Hey everyone. I’m trying to get tracing configured...
# ory-selfhosting
a
Hey everyone. I’m trying to get tracing configured with Kratos. I’m running Kratos in Kubernetes, with an opentelemetry agent sidecar for each pod. The agent is running otlp, jaeger, and zipkin receivers. My kratos tracing config looks like:
Copy code
tracing:
    service_name: "kratos"
    providers:
        jaeger: 
            propagation: w3c
            sampling:
                type: const
                value: 1
                server_url: <http://localhost:14250/sampling>
            local_agent_address: localhost:6831
    provider: jaeger
So the problems I’m seeing are: 1. The
Traceparent
header received from the request isn’t being propagated, so a new trace is started when it reaches Kratos. (Logs show that this header is present). 2. Kratos itself isn’t propagating the trace ID’s correctly. What should be multiple spans within a single trace are actually creating multiple traces with a single span in each. Has anyone run into this problem, know how to solve it, or have a working configuration?
Some examples of what I’m seeing. The
migration-status
trace works how I’d expect, with all the database transactions included in the same trace. However, for
/sessions/whoami
, an new trace is created for the top level request and for each resulting database transaction. And it doesn’t pick up the existing trace that was started by the proxy before it.
m
Hello @User, Apologies this issue fell off the radar, sometimes Slack is moving to fast to catch everything. I would suggest you open an discussion with your issue, that way we can track it easier and find a solution, Thanks! https://github.com/ory/kratos/discussions/new