---
title: "Troubleshoot private probes | Grafana Cloud documentation"
description: "Troubleshoot Synthetic Monitoring private probes by checking agent versions, authentication, API connectivity, logs, and metrics, and find community or account support."
---

> For a curated documentation index, see [llms.txt](/llms.txt). For the complete documentation index, see [llms-full.txt](/llms-full.txt).

# Troubleshoot private probes

If a Synthetic Monitoring private probe isn’t connecting or working as expected, use the following checks to review the agent’s configuration, logs, and metrics. For failures reported by a check, refer to [Troubleshoot a failing check](https://grafana.com/docs/grafana-cloud/observe-and-act/testing/synthetic-monitoring/analyze-results/troubleshoot-checks/).

1. Check the probe version using the `-version` flag and make sure you are running a supported version.
2. Make sure the probe has a valid probe access token.
3. Make sure the probe can connect to the Synthetic Monitoring API server.
4. Run the Synthetic Monitoring Agent in debug mode with the `-debug` and `-verbose` flags and check for errors in the logs.
5. Check the Synthetic Monitoring Agent metrics. The default address for Synthetic Monitoring Agent Prometheus metrics is `localhost:4050/metrics`.
6. If clustered agents don’t form a single cluster:
   
   - Check the agent logs for messages with `subsystem=cluster`. The `cluster peers changed` message lists the current members.
   - `bootstrapping single-node ring` means the agent couldn’t find or join any peer. Check `-cluster-join-addresses`, DNS resolution, and that TCP port `7946` is open between agents.
   - `address did not resolve` means one of the join addresses didn’t resolve to any IP address.
   - Make sure every agent uses the same `-cluster-name` and `-cluster-listen-port`.
   - Make sure the other agents can reach each agent’s advertise address. Set `-cluster-advertise-address` or `-cluster-advertise-interfaces` if needed.
7. If some scripted or browser checks never run in a cluster, make sure every agent has the same feature flags enabled, uses the same image variant, and has the same browser pool configuration. Refer to [Cluster requirements](https://grafana.com/docs/grafana-cloud/observe-and-act/testing/synthetic-monitoring/set-up/set-up-private-probes/clustering/#cluster-requirements).
8. If browser checks fail with a browser pool:
   
   - Check the agent logs for messages with `subsystem=browser_pool`.
   - If `sm_agent_browser_pool_acquires_total{result="exhausted"}` increases, no browser pool instance was free. Add instances.
   - If `sm_agent_browser_pool_instances` is `0`, the agent didn’t find any instance. Check `-browser-pool-addresses` and DNS resolution, and make sure the `crocochrome` Service is headless.
   - If `sm_agent_browser_pool_probes_total{result="error"}` increases, the agent can’t reach the instances or they don’t support session acquisition. Check that TCP port `8080` is open from the agents and that the instances run `crocochrome` `v0.13.0` or later.

If you are still facing issues with Private Probes, ask for help in the [community forum](https://community.grafana.com/). Refer to [Account types and support available for Grafana Cloud users](https://grafana.com/docs/grafana-cloud/account-management/support/#account-types-and-support-available-for-grafana-cloud-users) for more information.
