MaxScaleUptime() renamed to maxscale_uptime() and moved from
gateway.c (MaxScale main) to misc.c, which is included in the
maxscale_common library. That way the symbol will be available
at link-time and will thus not prevent the use of the linker
flags -Wl,-z,defs (resolve all symbols at link-time) when
linking a module that uses maxscale_uptime().
Due to an include conflict between /usr/include/poll.h and
maxscale/server/include/poll.h the latter was moved to
maxscale/server/include/maxscale/poll.h.
Once 1.4 is out, all maxscale header files will be moved to that
same sub-directory. That will prevent include conflicts and also
make it easy to see which include files belong to maxscale and
which do not.
Both the passwords and temporary files of the embedded library were stored
in the same directory. Now the directories are separated and the embedded
library uses the temporary directory. The datadir cleanup also now only
cleans up the temporary data directory.
Since maxscale-common is linked to the connector-c library
and no longer to the MySql embedded library, mysql_library_init
does not need any parameters.
There can be any sort of library behind qc_init. Hence the arguments
cannot be hardwired to be like the embedded library wants them.
Eventually it might make sense to allow passing arguments from
maxscale.cnf.
Now, qc_mysqlembedded is linked against MySQL's embedded library,
and MaxScale itself against Connector-C.
So, in order to build MaxScale, Connector-C must be installed.
This has been tested with Connector-C 2.2.1.
The build variable MYSQLCLIENT_LIBRARIES is no longer used.
The query_classifier library is now only a wrapper that loads an
actual query classifier implementation. Currently it is hardwired
to load qc_mysqlembedded, which implements the query classifier
API using MySQL embedded.
This will be changed, so that the library to load is specified
when qc_init() is called. That will then allow the query classifier
to be specified in the config file.
Currently there seems to be a conflict between the mysql_library_end()
call made in qc_mysqlembedded and the mysql_library_end() call made in
gateway.c. The reason is that they both finalize a shared library.
For the time being mysql_library_end() is not called in gateway.c.
This problem is likely to go away by switching from the client
library to the connector-c library.
The polling statistics collection used atomic_add to increment values. This
is not an optimal way to update statistical values. Moved to per thread
values which are summed up when they are read.
Moved the functions used to gather polling statistics to their own file and
created a specific data type for statistics.
The THREAD type was not used everywhere and pthread_t was used instead.
The thread creation function also returned the address of a stack allocated
value which isn't guaranteed to be usable.
Only the query classifier needs the functionality of the embedded
server, while the rest of MaxScale is content with the client
library or Connector/C.
This have now been rearranged so that query-classifier links with
the embedded static library and then explicitly exports its own
functions using the query_classifier.map linker script. That way
query classifier will use the embedded library, while the rest of
maxscale use the client library, and this without conflicts.
Currently, query_classifier is not linked to maxscale-common,
but executables must link to maxscale-common and query_classifier.
Earlier thread initialization was done in poll_waitevents(),
which was somewhat confusing, since the main thread also calls
into poll_waitevents().
Now there is a specific thread "main" function, which first
performs thread specific initialization and then calls into
poll_waitevents().
The manpage was out of date and was missing a BUGS section. The `--help` output
of MaxScale wrongly stated that the default is to log to shared memory as did
the documentation.
Currently the initialization function only prepares PCRE2 patterns for use.
Added the call to the new utils_init() function to MaxScale and all relevant
tests.
The replace_quoted now uses a shared PCRE2 pattern which should remove some
of the overhead of the function.
Earlier, by default, the error and message logfiles were written to
the filesystem and trace and debug logfiles to shared memory. Now,
with just one log-file the default must be the file-system. However,
if info and debug messages are logged, then the filesystem will
become a bottle-neck.
A reasonable approach is then as follows (in the config file)
syslog=true
maxlog=false
log_to_shm=true
With this set, the maxlog file will be created to shared memory, but
nothing will be written to it, since it is disabled. However, if there
is a need to investigate something, then a dba can from maxadmin turn
on maxlog logging and also enable info and debug messages. That is, it
will be possible to enable debugging output without restarting maxscale.
Incidentally, the way the config file and command line arguments are
handled should be rewritten. Currently, it is a mess.
