| name | starrocks-monitoring |
| description | Monitor and maintain a StarRocks cluster — health/metrics (backends, processlist, killing queries), compaction management, tablet repair/rebalancing, routine maintenance schedules, statistics collection for CBO, and troubleshooting commands. Use when checking StarRocks cluster health, diagnosing compaction or tablet issues, killing runaway queries, collecting statistics, or planning routine maintenance. |
| license | Apache-2.0 |
| metadata | {"author":"Tiansu Yu","version":"1.0"} |
StarRocks Monitoring and Maintenance
Observe and maintain a running StarRocks cluster: health metrics, compaction, tablet repair, statistics, and routine upkeep.
When to Use
- Checking cluster/BE health, disk usage, or tablet distribution
- Inspecting running queries and killing runaway ones
- Diagnosing or triggering compaction
- Repairing or rebalancing tablets/replicas
- Collecting statistics for the cost-based optimizer (CBO)
- Planning daily/weekly/monthly maintenance
Related skills: [starrocks-cluster-setup] for node topology and config; [starrocks-resource-management] for workload isolation; [starrocks-backup-recovery] for snapshots; [starrocks-query-optimization] for query-level diagnosis.
Key Metrics to Monitor
Cluster Health:
SHOW PROC '/backends';
SHOW PROC '/statistic';
SHOW PROC '/statistic/{be_id}';
Query Performance:
SELECT * FROM information_schema.loads ORDER BY StartTime DESC LIMIT 10;
SHOW PROCESSLIST;
KILL QUERY connection_id;
SHOW QUERY PROFILE '<query_id>';
Data Loading:
SHOW LOAD ORDER BY CreateTime DESC LIMIT 10;
SHOW LOAD WHERE State = 'CANCELLED';
SELECT * FROM information_schema.loads ORDER BY create_time DESC LIMIT 10;
Compaction Management
SHOW PROC '/compactions';
ALTER TABLE table_name COMPACT;
ALTER TABLE table_name COMPACT partition_name;
ALTER TABLE table_name CUMULATIVE COMPACT;
UPDATE information_schema.be_configs
SET value = "100"
WHERE name = "max_cumulative_compaction_num_singleton_deltas";
Tablet Repair and Rebalancing
ADMIN SHOW REPLICA STATUS FROM table_name;
SHOW TABLET FROM table_name;
ADMIN REPAIR TABLE table_name;
ADMIN REBALANCE DISK;
Routine Maintenance Tasks
Daily:
- Monitor load job failures
- Check query latencies
- Review disk usage
Weekly:
- Analyze slow queries
- Review partition retention
- Check compaction lag
Monthly:
- Review and optimize bucket counts
- Analyze data skew
- Update statistics (if using CBO)
Statistics Collection (for CBO)
ANALYZE TABLE table_name;
ANALYZE TABLE table_name (col1, col2);
SHOW ANALYZE STATUS;
SHOW META;
Troubleshooting Commands
ADMIN SHOW FRONTEND CONFIG LIKE 'log%';
SHOW VARIABLES;
SHOW SESSION VARIABLES;
ADMIN CHECK TABLET (tablet_id);
EXPLAIN VERBOSE SELECT ...;
EXPLAIN COSTS SELECT ...;