Menu
Amazon Relational Database Service
User Guide (API Version 2014-10-31)

Database Engine Updates 2016-04-06

Version: 1.6

This update includes the following improvements:

New Features:

  • Parallel read-ahead – Parallel read-ahead is now enabled by default for all Amazon Aurora DB clusters, and is not configurable. Parallel read-ahead was introduced in the December 2015 update. For more information, see Database Engine Updates 2015-12-03.

    In addition to enabling parallel read-ahead by default, this release includes the following improvements to parallel read-ahead:

    • Improved logic so that parallel read-ahead is less aggressive, which is beneficial when your DB cluster encounters many parallel workloads.

    • Improved stability on smaller tables.

  • Efficient storage of Binary Logs (lab mode) – MySQL binary log files are now stored more efficiently in Amazon Aurora. The new storage implementation enables binary log files to be deleted much earlier and improves system performance for an instance in an Amazon Aurora DB cluster that is a binary log replication master.

    To enable efficient storage of binary logs, set the aurora_lab_mode parameter to 1 in the parameter group for your primary instance or Aurora Replica. The aurora_lab_mode parameter is an instance-level parameter that is in the default.aurora5.6 parameter group by default. For information on modifying a DB parameter group, see Modifying Parameters in a DB Parameter Group. For information on parameter groups and Amazon Aurora, see DB Cluster and DB Instance Parameters.

    Only turn on efficient storage of binary logs for instances in an Amazon Aurora DB cluster that are MySQL binary log replication master instances.

  • AURORA_VERSION system variable – You can now get the version of your Aurora DB cluster by querying for the AURORA_VERSION system variable.

    To get the Amazon Aurora version, use one of the following queries:

    Copy
    select AURORA_VERSION();
    Copy
    select @@aurora_version;
    Copy
    show variables like '%version';

    You can also see the Amazon Aurora version in the AWS Management Console when you modify a DB cluster, or by calling the describe-db-engine-versions AWS CLI command or the DescribeDBEngineVersions API action.

  • Lock manager memory usage metric – Information about lock manager memory usage is now available as a metric.

    To get the lock manager memory usage metric, use one of the following queries:

    Copy
    show global status where variable_name in ('aurora_lockmgr_memory_used');
    Copy
    select * from INFORMATION_SCHEMA.GLOBAL_STATUS where variable_name in ('aurora_lockmgr_memory_used');

Improvements:

  • Improved stability during binlog and XA transaction recovery.

  • Fixed a memory issue resulting from a large number of connections.

  • Improved accuracy in the following metrics: Read Throughput, Read IOPS, Read Latency, Write Throughput, Write IOPS, Write Latency, and Disk Queue Depth.

  • Fixed a stability issue causing slow startup for large instances after a crash.

  • Improved concurrency in the data dictionary regarding synchronization mechanisms and cache eviction.

  • Stability and performance improvements for Aurora Replicas:

    • Fixed a stability issue for Aurora Replicas during heavy or burst write workloads for the primary instance.

    • Improved replica lag for db.r3.4xlarge and db.r3.8xlarge instances.

    • Improved performance by reducing contention between application of log records and concurrent reads on an Aurora Replica.

    • Fixed an issue for refreshing statistics on Aurora Replicas for newly created or updated statistics.

    • Improved stability for Aurora Replicas when there are many transactions on the primary instance and concurrent reads on the Aurora Replicas across the same data.

    • Improved stability for Aurora Replicas when running UPDATE and DELETE statements with JOIN statements.

    • Improved stability for Aurora Replicas when running INSERT … SELECT statements.

Integration of MySQL bug fixes:

  • BACKPORT BUG#18694052 FIX FOR ASSERTION `!M_ORDERED_REC_BUFFER' FAILED TO 5.6 (Port Bug #18305270)

  • SEGV IN MEMCPY(), HA_PARTITION::POSITION (Port Bug # 18383840)

  • WRONG RESULTS WITH PARTITIONING,INDEX_MERGE AND NO PK (Port Bug # 18167648)

  • FLUSH TABLES FOR EXPORT: ASSERTION IN HA_PARTITION::EXTRA (Port Bug # 16943907)

  • SERVER CRASH IN VIRTUAL HA_ROWS HANDLER::MULTI_RANGE_READ_INFO_CONST (Port Bug # 16164031)

  • RANGE OPTIMIZER CRASHES IN SEL_ARG::RB_INSERT() (Port Bug # 16241773)