Dear MySQL Users,
A new GA (general availability) version of MySQL Shell
(a component of the MySQL Server) has been made
available: MySQL Shell 1.0.10 GA. The MySQL Shell is provided
under Oracle's dual-license.
MySQL Shell is an interactive JavaScript, Python and SQL console
interface, supporting development and administration for the MySQL
Server. It provides scriptable API that supports the creation and
management of MySQL InnoDB clusters, as well as a modern fluent CRUD
API for the MySQL Document Store.
The AdminAPI provides an integrated solution for high availability and
scalability using InnoDB based MySQL databases, without requiring
advanced MySQL expertise.
The Document Store DevAPI enables developers and DBAs to create
"schema-less" JSON document collections and perform Create, Update,
Read, Delete (CRUD) operations on those collections with little to
no knowledge of the SQL language.
Additionally, MySQL Shell provides partial compatibility with the
mysql command line client by supporting many of the same command line
options. As with the mysql command line, MySQL Shell can also be used
to connect to and send SQL commands to pre-5.7.12 versions of the
MySQL server using the standard MySQL network protocol.
For more information about how to configure and work with an InnoDB cluster
see https://dev.mysql.com/doc/refman/en/mysql-innodb-cluster-userguide.html
For more information about how to use MySQL Shell and the MySQL Document
Store support see https://dev.mysql.com/doc/refman/en/document-store.html
For full documentation on MySQL Server, MySQL Shell and related topics,
see http://dev.mysql.com/doc/refman/en/
If you want to write applications that use the the CRUD based X DevAPI
use the latest MySQL Connectors for your language of choice. You can
also find documentation about Connectors in the MySQL documentation
site.
For more information about how to download MySQL Shell, see
http://dev.mysql.com/downloads/shell/
We welcome and appreciate your feedback and bug reports, see
http://bugs.mysql.com/
Enjoy!
Changes in MySQL Shell 1.0.10 (2017-07-28, General Availability)
Security Notes
* Security Fix: The linked OpenSSL library for MySQL Shell
Commercial has been updated to version 1.0.2l. Issues
fixed in the new OpenSSL version are described at
http://www.openssl.org/news/vulnerabilities.html.
This change does not affect Oracle-produced MySQL
Community builds of MySQL Shell, which use the yaSSL
library instead. (Bug #26284262)
* Security Fix: Running queries on servers using MySQL 8
against a table with a text column caused an unexpected
halt.
Functionality Added or Changed
* The clusterAdminType option of the dba.createCluster()
was unused and has been removed. (Bug #25585027)
* The --help message now displays all short-form options.
The --help message now displays complete information for
the --json option.
shell.help('options') now returns usable information
about mysqlsh options.
cluster.help('checkInstanceState') now reports correct
information.
cluster.help('addInstance') and
cluster.help('rejoinInstance') now report complete
information.
help() now reports complete information for all AdminAPI
and Shell commands. (Bug #24841493, Bug #24910540, Bug
#24841749, Bug #25701345, Bug #24977542, Bug #25268410,
Bug #25266846)
* Calling the modify() or remove() function without a
parameter caused the function to be executed against the
whole collection, which could cause unexpected results
such as deleting all rows in a table. To avoid this and
make the behavior consistent with update() and delete(),
a client-side exception is now thrown if the modify() or
remove() function is called without a parameter. Now, to
execute the modify() or remove() function against a
collection call them with an expression that evaluates to
true, for example remove('true') or modify('true').
Bugs Fixed
* When making cluster topology or membership changes,
AdminAPI was not taking into consideration the value of
group_replication_group_name, which could lead to
incorrect, non-deterministic results in scenarios such as
split-brains. Now, the following commands validate the
InnoDB cluster Metadata and the corresponding instance's
group_replication_group_name value:
+ dba.getCluster()
+ Cluster.rejoinInstance()
+ Cluster.forceQuorumUsingPartitionOf()
If the values of group_replication_group_name do not
match, the commands abort with an error.
dba.rebootClusterFromCompleteOutage() was also updated to
ensure that the group_replication_group_name variable has
not been changed before rejoining the instance. (Bug
#26159339)
* AdminAPI now always uses the active user value for the
current mysqlsh session, whether the value was explicitly
specified by the user or is the result of an implicit
default used by mysqlsh. (Bug #26132527)
* The addInstance() function was failing when run against a
server using MySQL 8. (Bug #26085799)
* Executing AdminAPI commands on a server with a version of
Python lower than 2.7 was failing without the correct
error message. (Bug #25975317)
* The checks performed by the AdminAPI upon issuing
dba.rebootClusterFromCompleteOutage() were more strict
than those required by Group Replication. Now, the
AdminAPI considers tables with a Primary Key Equivalent
(such as a Non Null Unique Key) as compatible, matching
the current requirement for Group Replication. (Bug
#25974689)
* Executing dba.createCluster() on an instance with
sql_mode=ANSI_QUOTES failed with an error. (Bug
#25966057)
References: See also: Bug #25990172.
