This commit refactors slave selection. The compare is still pair-wise but isolated into a small run_comparison() function. The function get_slave_candidate() is used when new connections are created, which I both moved and modified (had to move due to scoping), so diff is off.
The slave selection for routing: get_slave_backend() now has the filtering logic from old get_slave_backend() and compare_backends(), the latter of which is removed.
Backend functions mostly take shared_ptr<SRWBackend> in various forms (as is, const ref, in a container). Ideally the shared_ptr would be used only to where it is really needed, and either naked ptrs or references to RWBackend would be used. This refactor does not address that issue, but compounds it by using even deeper shared_ptr structures. Fixing that in a future commit.
The main piece of code, slave selection (backend_cmp_response_time), uses the available
method of pair-wise comparison of slaves. This will be changed to selection using all
available slaves, along with removal of hard coded values.
This is to support calculating the average from a session, and the slave selection criteria to be able to route based on averages. This commit, like the next one, have TODOs which you should feel free to comment on. Undecided things.
If secondary masters are defined, MASTER_USE_GTID=Slave_pos must
be specified for the default master.
If MASTER_USE_GTID is specified, only Slave_pos is allowed as
value.
The variable storing the configuration index renamed from
'current_config' to 'config_index'.
In the code the same terminology is used as in the documentation.
"Default" is is used for referring to the default connections
(earlier "primary" was used) and the secondary connections are
referred to as "secondary" (earlier "alternative" was also used).
- When specifying a configuration, if it already exists use
the existing configuration as the default, thus allowing you
to change only a specific configuration value.
- When specifying a new alternative configuration, use the
primary (aka first) configuration as the base.
- If MASTER_HOST is changed on the primary (first, nameless)
configuration, then all configurations are erased.
- At "SLAVE START", always start from the first configuration.
If the replication fails using the current config, we retry
immediately using another config, without waiting anything at
all.
Only when we have unsuccessfully tried with all servers, will
we wait a while before starting again.
0 cannot be used to indicate failure as 0 will be the waiting
period if we switch to an alternative server.
Also fix some minor issues; failure to initialize all members
and inverted success-check.
Separating the reading of the config from the process of
applying it allows us to use configs around and later apply
them (e.g. when we decide its time to switch master server).
At save time, assert that the configuration as dispersed around
blr and as stored in the config object are identical.
Later its the state from the config object that will be saved.
Alternative masters can now be specified like
CHANGE MASTER TO ...
CHANGE MASTER ":2" TO ...
CHANGE MASTER ":3" TO ...
Now only the data is stored, but it is neither used nor
saved.
This function applies a change master config. Currently is
used from blr_handle_change_master(), and subsequently will
be used when BLR switches to an alternative master.
This type will be used for storing the configuration
specified with CHANGE MASTER at a point where it has been
verified, to the extent possible, that the provided options
are valid.
In order to make it visible that the types contain C++
members, the following types were renamed:
CHANGE_MASTER_OPTIONS -> ChangeMasterOptions
MASTER_SERVER_CNF -> MasterServerConfig
Functions where also changed to take const refs were appropriate.
In order support the possiblity for having multiple alternative
masters for the binlog server, we need to have multiple configs
around.
Originally the config values were stored as 'char *':s, which
would have made the lifetime management of the strings laborious
and error prone.
Now, the options are stored as std::string:s, which makes the
lifetime management a non-issue.
When receiving a command, such as
CHANGE MASTER TO MASTER_HOST = '127.0.0.1' ...
when the value is extracted, the quotes are removed before the value
is returned.
Firstly, BLR does not anyway anywhere enforce that quotes are present
where they should be and, secondly, BLR currently does de-quoting all
over the place and with this change, all those places can be simplified.
The configuration updating in readwritesplit was the inspiration for the
mxs::rworker_local type. Due to this, taking it into use simply means that
the type changes from Config to mxs::rworker_local<Config>.
Given that worker.hh was public, it made sense to make routingworker.hh
public as well. This removes the need to include private headers in
modules and allows C++ constructs to be used in C++ code when previously
only the C API was available.
By storing the data gathere by readwritesplit inside the session, the
protocol will be aware of the state of the LOAD DATA LOCAL INFILE
execution. This prevents misinterpretation of the data which previously
led to closed connections, effectively rendering LOAD DATA LOCAL INFILE
unusable.
This change is a temporary solution to a problem that needs to be solved
at the protocol level. The changes required to implement this are too big
to add into a bug fix release.
The master reconnection in the binlogrouter used the housekeeper
recursively. This is no longer supported and the delayed worker call
mechanism should be used in its stead.
The maxscale_is_shutting_down function is used to detect when MaxScale
should stop. This fixes a race condition in the code where the workers has
not yet been initialized but a termination signal has been received. It
also replaces the misuse of the service_should_stop variable with a proper
function.
