When backend authentication failed due to errors other than wrong
credentials, the users were unconditionally reloaded. This caused a spike
of activity whenever authentication failed for other reasons.
Also fixed the test that checks for this to look for the correct error
message.
Added some code to detect server states in a consistent manner and created
the test. The multi-source replication appeared to cause problems for the
test system which needs to be resolved.
Added --force flags to most direct `mysql` calls to prevent errors from
stopping the processing of remaining commands.
The test now skips execution of inserts to separate nodes. This should
guarantee that the nodes are synced at all times.
The test now does the inserts and selects inside a transaction to force
the reads to the same node where the inserts are done.
The tests now print a stack trace to standard output if it receives a
fatal signal. This will help debug test failures caused by broken tests
and unexpected results.
The mm test was on the heavy side and the workload can be reduced for the
sake of testing only the functionality.
The mm_mysqlmon test did not check the initial state of the
server. Although not necessary, adding it will help detect test failures
caused by broken replication.
mxs1457_ignore_deleted did not stop the monitors before breaking the
replication and attempting to use MaxScale. As the test expects
authentication to work regardless of the actual state of the servers, the
monitors need to be disabled before a query is attempted.
The ssh_node function now supports printf style arguments. This is used to
simplify command execution on the nodes.
Curreltny, in addition to its old usage, it is used to drop extra
databases when replication is started.
The test now checks that both users with passwords and users without
passwords work through MaxScale when skip_authentication is enabled. In
addition to this, the test also makes sure that the wrong password and a
missing password are properly detected.
The result of the authentication should be ignored but the scramble that
is calculated as a side-effect still needs to be stored. This can be done
by altering the SQL used to get the matching row to only match on the
username, not the network address.
Also expanded the test case to cover the use of bad credentials.
The configuration for mxs1323_retry_read now uses shorter timeouts for
monitors. This should help the monitors detect the server failures before
the result of the SELECT returns. also increased the time the query sleeps
before returning.
The removal of a server from a service is intended to affect only new
sessions.
Added a test that checks that the connections are kept open even if the
server is removed from the service.
The test attempted to copy the generated CSV file from the MaxScale server
when in fact it was on the master server. Removed HEAVY label from the
test as it takes about 10 seconds to run.
Cherry-picked from commit f80cde2af149fda9c47d8f9131f82527b76af487.
The replication needs to be stopped before the binlogrouter is started. If
the replication is stopped after this, it is possible that two servers
with the same value of server_id attempt to register as slaves which
causes the later of them to fail.
The replication_manager is only designed for systems that have yum
installed which means it will always fail on non-RHEL/CentOS systems.
The query threads in mxs1323_stress were not checking whether the test had
ended while they were executing the queries. This caused test timeouts as
the queries can take a relatively long time.
The tests should not drop the test database and then recreate it. This
adds an unnecessary burden on replication which will cause false positives
when replication is not fast enough.
This is commit `2410c23e6664bef7e449dfd5a767121f309559fe` from develop
cherry-picked to 2.1.
When a backend is waiting for a response but no statement is stored for
the session, the buffer where the stored statement is copied is not
modified. This means that it needs to be initialized to a NULL value.
Added a test that checks that the behavior works as expected even with
persistent connections. A second test reproduces the crash by executing
parallel SET commands while slaves are blocked.
There is still a behavioral problem in readwritesplit. If a session
command is being executed and it fails on a slave, an error is sent to the
client. In this case it would not be necessary to close the session if the
master is still alive.
As the test does a lot of large inserts, the timeouts can be triggered at
times when the test is not actually stuck but just waiting for the insert
to complete. Increasing the timeout from 10 to 30 seconds should give it
enough time to finish processing.
When an internal connection is created, the SQL_MODE of the connection
should be set to a known default. The empty SQL_MODE allows consistent
functionality for all backend server versions.
The avrorouter failed to detect ALTER TABLE statements which caused a
regression. Extended the alter table tests to parse the JSON for more
strict validation of test results.
The test dropped the 'test' database which is heavily used both for
checking functionality and running tests. A better alternative is to use a
custom database that is only used by this test.
The masking filter now has a default fill value, so that if a
provided value does not match, then "X" is used.
The tests were modified to use "Y" as an explicitly provided
fill value (to distinguish from the default "X") and the results
were the default fill value would kick in were modified accordingly.
The timeout is hit very often if the cluster fixing takes some time. This
causes false positives so the timeout should be stopped before attempting
to fix the cluster.
The test seems to fail when executed outside of a debugger. A delay
between creating the connections and checking the connection count appears
to fix the problem.
Refactored the test to use stack allocated objects and favored assertion
style functions over conditional clauses with failing assertions inside
them.
