Instead of continuously updating the status of the server object,
the monitor should update the pending_status of the monitor object
and then only at the end update the status of the server object.
Instead of continuously updating the status of the server object,
the monitor should update the pending_status of the monitor object
and then only at the end update the status of the server object.
All monitors (but for MariaDBMon) now handle connection errors in
the same manner.
An exception is MMMon that modifies both the server status and the
monitor status. However, in fact, all other monitors do it the wrong
way and update the server status directly instead of only the monitor
status that then at the end of the loop should be copied to the server
status.
All monitors (but possible for MariaDBMon) now handle the
connecting to the server in the same way, that is, if the
connecting fails, then the status is updated and we return.
This makes the code clearer and reduces race conditions, as the monitor
could be writing SERVER->status while a router is reading it. Also,
the time during which the SERVER struct is locked drops to a fraction.
The configuring of the monitor instance is now performed in a
separate function. That is in preparation for the moving of the
start function to maxscale::MonitorInstance.
- All monitors (but MariaDBMon for the time being) inherit
from that.
- All common member variables moved to that class. Still
manipulated in derived classes.
In subsequent commits common functionality will be moved to
that class.
If the feature is enabled (default off), at the end of a monitor loop
(once server states are known), read_only is enabled on slaves servers
without it.
The logic was weird, as the permission checking function assumes a disconnected
server as fine. The checking is now done when starting the main loop and lacking
grants print errors but does not stop the monitor.
The parameter extraction caused a recursive lock of the server
spinlock. To work around this, an unlocked version of server_get_parameter
is needed.
Ideally, a lock-free setup would be used but due to this being a bug fix,
it will have to be done later on.
Since monitors are now freed at MaxScale exit, the server data should be freed. Also,
gtid domain variables are now initialized with a common constant.
