> ## Documentation Index
> Fetch the complete documentation index at: https://bloodhound.specterops.io/llms.txt
> Use this file to discover all available pages before exploring further.

# Performance Telemetry

> Configure and interpret OpenHound Okta collection performance telemetry.

![Applies to BloodHound Enterprise and CE](https://mintcdn.com/specterops/tTIczgde9H07oLXf/assets/enterprise-AND-community-edition-pill-tag.svg?fit=max\&auto=format\&n=tTIczgde9H07oLXf\&q=85\&s=ad49a576589f4d2a8081df77d07fdf56)

Use optional collection performance telemetry to diagnose and tune OpenHound Okta collection performance.

Telemetry is disabled by default and does not change the collection scope, schedule, Okta permissions, or number of Okta requests. When enabled, the collector writes bounded, value-free JSONL diagnostics for endpoint latency, HTTP outcomes, retries, quota observations, concurrency, pacing, pagination, and row counts.

## Configure Telemetry

Add the following section to `config.toml`:

```toml title="config.toml" theme={null}
[sources.source.okta.telemetry]
enabled = true
output_directory = "./telemetry"
reporting_interval_seconds = 60
max_file_bytes = 10485760
max_interval_records = 1440
queue_capacity = 16
```

Set `output_directory` outside the raw collection output directory. The collector rejects a nested location so the telemetry file is not treated as collection data or uploaded with it.

Run the Okta collection normally. The same configuration applies to interactive, scheduled, and containerized collections. Each collection writes one file named `openhound-okta-<opaque-run-id>.jsonl` in the configured output directory.

## Review Telemetry Records

Each JSONL line is an independently parseable record. The file starts with a `run_start` record, followed by periodic `interval` records, and normally ends with a `run_summary` record. Interrupted or incomplete collections retain records that were flushed before the interruption.

## Telemetry Settings

You can set the following values in `config.toml` or with the equivalent environment variable. Environment variables take precedence over `config.toml`.

Replace \{PREFIX} with `SOURCES__SOURCE__OKTA__TELEMETRY` in the environment variables below.

<table>
  <colgroup>
    <col />

    <col />

    <col />

    <col />
  </colgroup>

  <thead>
    <tr>
      <th>Parameter Name</th>
      <th>Environment Variable</th>
      <th>Default</th>
      <th>Description</th>
    </tr>
  </thead>

  <tbody>
    <tr>
      <td>`enabled`</td>

      <td>
        `{PREFIX}__ENABLED`
      </td>

      <td>`false`</td>
      <td>Enables telemetry. No telemetry file is created when this is `false` or omitted.</td>
    </tr>

    <tr>
      <td>`output_directory`</td>

      <td>
        `{PREFIX}__OUTPUT_DIRECTORY`
      </td>

      <td>`./telemetry`</td>
      <td>Directory for the JSONL file. It must be outside the raw collection output directory.</td>
    </tr>

    <tr>
      <td>`reporting_interval_seconds`</td>

      <td>
        `{PREFIX}__REPORTING_INTERVAL_SECONDS`
      </td>

      <td>`60`</td>
      <td>Interval at which the collector flushes aggregated diagnostic records. Must be greater than zero.</td>
    </tr>

    <tr>
      <td>`max_file_bytes`</td>

      <td>
        `{PREFIX}__MAX_FILE_BYTES`
      </td>

      <td>`10485760`</td>
      <td>Maximum telemetry file size. Must be at least `65536` bytes.</td>
    </tr>

    <tr>
      <td>`max_interval_records`</td>

      <td>
        `{PREFIX}__MAX_INTERVAL_RECORDS`
      </td>

      <td>`1440`</td>
      <td>Maximum number of interval records. Must be at least `1`.</td>
    </tr>

    <tr>
      <td>`queue_capacity`</td>

      <td>
        `{PREFIX}__QUEUE_CAPACITY`
      </td>

      <td>`16`</td>
      <td>Maximum number of records waiting for export. Must be at least `2`.</td>
    </tr>
  </tbody>
</table>

The collector bounds telemetry memory, queue depth, interval records, and file size. If the telemetry destination is unavailable, the exporter fails, or a limit is reached, the collection continues. The diagnostic records indicate dropped records, truncation, or exporter failure; an incomplete telemetry summary does not mean that the Okta collection failed.

## Telemetry Metrics

Telemetry separates the following measurements so you can identify where collection time is spent:

<table>
  <colgroup>
    <col />

    <col />
  </colgroup>

  <thead>
    <tr>
      <th>Measurement</th>
      <th>Description</th>
    </tr>
  </thead>

  <tbody>
    <tr>
      <td>`http_execution`</td>
      <td>Measures request execution through response-body transfer. It excludes JSON decoding and downstream DLT row conversion.</td>
    </tr>

    <tr>
      <td>`slot_wait`</td>
      <td>Measures time waiting for the Okta endpoint-family concurrency slot and the organization-wide `org_concurrency` gate.</td>
    </tr>

    <tr>
      <td>`proactive_pacing_wait`</td>
      <td>Measures waits caused by pacing against observed quota headers.</td>
    </tr>

    <tr>
      <td>`retry_backoff_wait`</td>
      <td>Measures the actual wait before a retry. Retry delay totals are also recorded by retry category.</td>
    </tr>
  </tbody>
</table>

The summary records the effective DLT extraction settings (`extract_workers` and `extract_max_parallel_items`) and the effective Okta performance settings, including page sizes, endpoint concurrency, retry window, and rate-limit reserve.

Page and row counts describe data yielded by Okta pagination; they do not confirm that DLT committed the rows to the raw collection.

> [!WARNING]
>
> Telemetry uses normalized endpoint templates and excludes credentials, tenant hostnames, raw URLs, cursors, object identifiers, response bodies, profiles, and exception text.
>
> Store the JSONL files as diagnostic artifacts and do not place them under the raw collection directory.
