Ubuntu Manpages

fapolicyd.metrics

fapolicyd runtime metrics report file

The /run/fapolicyd/fapolicyd.metrics file contains the most recent daemon metrics report requested through fapolicyd-cli --check-metrics, fapolicyd-cli --reset-metrics, or periodic interval reporting.

The metrics report answers where runtime hot paths and cache effects are. The state report, written to /run/fapolicyd/fapolicyd.state, answers whether the daemon is healthy and configured as expected. Each field is printed as a name: value line. Section headers end in a colon. Worker-specific drill-down fields are indented by two spaces under the related aggregate section.

The wall-clock time of the last successful metrics reset, or never when metrics have not been reset since daemon start. Reset reports show the counter window that is about to be reset; later metrics reports show the new reset time.
The active decision configuration generation at report time. The counters may span earlier generations when configuration was reloaded after the last metrics reset.
The active ruleset generation that the counters and rule hit table apply to.
The active trust database generation used by new decision reads at report time. This is a daemon-runtime generation that starts at 1 during startup. The counters may span earlier trust database generations when reloads occurred after the last metrics reset.
The active physical LMDB environment generation at report time. This increments only when fapolicyd closes the live environment and publishes a validated replacement environment for compaction or high-water reset. It complements the trust database generation: trust generations publish logical content inside an environment, while LMDB environment generations identify the storage file currently backing those logical generations.
The number of records in the active trust database generation at report time.

The number of policy decisions that allowed access.
The number of policy decisions that denied access.
The number of allow decisions produced by a matching rule.
The number of allow decisions produced when no rule had an opinion and the daemon used the default allow behavior. Detailed fallthrough dimensions are printed when this value is non-zero.

The number of fanotify permission events currently waiting in internal worker queues.
The highest per-worker internal event queue depth observed since the last metrics reset.
The number of failed enqueue attempts across all worker queues since the last metrics reset.
The oldest currently queued event age across all worker queues.
The number of fanotify permission events currently waiting in that worker's queue.
The highest queue depth observed for that worker since the last metrics reset.
The number of failed enqueue attempts for that worker since the last metrics reset.
The age of the oldest event currently waiting in that worker's queue. The value is formatted with a human-readable unit such as ms or s.
The cumulative number of fanotify permission events deferred since the last metrics reset because another process was still building subject pattern state in the same cache slot.
The highest number of concurrently deferred subject events observed since the last metrics reset.
The cumulative number of defer-array-full fallbacks since the last metrics reset. This field also appears in the state report because non-zero values are health indicators.
The number of subject-slot deferrals owned by one decision worker since the last metrics reset.
The highest number of concurrently deferred subject events observed in one decision worker since the last metrics reset.
The number of defer-array-full fallbacks in one decision worker since the last metrics reset.
The number of subject events currently parked in one decision worker's defer array.
The age of the oldest currently deferred event in one decision worker's defer array.

The number of subject cache hits.
The number of subject cache misses.
The number of populated subject cache slots whose full process identity did not match the current event and therefore had to be evicted before reuse.
The number of subject cache evictions and the eviction percentage relative to subject cache hits.
The number of subject cache entries evicted before process startup state was complete. This field also appears in the state report because non-zero values are health indicators.
The number of BUILDING subject cache entries evicted because the owning process was traced and could hold the slot indefinitely. This field also appears in the state report because non-zero values are health indicators.
The number of BUILDING subject cache entries evicted because their startup state stayed incomplete past the bounded stale window. This field also appears in the state report because non-zero values are health indicators.
The number of subject cache hits in one decision worker.
The number of subject cache misses in one decision worker.
The number of subject cache collisions in one decision worker.
The number of subject cache evictions in one decision worker.
The number of early subject cache evictions in one decision worker.
The number of traced BUILDING subject evictions in one decision worker.
The number of stale BUILDING subject evictions in one decision worker.

The number of object cache hits.
The number of object cache misses.
The number of populated object cache slots whose full file identity did not match the current event and therefore had to be evicted before reuse.
The number of object cache evictions and the eviction percentage relative to object cache hits.
The number of object cache hits in one decision worker.
The number of object cache misses in one decision worker.
The number of object cache collisions in one decision worker.
The number of object cache evictions in one decision worker.

The number of trust database lookup attempts since the last metrics reset.
The number of lookup attempts that could not open an LMDB read transaction because all LMDB reader slots were in use.

Trust database lookup latency is reported by fapolicyd.timing(5) only while manual timing collection is armed, so normal metrics collection does not add per-lookup clock reads.

One line per configured rule in rule order. The line includes the one-based rule number, hit count, and rule text. Rule hit counters are naturally scoped to the active ruleset generation and start at zero when a new ruleset is loaded. A metrics reset also clears the counters after reporting them, which allows focused tests against the currently loaded rules without forcing a rule reload.

One line is printed for each counted subject attribute in the form name requests=count lookups=count. The requests count increments when policy evaluation or syslog formatting asks for the attribute. The lookups count increments only when that attribute was absent from the event subject cache and fapolicyd had to compute or fetch it.

One line is printed for each counted object attribute in the form name requests=count lookups=count. The requests count increments when policy evaluation or syslog formatting asks for the attribute. The lookups count increments only when that attribute was absent from the event object cache and fapolicyd had to compute or fetch it.

Metrics resets affect counters in this report. Static configuration, current utilization snapshots, watched mounts, and health indicators are reported in fapolicyd.state(5) and are not reset by fapolicyd-cli --reset-metrics.

/run/fapolicyd/fapolicyd.metrics
Runtime metrics report file.
/run/fapolicyd/fapolicyd.state
Runtime state report file.
/run/fapolicyd/fapolicyd.timing
Manual decision timing report written when a privileged stop request ends an armed timing run, when a stop request finds timing unarmed, or when timing collection stops to avoid counter overflow.

fapolicyd(8), fapolicyd-cli(8), fapolicyd.conf(5), fapolicyd.rules(5), fapolicyd.state(5), fapolicyd.timing(5), and fapolicyd.trust(5).