馃煚 LOW SEV (5 servers) 路 Inference Consumer Down (metrics endpoint unreachable)
馃敟 inference-consumer scrape target is down or has disappeared
<|馃搱 graph> 路 <|馃搳 service dashboard> 路
馃毃 alarm definition 路
馃敃 silence
馃敟 inference-consumer scrape target is down or has disappeared
<|馃搱 graph> 路 <|馃搳 service dashboard> 路
馃毃 alarm definition 路
馃敃 silence
馃敟 inference-consumer scrape target is down or has disappeared
<|馃搱 graph> 路 <|馃搳 service dashboard> 路
馃毃 alarm definition 路
馃敃 silence
馃敟 inference-consumer scrape target is down or has disappeared
<|馃搱 graph> 路 <|馃搳 service dashboard> 路
馃毃 alarm definition 路
馃敃 silence
馃敟 inference-consumer scrape target is down or has disappeared
<|馃搱 graph> 路 <|馃搳 service dashboard> 路
馃毃 alarm definition 路
馃敃 silence
LOW SEV - cause-detector, not harm. A wedged metrics endpoint has ZERO customer impact (screen time and enforcement keep working, we are just blind), and a dead process is normally absorbed by the consumer-group rebalance with screen time backfilling behind it. The harm case - the rebalance did NOT absorb it - is caught as CRITICAL by 'Inference Partition Coverage Lost', including both-boxes-dead via NoData. Alerting critical on a fact that has not hurt anyone is how a channel gets muted. Correlate with 'Inference Consumer Unit Inactive' to tell a wedged endpoint from a dead process.
Grafana vmain-b1e6358