Quick Start
This guide connects Pilot to an existing Kafka cluster and walks through the key features.
1. Start Pilot
Point Pilot at your Kafka brokers:
docker run -d \
--name pilot \
-p 8080:8080 \
-e KAFKA_BOOTSTRAP_SERVERS=broker1:9092,broker2:9092,broker3:9092 \
calinora/pilot:latestFor SASL or TLS-secured clusters, see Configuration.
2. Open the Dashboard
Navigate to http://localhost:8080 in your browser.
The overview page shows:
- Cluster health - broker count, under-replicated partitions, offline partitions
- Broker status - each broker with its rack, partition count, and leader count
- Balance status - whether Pilot recommends a rebalance, and which broker is farthest from its fair share
3. Explore Your Cluster
Brokers
View individual broker details including partition assignments, rack placement, and disk usage.
Topics
Browse all topics with partition counts, replication factors, and per-partition activity rates.
Consumer Groups
See all consumer groups, their state, member count, and per-partition lag.
4. View the Current Proposal
Click Proposals in the navigation to see the current rebalancing proposal. Pilot continuously generates proposals in the background based on cluster state.
The proposal shows:
- Expected effect on each load - the broker farthest from its fair share now and after the plan
- Proposed moves with source and destination brokers
- Movement cost (leader-only, inter-broker, or cross-rack)
You can also fetch it via the API:
curl http://localhost:8080/api/v1/proposals/current5. Apply the Proposal
Review the proposal details. When ready, apply it from the UI or API (requires a license):
curl -X POST http://localhost:8080/api/v1/proposals/current/applyMonitor reassignment progress in real-time from the dashboard, or via API:
curl http://localhost:8080/api/v1/reassignmentsNext Steps
- Docker Quick Start - full stack with Docker Compose
- Environment Variables - tune balance thresholds and throttling
- Monitoring Setup - connect Prometheus and Grafana
- Self-Healing - enable automatic rebalancing