Apache Druid, time-series OLAP at any scale.

Queries Kafka supervisor consumer lag across topic partitions and verifies active worker slot allocations on middleManagers.

# Query active supervisor status, consumer lag, and partition assignments
curl -s http://localhost:8888/druid/indexer/v1/supervisor/telemetry_events_kafka/status | jq '{
  id: .id,
  generationTime: .generationTime,
  state: .payload.state,
  aggregateLag: .payload.aggregateLag,
  partitionLag: .payload.partitionLag
}'

# Inspect MiddleManager active tasks, capacity, and worker slots
curl -s http://localhost:8888/druid/indexer/v1/workers | jq '.[] | {
  host: .worker.host,
  capacity: .worker.capacity,
  currCapacityUsed: .currCapacityUsed
}'

Surfaces fragmented small segment counts by datasource and audits Historical node memory utilization.

-- Query segment count, total size, and average segment size per datasource
SELECT datasource,
       COUNT(*) AS total_segments,
       ROUND(SUM("size") / 1024 / 1024, 2) AS total_size_mb,
       ROUND(AVG("size") / 1024 / 1024, 2) AS avg_segment_size_mb
FROM sys.segments
WHERE is_active = 1
GROUP BY datasource
ORDER BY total_segments DESC;

-- Identify Historical nodes approaching max heap memory usage
SELECT host, curr_size, max_size,
       ROUND((curr_size * 100.0) / max_size, 2) AS utilization_pct
FROM sys.servers
WHERE server_type = 'historical';