gRPC Prometheus metrics not exposed for Python application in Beyla

[myeongseok-rpls]

Description:
I’m seeing that gRPC Prometheus metrics from my Python application are not being exposed by Beyla. I changed trace_printer to text for debugging, but no gRPC-related traces appear.

Configuration:

config:
  data:
    trace_printer: disabled
    profile_port: 6060
    routes:
      unmatched: heuristic
    log_level: error
    discovery:
      open_port: "8080,8000,3000,19530"
      services:
        - k8s_namespace: damoa-(personalize-ml-model-server-v2|webview|mobile-web)
          containers_only: true
    attributes:
      kubernetes:
        enable: true
        resource_labels:
          app: ["app"]
          app-version: ["app-version"]
      extra_group_attributes:
        k8s_app_meta: ["app","app-version"]
      select:
        http_client_request_duration_seconds:
          exclude:
            - "k8s.statefulset.name"
            - "k8s.replicaset.name"
            - "k8s.cluster.name"
            - "k8s.daemonset.name"
            - "k8s.deployment.name"
            - "k8s.job.name"
            - "k8s.owner.name"
            - "k8s.cronjob.name"
            - "k8s.pod.uid"
            - "k8s.pod.start.time"
            - "instance"
            - "job"
            - "k8s_node_name"
        http_server_request_duration_seconds:
          exclude:
            - "k8s.statefulset.name"
            - "k8s.replicaset.name"
            - "k8s.cluster.name"
            - "k8s.daemonset.name"
            - "k8s.deployment.name"
            - "k8s.job.name"
            - "k8s.owner.name"
            - "k8s.cronjob.name"
            - "k8s.pod.uid"
            - "k8s.pod.start.time"
            - "instance"
            - "job"
            - "k8s_node_name"
        http_client_request_body_size_bytes:
          exclude: ["*"]
        http_client_response_body_size_bytes:
          exclude: ["*"]
        http_server_request_body_size_bytes:
          exclude: ["*"]
        http_server_response_body_size_bytes:
          exclude: ["*"]
    filter:
      application:
        server_port:
          not_match: "{0,8126}"
        http_route:
          not_match: "/api/{healthcheck,version}"
    network:
      enable: false
    prometheus_export:
      port: 9090
      path: /metrics
      ttl: 1m
      instrumentations:
        - http
        - grpc

preset: application

Expected Behavior:
gRPC server/client metrics should be visible under the /metrics endpoint on port 9090.

Actual Behavior:
No gRPC metrics are exposed. Even after setting trace_printer to text, I don’t see any gRPC-related traces in the logs.

Steps to Reproduce:

Deploy the Python gRPC service with Beyla as a DaemonSet using the configuration above.

Access http://:9090/metrics.

Observe that only HTTP metrics appear; gRPC metrics are missing.

Notes:

trace_printer: text confirms no gRPC traces are captured.

Other HTTP metrics (client/server) are working as expected.

Can someone help diagnose why the gRPC instrumentation isn’t being picked up?

[grcevski]

Hi @myeongseok-rpls,

Thanks for reporting this issue. Can you please provide us with Beyla logs with the following options as set in the config file:

ebpf:
  bpf_debug: true
  protocol_debug_print: true
log_level: debug

These options will create a lot of logs, so perhaps run them for a short while, but enough to capture some of the gRPC traffic. With this information we should be able to tell if we are seeing any of the gRPC traffic and what might be going wrong. Please send as much logs as you can, the more we have the easier would be to debug the issue.

Alternatively, if you can, maybe you can make us a small example program in Python that mimics what your application is doing to we can try to reproduce?

Either option is fine, whatever is easier for you.

[myeongseok-rpls]

hank you for your feedback. I’ve collected the Beyla logs with the requested settings (eBPF bpf_debug=true, protocol_debug_print=true, and log_level=debug) and attached them here. I hope these logs help identify the root cause.

Thanks again for your support!

Logs-2025-07-11 15_24_05.txt

[grcevski]

Thanks, I believe I found the bug. I'm just doing a bit more verification but I'll proceed with a fix soon.