Summary

Closes GH-5040

Adds a new database-tips-appendix.adoc page to the reference documentation documenting database-specific caveats and performance tips that are not enforced by Spring Batch itself but that commonly surprise developers.

Contents

MySQL

JdbcCursorItemReader — streaming vs. in-memory ResultSet

• By default Connector/J loads the entire ResultSet into memory
• Option 1: fetchSize(Integer.MIN_VALUE) + verifyCursorPosition(false) for row-by-row streaming
• Option 2: useCursorFetch=true in JDBC URL with a specific page size

JdbcBatchItemWriter — rewriting batched statements

• rewriteBatchedStatements=true in the JDBC URL rewrites individual INSERT/UPDATE statements into multi-row form, typically yielding 2×–10× throughput improvement

PostgreSQL

SELECT FOR UPDATE SKIP LOCKED with JdbcCursorItemReader

• Holding an open cursor while the writer modifies locked rows in the same transaction causes lock conflicts on PostgreSQL
• Workaround: use JdbcPagingItemReader instead (pages are fetched in separate queries)

Cross-reference to FK indexing

• Links to the existing schema-appendix.adoc section on FK index recommendations

Test plan

• Verify AsciiDoc renders without errors
• Confirm nav entry appears correctly
• Check xref to schema-appendix.adoc#recommendationsForIndexingMetaDataTables resolves

🤖 Generated with Claude Code