Application Real-Time Monitoring Service:Collect Go pprof data with arms-obi

Code integration

In the main package of your Go service, import net/http/pprof and start a separate HTTP server:

import _ "net/http/pprof"

func main() {
    // Start the pprof server on a separate port to avoid conflicts with the business port.
    go func() {
        http.ListenAndServe(":6060", nil)
    }()
    // ... business logic
}

Container environment variables

In your application Pod's container spec, set the pprof port environment variable. The variable name defaults to ILOGTAIL_PROFILE_PORT and is configurable in the values:

containers:
  - name: myapp
    env:
      - name: ILOGTAIL_PROFILE_PORT  # Must match the port_env_name configured in arms-obi
        value: "6060"

arms-obi reads this environment variable to determine the target port.

Collector configuration

Profile collection URLs follow this pattern: http://<podIP>:<port><pprof_path>/<profile>

Filter configuration

Controls which Pods to collect from. All conditions are combined with AND; items within a whitelist are combined with OR.

LabelKV format: {key: "app", value: "myservice"}. If the value is empty, the agent only checks for the key's existence.

The filter pipeline runs in this order:

1. Node filtering: Collects only from Pods on the current node (automatic).

2. Pod Phase: Keeps only Pods in the Running phase with an assigned IP address.

3. Regular expression filtering: namespace / pod_name.

4. Label whitelist → Label blacklist.

5. Environment variable whitelist → Environment variable blacklist.

6. Port resolution: Reads the port value from the container environment variable specified by port_env_name.

ARMS configuration

Update the configuration

1. Log on to the Container Service for Kubernetes (ACK) console. In the left-side navigation pane, select Cluster List, and then click the name of the target cluster.

2. In the left-side navigation pane, select Components, find the arms-obi component, click Configure, and then modify the profiling collector configuration.

Agent logs show "arms sts not ready yet"

The agent failed to obtain an ARMS STS credential.

• Ensure that the cluster network can access the ARMS metadata API at <region_id>-intranet-arms.aliyuncs.com.

• You can run kubectl logs to view detailed error messages.

No Pods are discovered

Check your filter configuration:

• Confirm that the environment variable specified by port_env_name (default: ILOGTAIL_PROFILE_PORT) is set in your application's Pod containers.

• Confirm that the regular expressions for namespace/pod_name, as well as the label and environment variable whitelist and blacklist settings in the selector, are correct.

• Check the pod_count field in the agent logs to see how many Pods were discovered.

Collection fails

Common causes:

• The application Pod does not expose the pprof interface. Confirm that you have imported _ "net/http/pprof" and started the HTTP service.

• The pprof port does not match the port configured in the port_env_name environment variable.

• The pprof_path does not match the path prefix of your service.

• A network policy is blocking the HTTP connection from the agent Pod to the application Pod.

Upload fails

Common causes:

• The VPC private endpoint is unreachable.

• The ARMS STS credential has expired. The agent refreshes the credential automatically. Check the logs for STS refresh errors.