Returning the results of a query as a set of packets is currently more
efficient. This is mainly due to the fact that each individual packet for
single packet routing is allocated from the heap which causes a
significant loss in performance.
Took the new capability into use in readwritesplit and modified the
reply_is_complete function to work with non-contiguous results.
The backend protocol command tracking didn't check whether the session was
the dummy session. The DCB's session is always set to this value when it
is put into the persistent pool.
8 + 4 + 4 ensures 16 with 8 byte alignment, which means that
'data' is certain to be 8 byte aligned. 4 + 8 + 4 might result
in something else in some funky environment.
The GWBUF shared buffer and its data is now allocated in one
chunk so that the data directly follows the shared buffer.
That way, creating a GWBUF will involve 2 and not 3 calls to
malloc and freeing one will involve 2 and not 3 calls to free.
The result collection did not reset properly when a non-resultset was
returned for a request. As collected result need to be distinguishable
from single packet responses, a new buffer type was added.
The new buffer type is used by readwritesplit which uses result collection
for preparation of prepared statements.
Moved the current command tracking to the RWBackend class as the command
tracked by the protocol is can change before a response to the executed
command is received.
Removed a false debug assertion in the mxs_mysql_extract_ps_response
function that was triggered when a very large prepared statement response
was processed in multiple parts.
Inlined the getter/setter type functions that are often used. Profiling
shows that inlining the RWBackend get/set functions for the reply state
manipulation reduces the relative cost of the function to acceptable
levels. Inlining the Backend state function did not have as large an
effect but it appears contribute a slight performance boost.
The parameter handling for monitors can now be done in a consistent manner
by establishing a rule that the monitor owns the parameter object as long
as it is running. This will allow parameters to be added and removed
safely both from outside and inside monitors.
Currently this functionality is only used by mysqlmon to disable failover
after an attempt to perform a failover has failed.
KILL commands are now sent to the backends in an asynchronous manner. As
the LocalClient class is used to connect to the servers, this will cause
an extra connection to be created on top of the original connections
created by the session.
If the user does not have the permissions to execute the KILL, the error
message is currently lost. This could be solved by adding a "result
handler" into the LocalClient class which is called with the result.
When the LocalClient is constructed, it is possible to extract all the
needed information at that time. The only obstacle is the fact that the
LocalClient is constructed at the same time the session is. Since the
client DCB is created before the session, it is safe to extract the shared
data directly from it.
The total timeout for the retrying of interrupted queries can now be
configured with the `query_retry_timeout` parameter. It controls the total
timeout in seconds that the query can take.
The actual connection, read and write timeouts of the connector aren't a
good configuration value to use for abstracted queries as the time that it
takes to execute a query can be composed of both connections, reads and
writes. This is caused by the usage of MYSQL_OPT_RECONNECT that hides the
fact that the connector reconnects to the server when a query is
attempted.
The new `query_retries` parameter controls how many times an interrupted
query is retried. This retrying of interrupted queries will reduce the
rate of false positives that MaxScale monitors detect.
The stack traces weren't logged as the LOG_ALERT priority wasn't enabled
by default. As an alert is intended to be something that must leave a
trace somewhere, and as such, it must not be possible to disable it. For
this reason, it is acceptable to always log the message if the priority is
LOG_ALERT.
Added the -rdynamic linker flag so that all symbols are exported when
linking MaxScale.
As the stack trace is printed in a signal handler, the first attempt
should be to print the stack trace to the standard output. This way the
output is printed before an attempt to use malloc is made when it is
logged to the logfile.
Since the module command interface was expanded to include a JSON output
parameter, there is no longer a need for an output DCB. As the JSON can be
printed by both maxadmin and the REST API, this allows the removal of
explicit output formatting in module commands.
The `failover` and `failover_timeout` parameters are now declared as a
part of the mysqlmon module. Changed the implementation of the failover
function so that the dependencies on the monitor struct can be removed or
moved into parameters.
Split the state change processing and failover handling into two separate
functions and added a call to the failover function into mysqlmon. This
prevents unintended behavior when failover is enabled for non-mysqlmon
monitors. The parameter itself still needs to be moved into mysqlmon.
Moved the failover documentation to the mysqlmon documentation as it is
specific to this monitor.
The failover command is simulated by executing a call to /usr/bin/echo
with all possible monitor parameters. This allows testing of the failover
mechanism without actually using the failover command.
The timestamp of the last change from passive to active is now
tracked. This, with the timestamps of the last master_down and master_up
events, allows detection of cases when MaxScale was failed over but the
failover was not done.
Currently, only a warning is logged if no new master has appeared within
90 seconds of a master_down event and MaxScale was set to active from
passive.
The last event and when the event was triggered is now shown for all
servers. The latest change from passive to active is also shown.
When an event occurs on a server, it is now stored so that the last event
for each server is known. This allows a state change to trigger an event
even if, at the time of the event, no action was taken.
This change is only cosmetic as no functionality is implemented.
The `passive` parameter can be given in the configuration file or on the
command line. It is displayed in the diagnostic output and changes to it
are persisted.
The variable itself does nothing.
When readwritesplit is routing any queued queries, the currently executed
command of the protocol modules needs to be adjusted by
readwritesplit. This is not a true fix but more of a workaround to fix the
problems of queued query execution.
The correct solution would be to move the queued query handling into the
client protocol module so that all components see the same state.
The enums exposed by the connector are not intended to be used by the
users of the library. The fact that the protocol, and other, modules used
it was in violation of how the library is intended to be used.
Adding an internal mapping into MaxScale also removes some of the
dependencies that the core has on the connector.
Cleaned up the MaxScale version of the mysql.h header by removing all
unused includes. This revealed a large amount of dependencies on these
removed includes in other files which needed to be fixed.
Also sorted all includes in changed files by type and alphabetical
order. Removed explicit revision history from modified files.
Updated the Connector-C to 3.0.2.
Removed the `mysql_com.h` header from `maxscale/protocol/mysql.h` as it is
already included by `mysql.h` and including it causes errors with
Connector-C 3.0.2.
Organized includes in `maxscale/protocol/mysql.h`.
Basically it would be trivial to report far more operations
explicitly, but for the fact that the values in qc_query_op_t
currently, quite unnecessarily, form a bitmask.
In 2.2 that is no longer the case, so other operations will be
added there.
When a COM_CHANGE_USER was executed, it is possible that the server
responds with a AuthSwitchRequest packet instead of an OK packet. In this
case, the server sends a new scramble which must be used to create the 20
byte hash that is expected as the response.
Added missing SSL parameters to servers resource output as well as added
the processing of these parameters when servers are created. It is
preferable to define servers as either encrypter or plain and to prevent
the modification of this at runtime.
