When LEAST_BEHIND_MASTER routing criteria was used, the info level logging
function would fall through to the default case. In debug builds, this
would trigger a debug assertion.
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 GLIBC backtrace functionality doesn't generate file names and line
numbers in the generated stacktrace. This can to be done manually by
executing a set of system commands.
Conceptually doing non-signal-safe operations in a signal handler is very
wrong but as stacktraces are only printed when something has gone horribly
wrong, there is no real need to worry about making things worse.
As a safeguard for fatal errors while the stacktrace is being generated,
it is first dumped into the standard error output of the process. This
will function even if malloc is corrupted.
When the router requires statement based output, the gathering of complete
packets can be skipped as the process of splitting the complete packets
into individual packets implies that only complete packets are handled.
Also added a quicker check for stored protocol commands than a call to
protocol_get_srv_command.
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.
If a query was processed in the client protocol module when a prepared
statement was being executed by the backend module, the current command
would get overwritten. This caused a debug assertion in readwritesplit to
trigger as the result was neither a single packet nor a collected result.
The RCAP_TYPE_STMT_INPUT capability guarantees that a buffer contains a
complete packet. This information can be used to track the currently
executed command based on the buffer contents which allows asynchronicity
betweent the client and backend protocol. In practice this only comes in
play when routers queue queries for later execution.
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 multi-statement detection did not check for the existence of
semicolons before doing the heavier processing.
Calculcate the packet length only once for the result state management.
Replace the original version of the function with the reference version
and use it everywhere. Added runtime assertions to check that an invalid
DCB is never processed.
As the DCB passed as the clientReply parameter is guaranteed to match one
of the DCBs in the RWBackends. By using a reference, the need to copy a
shared_ptr is removed (along with the atomic operation that it implies)
thus reducing the overhead in the clientReply and the functions it uses.
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 result processing code did unnecessary work to confirm that the result
buffers are contiguous. The code also assumed that multiple packets can be
routed at the same time when in fact only one contiguous result packet is
returned at a time.
By assuming that the buffers are contiguous and contain only one packet,
most of the copying and buffer manipulation can be avoided.
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.
If a switchover_script parameter is given, its value will be used as
the switchover script. Otherwise the default will be used. Currently
just echo.
The MySQL Monitor now introduces two script variables, CURRENT_MASTER
and NEW_MASTER, that contain information about the current and new
master respectively.
Switchover is performed only if switchover has been enabled and MaxScale
is *not* in passive mode.
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.
To be able to do that, we need to get hold of the MXS_MONITORED_SERVER
corresponding to the SERVER specified as the new master.
So, instead of just return a boolean indicating whether the server was
found or not we return the MXS_MONITORED_SERVER pointer.
