As the stale status is not a real status bit and it's used to retain the
history of a master, there is no need to print it in any output. This
output will only confuse users now that the stale status will not be
cleared from masters that go down.
The log manager could send two messages if a log message was posted soon
before the log manager was stopped. This caused a debug assertion which
then manifested as a deadlock inside the log manager.
Make all modules lowercase and make module loading case
insensitive. Further, make command invocation case insensitive,
as far as the module name is conserned.
During switchover, MASTER_GTID_WAIT is now called on all slaves. This causes
switchover to complete slower than before but is safer if log_slave_updates
is not on on the new master server. Also, read_only is disabled on the
demoted server if waiting on slaves or promotion fails. This should
effectively cancel the failover for the old master.
The `purge` command can be used to reset the conversion
process. Currently, executing the `purge` module command and restarting
MaxScale is the only correct way to reset the conversion process.
Added more debug assertions to JSON serialization of the Avro schema. Also
checked that the column count for the TABLE_CREATE object matches that of
the TABLE_MAP object.
The test loads multiple modules in one call so we have to pre-load them
one by one to make sure that they are all present regardless of the
locations where the individual modules were built.
All modules now have an 8-bit range for capability flags. Currently only
the client side authenticator and protocol capability bits are loaded due
to the fact that backend versions of these modules don't relate to a
particular service.
A subset of the checks done at connection creation time need to be done at
query routing time. This guarantees that the connection is closed if the
server no longer qualifies as a valid candidate.
Added teset case that checks that a change in the replication topology
correctly breaks the connection.
When the unit tests were run without installing the libraries in their
final locations, the loading of the modules would fail. Using locations
relative to the build directory allows unit testing without having to
install the libraries.
The check for the current master backend did not see if the backend was
actually in use. This would be the case when the backend would be
originally in use but would be later disabled.
Since there is a concept called "listener" it is confusing that the
dcb "this" argument in some dcb functions is called 'listener' instead
of 'dcb' as it is called everywhere else.
Using the effective name means that a module command can be invoked
both using the deprecated name as well as the actual name. E.g.
both using mysqlmon and mariadbmon even though only the last one
actually exists as a module.
The trailing comment removal pattern unnecessarily required that a leading
space is present in all trailing comments.
Also, the pattern didn't match if no line ending was included in the SQL
statement. The subject ending should be the third valid terminator in
addition to UNIX and Windows style line endings.
If a shallow copy of the buffer is made, any modifications that are made
to the data after it has been queued will affect the queued query of the
LocalClient.
A copy-on-write mechanism would save the relatively expensive process of
copying the data but since the LocalClient is not often used, it is not
the most critical performance problem.
'mysqlmon' is still accepted but 'mariadbmon' is loaded instead.
This is done at runtime instead of e.g. by using a symbolic link,
so that a warning can be logged.
The warning is logged and the translation of the module name is
made by the code that loads the modules so that it's easy to do
the same thing for other modules as well.
In a subsequent commit the documentation is updated.
Change the ordering of the two flushes such that FLUSH LOGS comes last.
This seems to make sure gtid:s are updated to newest values before
the MASTER_GTID_WAIT-call. Without this fix, switchover does complete
succesfully, but some of the slaves may not be able to replicate due to
not having same events as new master. Exact reason for this still unclear.
Pre-loading users for all threads at startup significantly reduces the
chance for failures caused by the lazy initialization of the user database
done by the authenticators.
If users are not loaded at startup and the connection limit for all
servers is reached, authentication in MaxScale will fail not due to too
many connections but due to the lack of authentication data. This causes
repeated reloading of users, which floods the log with messages, and
unnecessary stress on the cluster itself.
For some reason, the source code of mysqlmon was split into C and C++
sources. This caused problems by effectively discarding all changes from
2.1 that are merged into 2.2.
This commit merges the changes into the correct file that were added to
the wrong file.
When a client connects to MaxScale and authentication fails, an error
about hostname resolution is logged. This happens because the
authentication first tries to resolve the address as an IP address, then
an IPv6-mapped-IPv4 address and finally as a hostname. If users have not
been loaded, the authentication is guaranteed to fail on the first attempt
due to the lazy loading of users.
The default value for strict_multi_stmt prevents compound statements and
atomic multi-statement commands from being executed without completely
disabling load balancing. As the new default value will have no practical
effect on all correct uses of readwritesplit, this is a relatively safe
thing to change.
As the object name formatting is automated, so should the formatting of
object names in arguments. This also makes 2.2 backwards compatible with
2.1 commands.
