MaxScale can now start without any defined monitors. This allows the core
services to be configured beforehand. With the changes to dynamic
modifications to servers, automatic scaling of slaves is possible.
The `add server` command accepts a server name and either a service name
or a monitor name. It will add servers to services and monitors. Since all
monitors use the MONITOR_SERVER structures directly, the monitors need
to be stopped before new servers are added to them
The debugcmd parameter processing didn't actually allow the maximum number
of parameters to be passed to the function. The detailed help text was
never printed and most of them were only duplicates of the short
description.
Before a query is routed to a backend, the status of the server reference
is checked. This allows the servers that are removed from a service to be
ejected from the list of active servers for active sessions.
The createInstace, newSession, closeSession and freeSession functions were
cleaned up and reorganized to be a bit clearer for the reader. Removed
unnecessary comments and replaced them with ones that explain what's
happening in the code.
Removed unused linked lists from both sessions and instances and replaced
them with better alternatives. This should improve performance since new
session don't have to acquire the instance level lock to put themselves
into the session list.
Removed the use of the BACKEND structure and replaced it with the use of
the SERVER_REF structure of the service. This allows dynamic changes to be
made to the list of servers.
Cleaned up parts of the code and removed obsolete or useless
functions. The schemarouter module could do with some refactoring since it
derives from readwritesplit.
The BACKEND structure in readconnroute is now replaced with the use of the
SERVER_REF structure of the service. This allows dynamic changes to the
list of servers to be made.
The readwritesplit now understands that the amount of servers can change
and some of the items in the list of server references aren't in use. This
allows dynamic changes to the number of servers used by readwritesplit.
Servers can now be added and removed from services which allows routers to
use them with new sessions. The routers don't fully use the new
functionality in the server references which prevents new servers from
being taken into use.
If shutdown has been initiated (via maxadmin or by sending a
SIGINT or SIGTERM) and a SIGTERM is received, the process
is terminated.
If shutdown has been initiated (via maxadmin or by sending a
SIGINT or SIGTERM) and a SIGINT (Ctrl-C) is received, a warning
is printed that shutdown is in progess. Then, if an additional
SIGINT is received, the process is terminated.
So, in practice:
- If MaxScale is running as a daemon, the first SIGTERM initiates
shutdown and a second one unconditionally terminates the process.
- If MaxScale is running in the console, the first Ctrl-C initiates
shutdown, the second prints a warning and the third terminates
the process.
This is to ensure that MaxScale can be forced to exit, in
case some thread is hung for whatever reason, and is thus preventing
the controlled shutdown.
The hangup code was refactored into a common function which should only be
used after the server states have been updated. This will remove erroneus
connections to already failed servers.
The master reference used by the readwritesplit sessions needs to be
reassigned if slave reconnection occurs. This happens because the
reference refers to a certain place in the backend reference array
instead of the actual backend reference and those places are mixed
when the array is sorted.
The qlafilter now has an option to log all messages to a single
file instead of session-specific files. Session ids are printed
to the beginning of the line when using this mode. Documentation
updated to match. Also, added an option to flush after every
write.
When persistent connections were used, it was possible that the injection
of COM_CHANGE_USER statements caused a crash when a DCB in the wrong state
was accessed.
For MySQL protocol modules, the `data` member of the client DCB points to
the shared session data, a MYSQL_session struct, but for sessions in the
persistent pool, it points to NULL. The boolean, `was_persistent`, tells
whether a DCB was just taken from the pool or it has been in use.
The `was_persistent` status wasn't properly reset for connections that
were put into the pool which caused a COM_CHANGE_USER statement to be
injected for stale connections in the pool which caused a crash when the
NULL `data` member was accessed.
The error logging is now more detailed and tells why the connection is
being closed. This should help the user figure out what is happening when
write fails and the connection is closed.
Some of the master server status checks didn't check whether the server
was actually running. The macros in server.h should always be used instead
of manually inspecting the server status.
If a readwritesplit session is active, it should never connect to a new
master. This will lead to unexpected results as the session states aren't
consistent.
If an illegal DCB close is done with a backend DCB, it will log the server
where it was connected. This allows us to know whether the DCB was
connected to a master or a slave.
Added more debug assertions to readwritesplit code. The DCBs should never
enter the DCB_STATE_DISCONNECTED.
Removed useless debug log messages. The messages usually just flood the
logs with no use to the developers.
The filter can detect SERVER_MORE_RESULTS_EXIST which means the server
is sending more result sets: example:
DROP PROCEDURE IF EXISTS multi;
DELIMITER $$
CREATE PROCEDURE multi() BEGIN
SELECT 1;
SELECT id FROM t2 limit 40;
set @a=4;
SELECT 2;
END$$
DELIMITER ;
MySQL> call multi()
A debug assertion failed due to a NULL buffer but a non-zero packet
length. This was caused by a missing reset of the packet length after
freeing the buffers.
If a Galera node has a nonpositive priority, the node will never be chosen
as the master. This gives the user more control over how the master is
chosen.
Some error messages were logged at INFO level and some had conditions that
prevent the logging. Removed these restrictions that an error situation is
always logged.
The luafilter didn't use a format string with dcb_printf which can lead to
unexpected results if the returned string contains printf special
characters.
If a master_failure_mode was set to error_on_write, a reconnection to the
old master would happen after the following events:
- Master server fails and the connection is closed
- The master server recovers
- A slave fails and the connection is closed
- A replacement for the slave is searched
If these events took place, the master would be taken back into use with
an inconsistent session state.
