Database Troubleshooting

Separate connection refusal, permissions, schema changes, locks, replication, and pool failures into server-state and client-configuration branches.

Scope to Record First

Preserve the product version, connection target, transaction state, execution plan, and recent schema changes before attempting a fix.

Use MySQL 8.4 rules and a schema comparison to distinguish INT, BIGINT and signedness mismatches from row-data failures.

A local hash-slot calculation separates SET expiry preservation from the same-slot requirement for multi-key scripts.

Compare session state with transaction lifetime and separate schema selection from version-dependent prepared-statement support.

Separate client-to-PgBouncer and PgBouncer-to-PostgreSQL authentication before changing passwords or pooling mode.

Correlate per-database/user waits with PostgreSQL query and lock time before increasing connection limits.

Distinguish unassigned.reason from current allocation deciders before choosing retries, capacity changes or recovery.

Combine alarm, queue and connection observations to distinguish consumer throughput, acknowledgements and connection pressure.

Separate RDB persistence failures, permissions, disk space and fork pressure before treating unblocked writes as recovery.

Compare UUID representations and MySQL/MariaDB differences before changing column width or binary conversion.

Two SQLite connections distinguish a held reader from a blocked second writer and show why neither alone proves a 504 cause.