* The server version is now displayed as part of the
"Connected to" message. (Bug #25824753)
* When using MySQL Shell on Windows any files created or
opened, for example those used during
dba.createSandboxInstance(), could not be deleted. (Bug
#25789094)
* Queries that used an expression of the form
collection.find(identifier IN identifier) caused an
abnormal mysqlsh exit. (Bug #25754078)
* If AdminAPI functions were interrupted, for example with
control-C on Unix, the exit code returned by the process
was 1 instead of the expected one for terminated
processes, for example SIGINT. (Bug #25728759)
* The previous limitation of auto_increment_increment being
set according to the random 32-bit unsigned integers used
for the server-id system variable has been resolved. Now,
when operating in single-primary mode
auto_increment_increment is set to 1 and
auto_increment_offset is set to 2. In multi-primary mode
auto_increment_increment is set to 7 and
auto_increment_offset is set to 1 plus the server_id
modulo 7. (Bug #25724564)
* The randomly generated passwords used by internal users
were not compatible with instances running the Password
Validation plugin. (Bug #25714751)
* The help for dba.configureLocalInstance(instance[,
options]) has been improved to describe the returned JSON
object. (Bug #25703028)
* The options in the MySQL Shell options dictionary are now
fully documented. (Bug #25701345)
* Issuing configureLocalInstance() when using a URI that
contained a user without the correct privileges resulted
in an incorrect new user being created. Now, if the user
in configureLocalInstance() URI does not have enough
privileges to grant all the necessary privileges for the
new user chosen during the interactive wizard
configuration the user is not created. (Bug #25614855)
* Issuing Cluster.rescan() resulted in non-determinstic
behavior which could produce incorrect JSON output,
showing an instance that was already part of the cluster
as belonging to the newlyDiscoveredInstances[] list and
to the unavailableInstances[] list. This also resulted in
MySQL Shell prompting to add or remove the instance from
the cluster. (Bug #25534693)
* AdminAPI functions now accept the standard connection
parameters as used by shell.connect. New validations have
been added for when require_secure_transport is ON, now
it is not possible to to create a cluster with
memberSslMode:DISABLED or to add an instance with
require_secure_transport=ON to a cluster where
memberSslMode:DISABLED. (Bug #25532298)
* The parsing of account names, for example when passing
the clusterAdmin option to dba.configureLocalInstance()
has been improved. (Bug #25528695)
* Issuing configureLocalInstance() twice could fail. (Bug
#25519190)
* When passing the rejoinInstances[] option to
dba.rebootClusterFromCompleteOutage(), if no
rejoinInstances[] option was specified then members were
being incorrectly handled during the rebuild. Now,
instances that are eligible to be added to the
rejoinInstances[] list but that are specified in the
removeInstances[] list are skipped by the interactive
wizard that tries to automatically build a
rejoinInstances[] list if one was not provided. This fix
also ensures that both interactive and non interactive
use of MySQL Shell correctly verify the rejoinInstances[]
list does not contain a non reachable instance. (Bug
#25516390)
* When using dba.deploySandboxInstance() and passing in
sandboxDir, the specified path must not exceed 89
characters. (Bug #25485035)
* shell.connect() did not report an error if an invalid
argument was used. An ArgumentError is now issued for any
invalid argument.
The following mutually exclusive pairs of options are now
checked, and an error is issued if both are specified:
+ --password and --dbPassword
+ --user and --dbUser
+ --port and --socket
(Bug #25268670)
References: See also: Bug #24911173.
* Issuing removeInstance() on the last member of a cluster,
and particularly the seed member, was resulting in a
cluster that could not be dissolved. Now, issuing
removeInstance() on the last member of a cluster results
in an error, and you must use dissolve() on that instance
to ensure the cluster is correctly dissolved. (Bug
#25226130)
* removeInstance() resulted in unexpected behavior in some
cases, for example when an empty password was passed as
part of the URI to the instance. (Bug #25111911)
* The default for prompts has been made more consistent.
