This article explains how to manage memory pressure in Amazon Aurora MySQL to prevent OOM-related instance restarts. It covers the use of built-in memory management features, the roles of shared and per-session memory structures, and strategies for diagnosing and mitigating memory exhaustion at the source.
Key points
Aurora MySQL avoids OS-level OOM kills by monitoring memory and applying recovery actions before the database crashes.
Major memory consumers include the InnoDB buffer pool, per-connection buffers, and memory-intensive query operations like sorting or joining.
Version 8.4 introduces aurora_enable_memory_management for dynamic, automated recovery compared to manual parameter configuration in older versions.
Effective OOM prevention involves combining automatic management with workload optimization, such as connection pooling and query indexing.