inlined-refs — consumer lag dashboarding

Step 1 — pick the exporter

We support two exporters. Read the comparison below before recommending.

Exporter comparison (from references/exporters.md)

kafka-exporter (Danielqsj)

Pros: lightweight, single binary, exposes per-group lag directly. Cons: stateful (consumes __consumer_offsets internally), can lag on large clusters.

Best for: clusters < 100 brokers, < 1000 consumer groups.

Config snippet:

kafka-exporter \
  --kafka.server=localhost:9092 \
  --kafka.version=3.5.0 \
  --web.listen-address=:9308

Confluent Control Center

Pros: official, covers full Confluent Platform metrics. Cons: paid licensing, heavier footprint.

Best for: enterprise Confluent Platform users.

Step 2 — Grafana dashboard

Use Grafana dashboard 7589 for kafka-exporter. Filter by consumer group, plot kafka_consumergroup_lag over a 5-minute window.

Step 3 — alerts

Set lag > 10,000 messages for 5 minutes as the default alert. Adjust per topic SLA.