Pushing Panel Data to Statsd

Purpose

Systems like cartography and starbase are “point in time”. This approach makes it easy to query the current state of the data, but sometimes having a view of the data over a period of time is needed to see trends. Seizu can, under some conditions, push query data into metrics systems, using statsd.

How to push metrics

Configure statsd in the backend

See the backend configuration for statsd

Set metric on panels

Panel configurations, including the metric field, are stored in the report store (DynamoDB or SQLModel). Configure panels through the Seizu UI or by seeding from the YAML file with make seed_dashboard. When a report version is saved, the stats worker’s index is updated automatically — no separate step is needed.

The metric setting on a panel is the prefix of the metric. The full name of the metrics being pushed depend on the panel type. For example, the following panel configuration would take the the total value from the count panel and push it to the cves.count.total metric.

        - cypher: cves
          details_cypher: cves-details
          caption: Total CVEs
          metric: cves.count
          type: count
          size: 2.4

The following panel configuration would take the numerator and denominator values from the progress panel and push it to the cves.severity.numerator and cves.severity.denominator metric.

        - cypher: cve-by-severity
          details_cypher: cve-by-severity-details
          caption: Critical CVEs
          metric: cves.severity
          type: progress
          threshold: 0
          size: 2.4

When parameters are used, the parameters would be pushed along with the metric as metric tags. For example, the following panel configuration would take the numerator and denominator values from the progress panel and push it to the cves.severity.numerator and cves.severity.denominator metric, with a metric tag of severity: CRITICAL:

        - cypher: cve-by-severity
          details_cypher: cve-by-severity-details
          params:
            - name: severity
              value: CRITICAL
          caption: Critical CVEs
          type: progress
          threshold: 0
          size: 2.4

If the parameter is an input, rather than an explicit value, the query will be run against every input value, and a metric will be pushed for every value, adding the input value as a metric tag. This can lead to stats with very high cardinality; see the limitation sections about this.

Run the dashboard-stats worker, in a cron

The worker can be run directly:

python -m reporting.dashboard_stats

The worker reads pre-computed stat descriptors directly from the configured report store (DynamoDB or SQLModel), so it needs access to the same store backend as the main seizu service. It does not read the YAML configuration file.

This worker runs and exits, so to push stats on a schedule, the worker should be run as a cron on the preferred schedule.

Limitations

• metric configuration is only support by specific panel types; currently: count, progress

• Panels with inputs can push metrics, but only for panels with a single input, and only if the input has less values than the configuration variable DASHBOARD_STATS_MAX_INPUT_RESULTS allows. This is to avoid sending stats with too much cardinality.

• The stats forwarder (or metrics backend) you have configured must support tags.