analyzing-range-distribution

Use SHOW RANGES FROM TABLE followed by your table name to list range counts, boundaries, and leaseholders via SQL. Grouping by lease_holder reveals distribution across nodes, and adding WITH DETAILS exposes range sizes, though DETAILS requires a LIMIT clause for safety.

Run SELECT lease_holder, COUNT(*) FROM [SHOW RANGES FROM TABLE your_table] GROUP BY lease_holder to compute per-node percentages. A single node holding over 40% of leaseholders in a balanced cluster indicates a hotspot, fixable with lease_preferences zone configs.

SHOW RANGES requires the admin role or the ZONECONFIG system privilege. Granting ZONECONFIG via GRANT SYSTEM ZONECONFIG TO user is the recommended least-privilege approach for read-only range analysis without write access.

WITH DETAILS computes span statistics on-demand, causing high CPU and memory usage proportional to range count. It is safe only when targeted at a specific table with a LIMIT of 50-100 during low-traffic windows; never run it cluster-wide without LIMIT.

Excessive small ranges indicate fragmentation from load-based splitting during high write throughput or sequential inserts. Calculate ranges per GB to assess severity: 1-15 is healthy, while 50+ suggests severe fragmentation that may warrant tuning range_max_bytes.

Query SHOW RANGES and filter where array_length(replicas, 1) is less than your replication factor, typically 3. Under-replicated ranges signal node failures, decommissioning operations, or zone config constraint mismatches requiring investigation.