Version End of Life. Not Yet Set
Release 6.1.4 contains a number of improvements and bug fixes, specifically for the tpm command line tool and stability improvements for Composite Active/Active topologies. In addition this release now fully supports the latest binlog compression feature of MySQL 8.0.20.
Improvements, new features and functionality
When a Primary is
SHUNNED, we no longer mark the whole
data service as
This will allow reading from Replicas to continue even if the Primary is offline.
In a Composite Active/Active topology, the relay hosts, by default, will pull THL from the remote Primaries. This can
be changed to pull from remote Replica(s) by use of the following setting
This is a change in behaviour. From version 6.0.4 up to and including 6.1.3, the default behaviour was configured to pull from the remote Replica(s).
The tpm diag command now captures cctrl ls output from the Composite dataservice when appropriate.
Fixes an issue where affinity specified in the configuration (
was ignored since it was not written to
user.map (this affects proxy configurations only).
Fixed a bug where tpm diag would fail to gather some MySQL information on a Composite Active/Active node.
Fixed a bug in tungsten_post_process where the
file would not be read properly. Also adds two warnings for skipped entries and corrects a typo in the property definitions.
Fixed an issue where connector client-list and the Proxy-mode command, tungsten show processlist,
NullPointerException errors when listing disconnected client applications.
When in proxy mode and fetching data source configuration at startup, the Connector will now properly parse, and use, affinity (and other URL options if any) to establish the connection.
Fixes an issue where connections with multiple affinity would not keep the current channel and reconnect to the database server while the connection could be reused.
In certain edge case situations, the manager could choose a manually shunned node as a viable node during failover.
cluster topology validate within cctrl would fail
if the cluster contained an Active Witness host.
Piping multiple commands to cctrl that would affect components in a remote cluster would fail, for example:
echo "use east; replicator db5 offline" | cctrl
Resolves an edge case in a Composite Active/Passive topology, with 2 or more Composite Passive dataservices, where a switch of a relay node in a single Replica service would incorrectly reconfigure all relays in the other Replica services.
Tungsten Clustering 6.1.4 Includes the following changes made in Tungsten Replicator 6.1.4
Release 6.1.4 contains a number of improvements and bug fixes, specifically for the tpm command line tool and improvements to the Redshift Applier. In addition this release now fully supports the latest binlog compression feature of MySQL 8.0.20.
Improvements, new features and functionality
Improves tpm performance by using more efficient routines to calculate paths.
Added the ability for tpm diag to skip both individual gather subroutines along with entire groups of gather subroutines.
Also added ability to list all gather groups and subroutines using
--listfor use with the
tungsten_provision_slave has been rewritten fixing a number of issues in the previous release. This version was previously released as the Beta script prov-sl.sh.
A number of performance improvements and fixes have been incorporated into the Drizzle Driver used for communication between components and MySQL, these include:
Better handling of large queries close to max network packet size.
Batch Support. Instead of sending statements one by one, the driver will be able to send multiple statements at once, avoiding round trips between the driver and MySQL server.
Fixes issues with interpreting useSSL on connect string URLs.
Issues: CT-1215, CT-1216, CT-1217, CT-1228
The tungsten_send_diag command now accepts arguments for the tpm diag command using
You must enclose the arguments in quotes, for example:shell>
tungsten_send_diag -c 9999 -d --args ‘--all -v’
debughas been disabled by default in the schemachange filter. Resulting in reduced noise in the replicator log file.
A new replicator role
thl-clienthas been added. This new role allows the replicator to download THL from a Primary, but not apply to the target database.
delayInMsfilter has been added which allows the applying of THL to a Replica to be delayed. The filter allows millisecond precision. This filter works in the same way as the
TimeDelayFilter, however that filter only allow second precision.
The filter works at ROW level and allows the filtering out of rows based on one or more column/value matches
When configuring the Redshift applier, you can now configure which tool the applier will use for posting CSV files to S3. Options are s3cmd (default), s4cmd or aws.
A number of improvements have been made to the Redshift Applier to allow optional levels of table locking.
This is particularly useful when you have multiple Redshift Appliers in a Fan-In topology, and/or very high volumes of data to process.
The additional locking options reduce the risk of Redshift Serializable Isolation Violation errors occuring.
Full details of how to utilise the new options can be read at Handling Concurrent Writes from Multiple Appliers
Tungsten Replicator can now correctly extract and parse Binary Log entries when the MySQL option
binlog-transaction-compressionhas been enabled.
binlog-transaction-compressionis a new parameter introduced from MySQL 8.0.20.
In certain edge cases, tungsten_provision_slave would fail to detect if mysql was shutdown.
tpm diag now collects directories specified with
Fixes the tpm update command, which would exit with the error:Argument " (error code 1)" isn't numeric
tpm diag now collects any files specified by
!includedirectives in the
tpm diag also looks in
Fixes a bug which prevented tungsten_send_diag from uploading a self-generated diagnostic zip file.
tpm diag now properly derives the correct target path to the releases directory if the home directory in the configuration points to a sym-link.
Removed tpm diag call to sudo for gathering ifconfig and lsb_release commands.
tpm update would fail and report errors if deployall had been executed.
tpm diag no longer requires the mysql command-line client when running on non-MySQL Applier nodes, and no longer attempts to gather any MySQL diagnostic information.
Removes the requirement to execute component start/stop commands with sudo when called through systemd.
This is specific to start/stop actions following the use of the deployall scripts.
Fixes cases where tpm fails when the OS
hostnamecommand returns a different string than what is used in the configuration (i.e.
hostnamereturns a FQDN, yet the configuration contains shortnames like db1, db2, etc.).
In certain cases, after a reprovision, tungsten_provision_slave didn’t always run the steps to reset the local replicator service. This made the replicator go into an error state after provision had completed.
The tpm diag command now handles the cluster-slave topology more gracefully, and properly handles cluster nodes without the Connector installed.
Improved output text clarity by converting multiple verbose-level outputs to debug, and warnings to notice-level.
tpm diag now gathers sym-linked files correctly.
ddlscan now sets the datatype for sequence number columns to BIGINT when generating staging table DDL for Redshift deployments.
Fixes a situation where tpm update exits with a
In heterogeneous replicator deployments, the
convertstringfrommysqlfilter would fail to convert strings for alphanumeric key values.
Tungsten Replicator could wrongly think it is already in
DEGRADEDmode when trying to put it into
Tungsten Replicator now recognises Amazon AWS SSL Certificates to enable SSL communication with AWS Aurora.
When host server time (and thus MySQL time) is not configured as UTC, issuing cluster heartbeat or trepctl heartbeat in the first hours around daylight savings time would create an invalid time in MySQL.
For more information on timezones when issuing heartbeats, see trepctl heartbeat Time Zone Handling
The replicator would fail to apply into Vertica when configured as an offboard installation due to the applier incorrectly expecting the csv files to exist locally on the remote Vertica host.
Due to a change in the Binary log structure introduced in MySQL 8.0.16, the replicator would fail to extract transactions for Partitioned tables.
The replicator would fail to correctly parse statements with leading comment blocks in excess of 200 characters.