Odigos Own Metrics

Overview

Odigos collects own metrics about its pipeline — collector throughput, export rates, eBPF instrumentation counters, and other internal telemetry. These metrics describe Odigos itself, not your application workloads. By default, the bundled metrics store is disabled (ownTelemetry.metricsStore.disabled: true). When you enable it, Odigos deploys a VictoriaMetrics instance (odigos-victoriametrics) in the Odigos namespace and routes own metrics there so the UI can query them.

What enabling the metrics store does

When ownTelemetry.metricsStore.disabled is set to false:

VictoriaMetrics is deployed — a single-replica odigos-victoriametrics Deployment with 1-day retention stores own metrics inside the cluster.
Collectors report own telemetry — node collectors and the cluster gateway scrape their OpenTelemetry internal /metrics endpoints and forward the data through the gateway to VictoriaMetrics.
Odiglet instrumentation metrics are collected — node collectors scrape eBPF instrumentation counters from odiglet and forward them through the gateway to the store.
Odigos UI representation — metrics are display under “collectors pipeline”.

Separately from the metrics store, collectors always export traffic-size metrics directly to the UI server over OTLP. That path powers per-source and per-destination throughput (throughputBytes, totalDataSent) in the UI without requiring the metrics store.

Enabling

The metrics store is disabled by default to keep the Odigos footprint minimal on small clusters.

Helm yaml
Helm command

ownTelemetry:
  metricsStore:
    disabled: false

helm upgrade --install odigos odigos/odigos \
  --namespace odigos-system \
  --reuse-values \
  --set ownTelemetry.metricsStore.disabled=false

Configuration options

Export to external metrics destinations

You can also send own metrics to your existing metrics backends by enabling Collect Odigos own metrics on a metrics destination. When either the metrics store or at least one such destination is enabled, collectors start the own-metrics collection pipeline. See Odigos own metrics settings in the API reference for the full OdigosOwnMetricsSettings schema.

Resource footprint

The bundled VictoriaMetrics instance requests roughly:

Resource	Request	Limit
CPU	100m	500m
Memory	512Mi	2Gi

Metrics are retained for 1 day on ephemeral storage (emptyDir).

Using own metrics for sizing

Own metrics help you understand how much telemetry Odigos is processing before you change ResourceSizePreset or collector resources.

Enable the metrics store.
Instrument representative workloads and let traffic run for a few minutes.
Check source throughput and destination throughput in the Odigos UI.
Compare total span throughput (bytes/sec) with the Benchmarks preset tables to pick size_s through size_xl.

Getting Help

If you have any issues, or require our assistance, please open an issue in GitHub, or reach out to us in Odigos Slack

​Overview

​What enabling the metrics store does

​Enabling

​Configuration options

​Export to external metrics destinations

​Resource footprint

​Using own metrics for sizing

​Getting Help