The test fails since slaves are constantly going out-of-sync and cannot rejoin
the cluster. After three failovers, the cluster has three standalone server and
the monitor gets confused which server is the master. For now, limit the test to
two failovers.
The test expects the monitor to respond within 5 seconds. This is not
guaranteed to happen so waiting for a monitor interval instead of a
hard-coded duration provides for a more stable test.
If the test would otherwise succeed but a coredump is found when the logs
are copied in the destructor, the test would pass. To prevent this, the
destructor should always exit with a non-zero status when it detects an
error.
Streamlined the test to perform as much testing as fast as possible. The
gradual ramp up did not provide any concrete benefits compared to testing
everything at once.
Replaced structures with C++11 alternatives where possible and removed
unused, redundant or dead code.
Before the state of the backend servers is checked, MaxScale needs to be
stopped to prevent the automated failover from interfering in the start-up
process.
Removed the excessive comments in favor of a simplified description. Use
stack-allocated TestConnections and simplify assertions.
The main change is the different SQL used to update the user with the old
password. Direct modification of the `mysql`.`user` database isn't very
neat but it guarantees that the value is updated.
The test should stop MaxScale at the start unless the manual debug flag is
given on the command line. This fixes the connection failure of mxs1719
but reveals a problem with the filter itself.
When a test is checking the status of the nodes, the output is relatively
verbose.
Also changed dropping of users to use the IF EXISTS syntax. This will
remove the errors if the users do not exist.
Don't test failover functionality when it is not needed. The bug is only
about the extra events that appear when a master is demoted and a slave is
promoted.
