starrocks-explain-plan

To diagnose a slow StarRocks query, run EXPLAIN ANALYZE to get runtime metrics, then systematically check OlapScanNode for partition pruning, verify runtime filter IDs between HashJoinNode and scan nodes, and review join strategy and shuffle costs to pinpoint bottlenecks.

EXPLAIN VERBOSE shows the physical execution plan structure including fragments and pipelines, while EXPLAIN COSTS provides the costed view with cardinality estimation and statistics. Both help interpret distributed ExchangeNode communication and optimizer decisions for performance tuning.

High shuffle cost in StarRocks execution plans often stems from suboptimal join strategy selection or missing COLOCATE joins. Decode fragment and ExchangeNode communication types to identify where distributed data movement is introduced, and verify cardinality estimation correctness.

To verify materialized view rewrite in StarRocks, check the rollup and table fields in your EXPLAIN output. The plan should indicate whether the scan reads from an MV rollup instead of the base table, confirming the optimizer successfully applied the rewrite.

Yes, you can check runtime filter pushdown in StarRocks EXPLAIN by confirming runtime filter IDs match between HashJoinNode and scan nodes. This verifies that filters are properly propagated to scan operators for early data pruning.

To find the slowest operator in StarRocks, use EXPLAIN ANALYZE metrics including OperatorTotalTime, PeakMemoryBytes, and row movement patterns (Pull/Push). These operator-level runtime metrics directly locate performance bottlenecks within the pipeline.