Functionality Added or Changed
A new Unreachable Agents section was added to the MySQL Instances dashboard. It is only visible if there are unreachable Agents. This section displays information about the unreachable Agents, and provides a Delete Agent action. (Bug #17662013)
mysql-identity-source Agent Installer
agent.sh option was added.
It determines how the unique identity for the MySQL instance is
default option uses either
the "server_uuid" variable if present, or generates a random new
one. Passing in
host_plus_datadir uses a hash
of the host identity and the path to the MySQL instance's data
For example, passing in
allows a MySQL instance to preserve its history in MySQL Enterprise Monitor if
it is rebuilt, as the host and datadir are the same, thus the
hash will be identical.
host_plus_datadir is not allowed when
the Agent is remote monitoring a MySQL instance, as MySQL Enterprise Monitor is
unable to definitively compute a known-unique host identity in
(Bug #17558993, Bug #70507)
--socket usage improvements with
agent.sh, the testing suite, and the Agent
Installer. Under certain conditions, using
--socket would fail when used with an
unattended installation, or with
When the service manager was unavailable, and when the HTTP threads were hung waiting for timeouts, the inflow of collected data could have exhausted the heap, which caused the Agent to run out of memory and shut down. (Bug #17790128)
Text on the InnoDB Buffer Pool Usage graph was updated, and now reads as:
It is not recommended to run this report on instances with large InnoDB buffer pools (over approximately 32 GB). Generating this data incurs a mutex lock for every 10,000 pages read within the buffer (in a loop), which can decrease throughput/concurrency of connected clients. If no data is returned within 2 minutes this report will automatically time out. (Bug #17669455, Bug #70729)
Added a better explanation for the case when Example Query is enabled but example queries are not available. The new text:
If you are using the Performance Schema, you need to adjust the server's configuration by running this command: UPDATE performance_schema.setup_consumers SET enabled = 'YES' WHERE name = 'events_statements_history_long'; To make this change permanent, adjust your /etc/my.cnf file (location may vary) to include the following line: performance_schema_consumer_events_statements_history_long = ON Please note that the query may also have been removed due to configured table size limitations, look for the performance_schema_events_statements_history_size system variable on the same file and adjust appropriately.
For more information, see Section 7.1.1, “Using the MySQL Performance Schema”. (Bug #17666148)
The binary log related advisors now check whether binary logging is enabled before performing related tests. Binary log events are no longer generated if this feature is disabled. (Bug #17697324)
The Upgrade Installer now checks for available disk space before
initiating an upgrade, and halts operation if it determines that
there is not enough. This required amount is logged as "Total
required space: XXXXMB" in
Disabling, and then enabling an Advisor would lose the custom configurations that were made to the Advisor. (Bug #17670151)
On Linux and Mac OS X, with the MySQL Enterprise Service Manager Installer, attempting to change the default ports to values such as 80/443 would fail, and report an error about not being able to bind to ports below 1024. (Bug #17611217)
On Solaris 10 with the Solaris Studio installed, the Installation would fail when attempting to use an existing MySQL Instance as the MySQL Enterprise Service Manager Repository Database. (Bug #17656476)
User authentication failures are handled more gracefully, as sometimes unhandled exceptions were thrown. (Bug #17648304)
The Bad Connections panel was improved. Error messages now show the actual error instead of masking them with simple titles like "General Error", the data of the last error was added, and now each error lists its own timestamp. (Bug #17632841, Bug #70670)
jsvc utility is now used to start/stop
Apache Tomcat. It provides general improvements, such as timing
the Apache Tomcat "stop" so it always removes the associated PID
(Bug #17595732, Bug #70600)
config.sh without any arguments
would generate a stack trace.
Unix domain sockets are now supported on Solaris SPARC systems.
--socket may now be passed to
agent.sh. The required
libjunixsocket library was missing.
The shutdown port,
18085, is no longer used
as MySQL Enterprise Service Manager now uses
Microsoft Windows, and
jsvc on Mac OS X and
Linux, to launch and control the Service Manager.
18085 was a locally bound port
mysqlmonitorctl.sh to stop and
start Apache Tomcat.
(Bug #17559033, Bug #70505)
With a large amount of assets (around 400+), the Asset
Selector would not load, as it could reach the
lz-string library is now used to manage
When there was large amount of replication topologies, adding more topologies would increase the CPU dramatically, and slow the application. (Bug #17512054)
--test-privileges option was added to
agent.sh. It works similarly to
--test-credentials with the exception that it
only tests the Administrative user, and it tests both its
privileges (to CREATE USER) and credentials. An error code of
4 is returned when the Administrative user is
CREATE USER privilege.
StandardSocketFactory was extended to
include IPv6 support.
All of the older MySQL Enterprise Backup rules and graphs were migrated to MySQL Enterprise Monitor 3.0 as a single MySQL Enterprise Backup Advisor with individual thresholds, and as a single Enterprise Backup Run Time graph that displays total and lock times.
A help tooltip was added to the Asset Selector search function. It notes that the search is case-insensitive, and that a search string must be 3+ characters in length.
Agent replication connection credential/privilege problems are now only logged once as a warning, and are logged as exceptions in debug mode. Before it could clutter the standard logs with exceptions. For example, a slave requires special permission to access a master's UUID.
After adding a MySQL instance to monitor with all wrong credentials, and editing it with the correct credentials (so it was monitored normally), repeating this process after removing the MySQL instance from being monitored would cause it to show up in the monitored MySQL instances table. In this case, it would show up as both monitored and unmonitored.
The MySQL Enterprise Plugin for Connector/J logged the wrong version and build number, as it reported 18.104.22.168 instead of 22.214.171.124.
New Advisors, which are added when MySQL Enterprise Service Manager is upgraded, are now auto-scheduled to start.
When a custom collection had both delta and regular variables, the rule evaluation would occur for both.