New Features

Several optimizations have been added to improve performance on partitioned tables, including:

Additional partitioned table performance optimizations may be added in future releases.

Previously, for externally started database processes, the database processes may not distribute evenly across the available Admin processes (APs). A new start te and start sm parameter is used to supply the Admin with distribution policies used to ensure better distribution.

The Asynchronous SM feature allows administrators to configure NuoDB in an active / passive configuration across long distance regions for operational and disaster recovery protection in the event of a primary site outage. The Asynchronous SM feature replicates all database changes to a remote passive data center asynchronously so that there is limited impact due to higher network latency that is common to Wide Area Networks (WANs). In the event of a primary site disaster, the Asynchronous SM can be converted to a regular SM so that the database can be restarted quickly and put back into service. For more information, see Asysnchonous SM.

This new capability allows users to create a hot copy backup, using nuocmd hotcopy with the --type simple argument, and that backup (or a copy of that backup) can then be used directly to start a database. This feature was previously available in the NuoDB Nuoagent administrative tier and has been added to the NuoDB Admin administrative tier.

The new system-wide database DEFINER security option allows a user to read a database VIEW object after being granted a single Grant Select on View privilege. This new option eases database administration and improves security when using a DEFINER security model is preferred. The database default remains the INVOKER security option. To enable the DEFINER security option, the new system property DEFAULT_VIEW_SECURITY must be set to DEFINER.
For more information when using the Scalar Execution Engine, see SQL System Properties.
For more information when using the Vector Execution Engine, see SQL System Properties.

The service streamlines configuration with third-party Key Management Systems and is used for TDE initial password configuration as well as password rotation.

Added two new Transaction Engine (TE) timing metrics, SQLListenerIdleTimeStatement and SQLListenerIdleTimeTransaction that narrow down SQLListenerIdleTime to when statements or transactions are open respectively. A new "Time" monitor metric was also added that timestamps when each monitoring period ended expressed in milliseconds. For more information, see Metrics Published by Database Processes.

Previous behavior could lead to a full database shutdown during failure detection and resolution processing when the database environment is experiencing a heavy network congestion event. Failure resolution processing will now continue to run and complete (possibly taking a little longer in some scenarios) but will avoid an unwanted forced database shutdown event.

Previously the number of index builds which may run concurrently on a single SM was not limited. If many concurrent long-running index builds were issued, the Storage Manager (SM) could become resource constrained and slow end user application response time. To avoid excessive index build resource consumption, a new limit of 4 concurrent index build jobs is now enforced. For more information, see either reference,
Scalar Execution Engine: CREATE INDEX.
Vectorized Execution Engine: CREATE INDEX.

Previously Raft log compaction was performed in the critical path of commits whenever the number of committed and uncompacted log entries exceeded raft.compactionThreshold (1024 by default). This would delay the command response that triggered compaction and also disrupt leadership. Now compaction is performed by a task that is scheduled every 30 seconds by default and is configurable by the nuoadmin.conf file raft.compactionIntervalSec property. If the compaction task detects that the number of committed and uncompacted entries in the Raft log is equal to or greater than raft.compactionThreshold, compaction is performed.

New switch options added are --wait-for-acks, --fail-on-process-exit, and --check-converged. In addition, the nuocmd check process command now will fail-fast if the process exits during the command invocation, and a new nuocmd check subcommand server has been added. To learn more about these enhancements, and when to use them, see NuoDB Command Reference.

A new auditing capability has been added in NuoDB Admin which allows logging of all REST API requests and responses. The type and verbosity of the logged messages are configurable. By default all authentication and authorization failures are logged, and the type and verbosity of the logged messages are configurable. For more information, see the reference documentation for the NuoDB Admin Audit Log file.

When a network connection between a Transaction Engine and a SQL client is terminated, some virtual environments do not notify the engine of the socket termination. Without a TCP notification, the engine will keep the socket open until a timeout is reached. On most systems, the default timeout is two hours. The new --sql-client-tcp-keepalive allows users to override the system defaults using the TCP_KEEPIDLE:TCP_KEEPCNT:TCP_KEEPINTVL format.