmatic roadmap for near-term improvements.
-
Current setup (summary)
- Export: Prometheus exposition (default), optional OTLP (gRPC)
- Existing instruments:
- Sites: newt_site_registrations_total, newt_site_online (0/1), newt_site_last_heartbeat_timestamp_seconds
- Tunnel/Traffic: newt_tunnel_sessions, newt_tunnel_bytes_total, newt_tunnel_latency_seconds, newt_tunnel_reconnects_total
- Connection lifecycle: newt_connection_attempts_total, newt_connection_errors_total
- Operations: newt_config_reloads_total, process_start_time_seconds, newt_build_info
- Operations: newt_config_reloads_total, process_start_time_seconds, newt_config_apply_seconds, newt_cert_rotation_total
- Build metadata: newt_build_info
- Control plane: newt_websocket_connect_latency_seconds, newt_websocket_messages_total, newt_websocket_connected, newt_websocket_reconnects_total
- Proxy: newt_proxy_active_connections, newt_proxy_buffer_bytes, newt_proxy_async_backlog_bytes, newt_proxy_drops_total, newt_proxy_accept_total, newt_proxy_connection_duration_seconds, newt_proxy_connections_total
- Go runtime: GC, heap, goroutines via runtime instrumentation
-
Main issues addressed now
- Attribute filter (allow-list) extended to include site_id and region in addition to existing keys (tunnel_id, transport, protocol, direction, result, reason, error_type, version, commit).
- site_id and region propagation: site_id/region remain resource attributes. Metric labels mirror them on per-site gauges and counters by default; set
NEWT_METRICS_INCLUDE_SITE_LABELS=falseto drop them for multi-tenant scrapes. - Label semantics clarified:
- transport: control-plane mechanism (e.g., websocket, wireguard)
- protocol: L4 payload type (tcp, udp)
- newt_tunnel_bytes_total uses protocol and direction, not transport.
- Robustness improvements: removed duplicate clear logic on reconnect; avoided empty site_id by reading NEWT_SITE_ID/NEWT_ID and OTEL_RESOURCE_ATTRIBUTES.
-
Remaining gaps and deviations
- Some call sites still need initiator label on reconnect outcomes (client vs server). This is planned.
- Config apply duration and cert rotation counters are planned.
- Registration and config reload failures are not yet emitted; add failure code paths so result labels expose churn.
- Document using
process_start_time_seconds(andtime()in PromQL) to derive uptime; no explicit restart counter is needed. - Metric helpers often use
context.Background(). Where lightweight contexts exist (e.g., HTTP handlers), propagate them to ease future correlation. - Tracing coverage is limited to admin HTTP and WebSocket connect spans; extend to blueprint fetches, proxy accept loops, and WireGuard updates when OTLP is enabled.
-
Roadmap (phased)
- Phase 1 (done in this iteration)
- Fix attribute filter (site_id, region)
- Propagate site_id (and optional region) across metrics
- Correct label semantics (transport vs protocol); fix sessions transport labelling
- Documentation alignment
- Phase 2 (next)
- Reconnect: add initiator label (client/server)
- Config & PKI: newt_config_apply_seconds{phase,result}; newt_cert_rotation_total{result}
- WebSocket disconnect and keepalive failure counters
- Proxy connection lifecycle metrics (accept totals, duration histogram)
- Pangolin blueprint/config fetch latency and status metrics
- Certificate rotation duration histogram to complement success/failure counter
- Phase 1 (done in this iteration)
-
Operational guidance
- Do not double scrape: scrape either Newt (/metrics) or the Collector’s Prometheus exporter (not both) to avoid double-counting cumulative counters.
- For high cardinality tunnel_id, consider relabeling or dropping per-tunnel series in Prometheus to control cardinality.
- OTLP troubleshooting: enable TLS via OTEL_EXPORTER_OTLP_CERTIFICATE, use OTEL_EXPORTER_OTLP_HEADERS for auth; verify endpoint reachability.
-
Example alerts/recording rules (suggestions)
- Reconnect spikes:
- increase(newt_tunnel_reconnects_total[5m]) by (site_id)
- Sustained connection errors:
- rate(newt_connection_errors_total[5m]) by (site_id,transport,error_type)
- Heartbeat gaps:
- max_over_time(time() - newt_site_last_heartbeat_timestamp_seconds[15m]) by (site_id)
- Proxy drops:
- increase(newt_proxy_drops_total[5m]) by (site_id,protocol)
- WebSocket connect p95 (when added):
- histogram_quantile(0.95, sum(rate(newt_websocket_connect_latency_seconds_bucket[5m])) by (le,site_id))
- Reconnect spikes:
-
Collector configuration
- Direct scrape variant requires no attribute promotion since site_id is already a metric label.
- Transform/promote variant remains optional for environments that rely on resource-to-label promotion.