Documentation
¶
Overview ¶
Package metrics provides Prometheus metrics specific to awl-bootstrap-node.
Most of the useful telemetry for a bootstrap node (relay service, autonat, swarm, resource manager, identify, holepunch, ...) comes from libp2p's own built-in Prometheus metrics, enabled via libp2p.PrometheusRegisterer. This package only adds what libp2p does not expose: DHT routing table size, node bandwidth by protocol, node info/uptime, peerstore size, connected peers and DHT requests by client, and relayed connection durations with buckets that fit the relay limits.
Index ¶
- Variables
- func CountDHTRequests(ps peerstore.Peerstore) dht.Option
- func RelayMetricsTracer(reg prometheus.Registerer) relay.MetricsTracer
- func SetBandwidthStats(s BandwidthStats)
- func SetHost(h host.Host)
- func SetNodeInfo(version, peerID string)
- func StartBackgroundUpdater(ctx context.Context, p2pMetrics P2pMetrics)
- type BandwidthStats
- type P2pMetrics
Constants ¶
This section is empty.
Variables ¶
var ( NodeInfo = promauto.NewGaugeVec(prometheus.GaugeOpts{ Namespace: namespace, Subsystem: subsystemNode, Name: "info", Help: "Static node information.", }, []string{"version", "peer_id", "go_version", "os", "arch"}) NodeUptimeSeconds = promauto.NewGauge(prometheus.GaugeOpts{ Namespace: namespace, Subsystem: subsystemNode, Name: "uptime_seconds", Help: "Node uptime in seconds.", }) NodeStartTimestamp = promauto.NewGauge(prometheus.GaugeOpts{ Namespace: namespace, Subsystem: subsystemNode, Name: "start_timestamp", Help: "Unix timestamp of node start.", }) )
var ( // P2PDHTRoutingTableSize is not provided by libp2p Prometheus metrics // (go-libp2p-kad-dht is instrumented via OpenTelemetry and does not export // routing table size), so we expose it ourselves. P2PDHTRoutingTableSize = promauto.NewGauge(prometheus.GaugeOpts{ Namespace: namespace, Subsystem: subsystemP2P, Name: "dht_routing_table_size", Help: "DHT routing table size.", }) P2PBootstrapPeersConnected = promauto.NewGauge(prometheus.GaugeOpts{ Namespace: namespace, Subsystem: subsystemP2P, Name: "bootstrap_peers_connected", Help: "Number of connected bootstrap peers.", }) )
var P2PDHTRequestsTotal = promauto.NewCounterVec(prometheus.CounterOpts{ Namespace: namespace, Subsystem: subsystemP2P, Name: "dht_requests_total", Help: "DHT requests received, by message type and by the client of the peer that sent them.", }, []string{"message_type", labelNameApp, labelNameVersion})
P2PDHTRequestsTotal adds the client of the sender to what libp2p_kad_dht_rpc_inbound_messages_total counts: the version shows how much of the DHT load comes from old clients.
var P2PRelayCircuitDurationSeconds = promauto.NewHistogram(prometheus.HistogramOpts{ Namespace: namespace, Subsystem: subsystemP2P, Name: "relay_circuit_duration_seconds", Help: "How long relayed connections lasted. Attempts that failed to reach the destination are counted too, with a duration under 1 s.", Buckets: []float64{1, 10, 30, 60, 110, 125, 300, 590, 610, 1200, 3600}, })
P2PRelayCircuitDurationSeconds repeats libp2p_relaysvc_connection_duration_seconds with buckets that fit relayed connections: the libp2p one has the default buckets, the largest is 10 s, so it can't show circuits cut by the relay limit. The buckets are around 2 and 10 minutes, the current limit and the next one.
libp2p reports a circuit as closed also when the relay could not reach the destination, so failed attempts land in the first bucket. To get shares of established circuits, divide by libp2p_relaysvc_connection_request_response_status_total{status="ok"}.
Functions ¶
func CountDHTRequests ¶
CountDHTRequests returns a DHT option that counts the received requests in P2PDHTRequestsTotal. The user agent of the sender is taken from ps.
func RelayMetricsTracer ¶
func RelayMetricsTracer(reg prometheus.Registerer) relay.MetricsTracer
RelayMetricsTracer returns the relay metrics tracer of libp2p registered with reg, which also observes P2PRelayCircuitDurationSeconds. Pass it with relay.WithMetricsTracer: it replaces the tracer libp2p sets up itself.
func SetBandwidthStats ¶
func SetBandwidthStats(s BandwidthStats)
SetBandwidthStats sets the source of the traffic metric. The metric is absent until it is set.
func SetHost ¶
SetHost sets the host for the metrics computed on scrape: connected peers and peerstore size. They are absent until it is set.
func SetNodeInfo ¶
func SetNodeInfo(version, peerID string)
SetNodeInfo sets the static node info gauge and the start timestamp.
func StartBackgroundUpdater ¶
func StartBackgroundUpdater(ctx context.Context, p2pMetrics P2pMetrics)
StartBackgroundUpdater periodically updates gauge-type metrics from their data sources.
Types ¶
type BandwidthStats ¶
BandwidthStats is the source of the traffic metric. It is satisfied by *github.com/anywherelan/awl/p2p.P2p.
type P2pMetrics ¶
type P2pMetrics interface {
RoutingTableSize() int
BootstrapPeersStats() (total int, connected int)
}
P2pMetrics is an interface for getting p2p stats used by the background updater. It is satisfied by *github.com/anywherelan/awl/p2p.P2p.