(Bug #25039886)
* A number of issues with the output of
shell.help("prompt") have been corrected. (Bug #25026855,
Bug #25242638, Bug #25676343, Bug #25176769)
* The usage of the terms global session and active session
have been made more consistent in the user interface.
(Bug #24954546)
* The removeInstance() was not working on stopped instances
and it was not possible to remove an unavailable instance
from the cluster. The fix adds a new option force to the
removeInstance() command to enable you to remove
instances from the metadata that are permanently not
available, avoiding obsolete data from being kept in the
metadata of the cluster. In addition the error message
provided when not using the force option has been
improved and the online help for the removeInstance() was
also updated accordingly. (Bug #24916064)
* Unsigned data could be incorrectly read from the
database. (Bug #24912358)
* MySQL Shell now displays an invalid year as 0000,
matching the behavior of the MySQL prompt, rather than as
0. (Bug #24912061)
* MySQL Shell did not display fractional seconds for values
in DATETIME columns. (Bug #24911885)
* The parsing of Unix sockets provided as part of a URI has
been improved. (Bug #24905066)
* Creating Classic sessions that connect using Unix sockets
now uses the correct defaults such as hostname. This
resolves the previous limitation of using Unix sockets to
connect to InnoDB cluster instances. See MySQL Shell Connections
(http://dev.mysql.com/doc/refman/5.7/en/mysql-shell-connections.html)
for information on how the defaults are
applied to socket connections. (Bug #24848763, Bug
#26036466)
References: See also: Bug #24911068.
* Some issues with the MySQL Shell command line help output
were fixed. (Bug #24841749, Bug #24841493, Bug #24910540)
* If an error occurs now both the error number and error
text are correctly displayed. (Bug #24816578)
* URIs were incorrectly parsed in MySQL Shell when
passwords were hidden. (Bug #24793956)
* The error messages generated by issuing
dba.deployLocalInstance() against an unsuitable or
incompatible instance have been improved. (Bug #24598272)
* The dba.createCluster(), dba.getCluster(), and
dba.rebootClusterFromCompleteOutage() functions have been
updated to validate the cluster name, using the following
rules:
+ Name must start with a letter or the _ character
+ Name can only contain alphanumeric characters and
the _ character
+ Cannot be longer than 40 characters
+ Cannot be empty
The Cluster.addInstance() function has been updated to
validate the label used on an instance in the cluster,
using the following rules:
+ Label can only contain alphanumerics or the _
character
+ Cannot be longer than 256 characters
+ Cannot be empty
(Bug #24565242)
* Using the --dba enableXProtocol option to install X
Plugin was failing. (Bug #23508428)
* mysqlsh stopped responding if the \source command was
given a directory (rather than file) argument. (Bug
#23097932, Bug #81060)
* When using cluster.rejoinInstance() any user specified in
the connection string passed in was being ignored and the
connection was using the root user. (Bug #85567, Bug
#25786495)
* On an instance configured as a multi-threaded slave, in
other words slave_parallel_workers set to greater than 0,
and with slave_parallel_type=DATABASE,
dba.checkInstanceConfiguration() was not detecting that
the instance was not correctly configured for InnoDB
cluster usage.
* If removeInstance() failed due to a connection error, an
error was reported but the instance was incorrectly
removed from the InnoDB cluster metadata, and remained
part of the replication group. The fix ensures the
metadata is correctly updated according to the result of
removeInstance().
* In a situation where a new primary instance was elected,
adding a new instance to the cluster resulted in an error
due to a failed connection to the previous primary
instance.
* The functions that modify server variables, such as
dba.createCluster() and dba.validateInstance() now
provide more information in interactive mode output and
log output about server variables which are changed when
executed.
* Deploying instances to paths with directories that
contained spaces was failing without error. Use double
backslash to specify such paths, for example
D:\\Cluster\\foo bar.
* The Cluster object obtained from functions such as
dba.createCluster() or dba.getCluster() became unusable
once the Shell session in which the object was created is
was connected to a different server. The fix modifies the
Cluster object so that:
+ The Cluster object holds an internal reference to
the Session from which it was created or retrieved.
+ AdminAPI functions that modify the Cluster are made
using the session referenced by the object.
On Behalf of the MySQL/ORACLE RE Team
Balasubramanian Kandasamy