############# Release Notes ############# 4.6.5 ===== Bindings -------- * Java bindings now perform marshaling off of the network thread. 4.6.4 ===== Features -------- * Added ability to use --io_trust_seconds in a warn-only mode, which logs a trace event rather than failing the process when a disk operation takes a long time. This is enabled with --io_trust_warn_only. Fixes ----- * Disk operation timeouts now cause the process to restart rather than hang indefinitely. * ``fdbdr switch`` did not start the DR in the opposite direction correctly, resulting in mutations being lost. * Lowered backup and DR batch sizes to avoid large packet warnings. * Remove partial pipelining of tlog commits. 4.6.3 ===== Features -------- * Added the ability to run a consistency check of a database using a new server role. Fixes ----- * Added the ability to automatically shutdown processes if a disk operation takes a long time to complete. This is enabled with --io_trust_seconds. * Too many outstanding storage recruitment requests causes the cluster controller to hang. * Corrected issue with Ubuntu installer package on Ubuntu 16.04 not starting daemon. * Package non-Linux builds of JNI component into Java jars. * Published backup-related binaries on macOS were incorrectly pointing to symbolic link specification files. Performance ----------- * We no longer fsync trace files. * Lowered the default bandwidth shard splitting knobs for better performance with hot key ranges. 4.6.2 ===== Fixes ----- * The tlog could commit more than 100MB at a time * Metrics with filename component not present in trace events * Setting new locality information causes missing process metrics in status details * FDB processes killed via CLI could hang while killing themselves * Enabled recovery of on-disk data files in the event of a very specific rare corruption situation * Process messages get reported as errors by status, but don't get attributed to a process in the status details list * DR prematurely reported progress for work that needed to be retried Performance ----------- * Storage engine performance improvements to reduce the overhead that ssd-2 requires for its benefits over ssd-1 * Lowered the default fetch keys parallelism to slow down data distribution 4.6.1 ===== Fixes ----- * Starting a new DR on a large database can cause the secondary cluster to lose availability * Secondary clusters that have been upgraded were reporting "primary" metrics * Backup and DR could get stuck if too many tasks timed out simultaneously 4.6.0 ===== Features -------- * Added a new storage engine type ``ssd-2`` that includes page checksums and more efficient storage of large values. The previous storage engine has been renamed ``ssd-1``, and the storage engine ``ssd`` is an alias for ``ssd-2``. * DR and Restore won't overwrite a non empty database Performance ----------- * Improve performance of the ssd storage engine in databases with large keys or values * Improved cluster recovery speed * Restore is faster due to better load leveling across the keyspace * Reduced the conflict ranges applied for get range calls in rare cases Fixes ----- * Backup to Blobstore sends and verifies MD5 sums for uploads * Backup restoration could be unrestorable in certain cases * Clients using the multi-version client functionality would incorrectly report incompatible connections in status * Backup and DR network metrics were incorrectly reported as 0 * Java: fix race condition when removing an empty directory which could lead to a NoSuchElementException * Fixed a source of potential crashes in fdbcli Bindings -------- * API version updated to 460. There are no behavior changes in this API version. See the :ref:`API version upgrade guide ` for upgrade details. Status ------ * The following fields were added: cluster.data.moving_data.total_written_bytes, cluster.qos.limiting_queue_bytes_storage_server, cluster.qos.worst_version_lag_storage_server, cluster.qos.limiting_version_lag_storage_server, cluster.qos.transaction_per_second_limit, cluster.qos.released_transactions_per_second, cluster.qos.performance_limited_by.reason_id, and cluster.database_available Earlier release notes --------------------- * :doc:`4.5 (API Version 450) ` * :doc:`4.4 (API Version 440) ` * :doc:`4.3 (API Version 430) ` * :doc:`4.2 (API Version 420) ` * :doc:`4.1 (API Version 410) ` * :doc:`4.0 (API Version 400) ` * :doc:`3.0 (API Version 300) ` * :doc:`2.0 (API Version 200) ` * :doc:`1.0 (API Version 100) ` * :doc:`Beta 3 (API Version 23) ` * :doc:`Beta 2 (API Version 22) ` * :doc:`Beta 1 (API Version 21) ` * :doc:`Alpha 6 (API Version 16) ` * :doc:`Alpha 5 (API Version 14) `