Notes / System Design / 15 Complete Case Studies / 01 Telemetry Ingestion Pipeline

3.5 Failure Modes and Mitigations

Failure modes and mitigations across the telemetry ingestion pipeline — gateway crashes, broker failure, processor crashes, span explosion, cardinality rejection, storage saturation, and clock skew.

Appears in: Telemetry Ingestion Pipeline — §3, Deep Dives — this is §3.5.

3.5 Failure Modes and Mitigations

FailureImpactMitigation
Gateway pod crashIn-flight requests lost (small window)Agents retry with backoff; gRPC deadline ensures fast failure; stateless so restart fast
Kafka broker failurePartition leaders re-elected (10–30s)RF=3, unclean.leader.election=false; producers retry; consumer lag grows briefly
Processor crash mid-batchMessages re-consumed from Kafka offsetAt-least-once; downstream TSDB deduplicates
Trace assembler OOM (span explosion)Incomplete traces flushed or droppedHard memory limit + backpressure to Kafka; alert on orphan span rate
TSDB write rejection (cardinality)Metric data loss for offending tenantCardinality check upstream (processor) to reject early; platform alert to tenant
Storage tier fullWrite rejection cascadeBackpressure to Kafka → gateway 429 → agent backoff; auto-scale storage (object store)
Clock skew between agentsOut-of-order samples rejected by TSDBMimir/Prometheus accept samples up to 1h out of order with out-of-order ingestion enabled

Local graph

Full graph →