If a listener is defined in a static configuration file, it can now be
destroyed at runtime. If MaxScale is restarted, the listener will be
created again unless the configuration file is modified.
If a value that is not of key-value format was given to the create
commands, the code previously threw an exception. This causes tests to
misbehave so changing it to a rejected promise is more appropriate.
The configuration used an invalid parameter name which caused the two
MaxScales to use the same directory for persisted configurations.
The test also tried to sync with the wrong server causing the same
listener to be created twice on the same port but on different MaxScales.
Cleaned up the startup and shutdown scripts to always start from a clean
state.
Each `show` type command that takes a resource name now also has a version
that prints all resources of that type.
Added test cases for newly added commands.
Test that `list servers` works with both monitored and unmonitored
servers. The test doesn't check that the GTID is present as the test setup
uses file and position based replication.
The enabled admins for the Linux users were shown as basic users. This was
caused by the separation of the two admin types.
Added tests that check that enabled Linux accounts show the correct type
in the output.
The type of the created user can now be specified with the --type option.
Expanded tests that cover the user creation. Also added a test case that
checks that basic users are only allowed to read through the REST API.
Only the protocol, port and address of the listener were used to check if
a listener exists. The check should also use the name of the listener to
be sure that each name is unique.
Expanded tests to check that the creation of duplicate listeners is
detected. Did minor improvements to related test code.
The `cluster diff` command is now tested by doing add/alter/remove
operations on servers. As the same methods are used for all objects, they
should detect all sorts of modifications between two Maxscales.
The `execute` command now returns the output of the command instead of
printint it. This allows the tests to actually test the output of the
commands instead manually verifying that it is correct. It also allows the
library part to be used as an actual library that only returns data.
The monitors should only be reused if they have the same name and they use
the same module. This way the only difference is in configuration.
Fixed MaxCtrl detection of bad options and altered monitor creation test
to expect correct results. Also improved some of the error messages.
The test tests the internals of the cluster diff generation
algorithms. This should guarantee that the diff calculation and the
resulting synchronization will work as expected without having to test it
on a live cluster.
All errors are now returned as rejected Promises. This will make the error
reporting more consistent with the tested output as well as the actual
output.
Added more tests for error cases and fixed minor argument processing bugs.
The tests are now sorted by command type. This allows for more consistent
test creation as tests can be done per source code file instead of per
object type.
Added helper functions for common testing operations. The two main ones
are doCommand, which executes a MaxCtrl command and returns a result, and
verifyCommand, that executes a command and verifies the result via the
REST API.
Making the core a module proved to bring more problems than it solved. For
the sake of simplicity in installation and code coverage reporting, the
core is now completely contained in the `lib/` directory.
Also added preliminary coverage reporting with nyc.
As the REST API is a part of the core, it is more appropriate for the
tests to reside there as well. Further refactoring of the testing needs to
be done to allow multiple components to use the same framework but with
different tests.
Before creating the users, the master will do a RESET MASTER to remove the
automatically generated binlogs. The servers should now start a lot faster
since they no longer need to replicate the large amounts of data that were
in these generated binlogs.
The service and server tests were deleting the relationships object and
posting a body without it instead of posting an empty relationships object
to delete relationships.
