Platform Admin Service — Observability

Status: populated Owner: TBD Last updated: 2026-04-18 Companion: 12 Observability

1. SLIs and SLOs

SLI	SLO	Window
Internal flag evaluate availability	≥ 99.9%	30-day rolling
Internal flag evaluate p95 latency	≤ 120 ms	24 h
Aggregate health endpoint availability	≥ 99.5%	30-day
Aggregate health p95 response time	≤ 2 s	24 h
Config write p99 latency	≤ 300 ms	24 h

2. Key metrics

Metric	Type	Labels	Description
pltadm_flag_evaluate_duration_seconds	Histogram	key, decision, reason	Flag evaluation latency
pltadm_flag_cache_hits_total	Counter	key	Redis cache hits for flag evaluation
pltadm_config_mutations_total	Counter	key, scope	Config upsert/archive events
pltadm_health_aggregate_status	Gauge	overall	0=healthy, 1=degraded, 2=unhealthy
pltadm_health_source_status	Gauge	service_id	Per-service health status
pltadm_health_poll_duration_seconds	Histogram	service_id	Health probe duration
pltadm_outbox_unpublished_age_seconds	Gauge	—	Age of oldest unpublished event

3. Alerts

Alert	Condition	Severity
FlagEvaluateLatencyHigh	p95 > 150 ms for 5 min	High
FlagCacheHitRateLow	Cache hit rate < 50% for 5 min	Medium
PlatformHealthUnhealthy	pltadm_health_aggregate_status = 2 for > 2 min	Critical
HealthSourceStale	Any source not polled within 2× staleness threshold	High
OutboxUnpublishedOld	Age > 60 s	High

4. Dashboards

Dashboard	Description
Feature Flag Operations	Evaluate latency; cache hit ratio; flag mutation frequency
Platform Health	Aggregate status heatmap; per-service status over time
Config Governance	Config mutation frequency; history growth
Outbox Health	Unpublished event age; publish failure rate

5. Traces

Span	Description
pltadm.flag.evaluate	Full flag evaluation including cache check
pltadm.health.aggregate	Health aggregation query
pltadm.health.poll	Per-service health probe
pltadm.config.upsert	Config write + history row
pltadm.outbox.publish	Event publish via outbox relay