| name | logging-noise-reduction |
| description | Identify noisy log categories and apply level overrides / burst suppression per the logging spec, plus container log rotation. Use when logs are too verbose or a category floods the output. |
| trigger | reduce log noise, noisy logger, burst suppression, log level override, logging profile, Serilog config, log rotation, quiet logs |
Logging Noise Reduction
Goal
Quiet a noisy log category by adjusting its level and/or applying burst
suppression per the logging spec, and ensure container logs rotate so they don't
fill disk — without losing the signal needed for triage.
Inputs
- The noisy logger category and the desired level / suppression threshold.
- Key references:
- Logging spec:
docs/Spec/11_LOGGING.md (profiles Quiet/Normal/Debug/Trace,
burst-suppression rules, per-container defaults, required structured fields).
- Service logging config:
Services/<service>/Program.cs (Serilog host config)
and Config/LoggingProfile.json.
- Container log driver:
docker-compose*.yml (logging max-size / max-file).
- Area rules:
.github/instructions/services.instructions.md (+
.github/instructions/infrastructure.instructions.md for compose).
Note: Spec/11 references Exports/Telemetry/{StructuredLogging,BurstSuppressor}.cs
which may not exist on this branch yet — apply level overrides via the existing
Serilog config until the suppression helper lands; treat Spec/11 as authoritative.
Preconditions
- You have identified the exact category string (e.g.
"WoWStateManager.Movement") from the actual log output.
- The owning service builds green.
Procedure
- Confirm the noisy category from real logs (use Codex to scan large log files
rather than reading them inline).
- In the service's logging config /
Config/LoggingProfile.json, set a category
override (e.g. "WoWStateManager.Movement": "Warning") per the Spec/11 profile.
- For bursty categories, add a suppression entry with a
MaxPerMinute threshold
and emit a [SUPPRESSED] summary metric (wwow.log.suppressed_total{category})
so suppression is observable (see [[metrics-instrumentation]]).
- Set container log rotation in compose:
logging.driver: local, max-size,
max-file per the Spec/11 per-container table.
- Re-run and confirm the signal you need is still present at the chosen level.
Verification
- Build:
.\scripts\build.ps1; fast suite: .\scripts\test-fast.ps1.
- Run the service and confirm the category volume dropped and important events are
still logged.
- Confirm container logs rotate at the configured size.
Outputs
- Updated logging config /
Config/LoggingProfile.json + (optional) compose log
rotation.
[SUPPRESSED] metric if burst suppression was added.
Failure modes and recovery
- Over-suppressing hides real errors — never drop Warning/Error for a category
that can fail meaningfully.
- No rotation → disk fills from container logs.
- Suppressing silently — always emit the suppression summary so the loss is
visible.
Related skills
- [[metrics-instrumentation]] — the suppression-count metric.
- [[config-hot-reload-subscriber]] — make logging levels live-reloadable.
- [[docker-stack-extension]] — set rotation when adding a container.
- Reference:
docs/Spec/11_LOGGING.md.
