OpenTelemetry traces
Your services keep sending traces the way they do today. Varqeni reads only counts the Collector derives from them, never the traces themselves.
AvailabilityAvailable to everyone with early access to Varqeni.
How it works
- Setup: the Collector runs its span metrics and service graph connectors and exposes them to the Prometheus Varqeni already reads. Each service is matched by its OpenTelemetry service name.
- Error rates: the share of a service's server spans that end in an error, over two minutes, judged like a service that counts its own requests.
- Dependencies: when a service's traces show it calling another, that is a dependency even if nobody wrote it down. A failure that a service only passes on is named once, at the service where it started.
- Fixes: the same as for any service: roll back a bad deploy or restart a hung workload after a named approval.
- In the lab, on two services that send only traces, Varqeni named the right cause and service in 30 of 30 blind trials.
- Logs and metrics: see OpenTelemetry logs and metrics.
What it can and cannot do
Can
- Read request errors per service and the calls between services from your OpenTelemetry Collector, and name a failure once, where it started.
Cannot
- Read the traces themselves.
Related articles
- Read your services under ServiceUnder Service, the console shows each service you use, whether it answers and how fast, how many of its requests fail, which services it calls, your scheduled jobs, and what was deployed in the last 30 minutes.
- OpenTelemetry logs and metricsService Operations names a service that still answers but logs errors, and one whose own metric crosses a line you set, from your OpenTelemetry Collector.
- Infrastructure through PrometheusInfrastructure Operations reads hosts, virtual machines, Kubernetes, storage, data services and certificates from metrics you already collect.
Still stuck? Contact support or email info@fortisdatagroup.com.