There were some variance regarding the way the MaxScale log (i.e.
the file log) was called; "maxlog" in configuration file and
"maxscalelog" at the command line and maxadmin interface. Now it
is uniformly referred to as "maxlog" in the configuration file, at
the command line, from maxadmin and in the code.
Since localtime is not thread-safe it should not be used in multithreaded
contexts. For this reason all calls to localtime were changed to localtime_r
in code where concurrency issues were possible.
Internal tests were left unchanged because they aren't multithreaded.
Claiming that the loading of maxscale.cnf failed in case of any
error was misleading. Maxscale may not succeed in opening it,
reading it or processing it.
LOGIF and skygw_write_log removed from server/core/*.c and
replaced with calls to MXS_(ERROR|WARNING|NOTICE|INFO|DEBUG).
This is a mechanism change, no updating of the actual message
has been performed.
Currently this causes a very small performance hit, since the
check whether the priority is enabled or not is performed in
the function that is called and not before the function is called.
Once all LOGIFs and skygw_write_logs have been replaced, the
behaviour will be altered back to what it was.
skygw_logmanager_init renamed to mxs_log_init and skygw_logmanager_done
renamed to mxs_log_finish. skygw_logmanager_exit removed alltogether as
all it did was to call skygw_logmanager_done. That appears to have been
a source for confusion as in many places a call to skygw_logmanager_done
was followed by a call to skygw_logmanager_exit. In addition, the function
skygw_log_done was removed from the header, since it lacked an
implementation.
With only one log-file no arguments are needed. The maxadmin command
'flush log' still accepts all the previous arguments, but warns about
them being deprecated.
Earlier, the global setting for the syslog decided whether syslog
was enabled when skygw_logmanager_init was called, but not whether
logging to syslog actually was made.
Now syslog logging is enabled by default and the global setting
decides whether or not syslog logging actually is made. That is,
this opens up the possiblity for making it possible to turn on
and off sysloging at runtime.
Further, although the API led you to believe otherwise, it was
hardwired that LOGFILE_ERROR and LOGFILE_MESSAGE messages were
written to syslog.
The changed removed the need for passing an argv array explicitly.
The syslog ident must be provided explicitly when calling
skygw_logmanager_init (and not provided via the argv array).
It can be NULL, in which case it automatically will be the program
name.
The openlog() call is now always made, irrespective of what the
value of the global syslog flag is. That way it will be possible
to turn syslog logging on or off after the fact.
Whether the log-file should be written to the filesystem or to
shared memory must now be explicitly defined when calling
skygw_logmanager_init() (instead of passing that via the argc/argv
construct).
Also, the meaning of '-l' when invoking maxscale has been changed.
Earlier -l [file|shm] specified whether the trace and debug logs
should be written to shared memory (while the error and message
logs always were written to the filesystem) and the _default_
was to write them to shared memory.
Now, with only one file, '-l' has still the same meaning, but it
decides whether the one and only logfile should be written to shared
memory, or the filesystem and the _default_ is to write it to the
filesystem.
The previous interface of skygw_logmanager_init was conceptually
broken. With -o you could specify that logging should be done to
stdout. However, even if you did that, the log manager still checked
that the logging directory could be accessed. Unless it had been
specified using -j <path> the default was /var/log/maxscale.
That is, unless the program calling skygw_logmanager_init was invoked
by a user that had write access to /var/log/maxscale, there would be
a complaint even if nothing was ever written to that directory.
In practice this meant that even if -o was used you had to provide
a -j with a path that surely is writeable (e.g. "/tmp").
This has now been changed so that you explicitly must provide the
log directory and the flags -j and -o are removed.
bool skygw_logmanager_init(const char* logdir, int argc, char* argv[]);
If /logdir/ is provided then logged messages are written to a log file
in that directory. If /logdir/ is NULL then messages are logged to stdout
and no checks for access to any directory is not made.
The log manager variables lm_enabled_log_files_bitmask, log_ses_count
and tls_log_info that earlier were declared separately in every
c-file are now declared in the log_manager.h header.
There were many combinations of tab-width used so making the
indentation look right in the editor wasn't really possible.
The changes made:
- All tabs replaced with spaces.
- Indentation depth 4.
- , followed by space.
- Most binary operators (*, =, -) surrounded by one space.
- No space following ( or before ).
- Keywords follwed by 1 space.
The log manager is initialized only once and skygw_log_sync_all now checks if the log manager has been successfully started before interacting with the log manager
