Quickstart • Administrator docs • Development docs •
Observability Operator is a Kubernetes Operator that manages and orchestrates the infrastructure to collect and relay Observability data to several destinations in multi-teant Kubernetes clusters.
It aims to require minimal configuration and reduce the toil of a feature team to get Observability data from their workloads by providing sane defaults and automatically discovering as much information as possible on behalf of the user.
It captures telemetry of tenant workloads (won't instrument the entire cluster), can capture metrics, logs and traces and will relay the telemetry to the tenant systems.
- Metrics: scrape Prometheus format from Pods
- Logs: any format, preferably JSON
- Traces: OpenTelemetry format
- Destination:
- Metrics: any Prometheus RemoteWrite-compatible destination
- Logs: any Fluentd supported format
- Traces: any OpenTelemetry-compatible receiver
Observability Operator relies on the cluster already providing some services. Cluster admins are expected to fulfill these needds in advance:
For traces we use Grafana Alloy but we manage its deployment directly in this operator, so there are no dependenciies to install.
Refer to Please refer to the code contribution guide.
You can find further details of the operator in our development documentation.
This project is released under MIT license. You can find a copy of the license terms in the LICENSE file.