====== OS Agent - Statistics ======

The **Statistics** tab shows counters about the OS Agent push activity, useful for diagnosing problems on the receiving side.

===== Counters =====

| Counter | Meaning |
| **Metrics Received** | Total number of metric points the Collector has accepted and processed since the last restart|
| **Push Requests** | Total number of HTTP push requests received. |
| **Auth Failures** | Number of pushes rejected because of an invalid or missing API key. A non-zero value means an agent is misconfigured **or** someone is probing the endpoint. |
| **Unmatched Hosts** | Number of pushes from a host that could not be linked to any System. This is normal for standalone hosts - it is informational, not an error. |
| **Parse Errors** | Number of pushes the Collector could not parse - typically malformed Influx Line Protocol or read errors. Should stay near zero. |

The counters are highlighted in red when they are above zero - this draws the eye to errors when the page is open.

==== Reading the counters ====

  * **Push Requests** rising and **Metrics Received** rising → everything works
  * **Push Requests** rising but **Metrics Received** stuck at zero → the body is parsed but discarded - usually a wrong ''host'' tag (rejected at host registration), or //Active// flag is OFF
  * **Auth Failures** rising fast → an agent uses an old key, or someone is brute-forcing
  * **Parse Errors** rising → an agent sends a wrong format - check the Collector log for the exact line

===== Refresh =====

The counters do not auto-refresh. Click **Refresh** at the top of the tab to reload them.

===== Clear =====

The **Clear** button resets all counters to zero. Useful before a test:

  - Click Clear
  - Run your test (push a metric, restart Telegraf, etc.)
  - Click Refresh
  - Read the new values - they reflect only the test activity

The reset only affects the counters in memory - it does not delete metrics from the pipeline or remove hosts.

===== Server-side log =====

The Collector also logs a few key events:

  * Each successful auto-discovery: ''Auto-discovered OS host: <hostname> (system=<id>)'' or ''(standalone)''
  * Auth failures: ''OS Agent: rejected push from host=<...> - <reason>''
  * Host cap reached: ''OS Agent: host cap reached (10000), rejecting auto-discovery of <hostname>''
  * Parse errors: ''Failed to parse influx body: <error>''

Look for these messages in the Collector log when the counters do not match what you expect.

===== Combining with Telegraf logs =====

When a problem appears, check both sides:

  * **Telegraf** (sender) - ''journalctl -u telegraf -f'' on Linux or Event Viewer → //Application// on Windows. Useful messages: ''Error writing to outputs.http'' (network or auth issue), ''connection refused'' (Collector down or wrong URL), ''received status code: 401'' (wrong API key), ''received status code: 403'' (host deactivated).
  * **Collector** (receiver) - log file under ''logs/'' or wherever your Collector logs are. Look for ''OS Agent'' lines.

Mismatched counts (Telegraf says it pushed N, Statistics tab says fewer were received) are a sign of a network drop or a Collector restart - the //Push Requests// counter resets when the Collector restarts.