Instead of looking at the server version, the actual error message should
be inspected. This guarantees that the correct error message is logged
even with custom builds.
The intention was to send the lowest backend version string automatically
to the client instead of the default handshake version. This did not work
as the service version string was used instead of the server version.
When a connection to a server is lost and the session command history is
disabled, the session will continue as long as at least one connection is
open. Previously the open connection calculation used the same code that
was used when a new session was created which only inspected the
configured server count instead of the actual open connection count.
The table creation was not detected as the function used to extract the
table name did not return the fully qualified names. Even if it did return
a fully qualified name, it wouldn't have been correctly processed.
The test did not use the wait_for_monitor function to sync with the
monitor. This function speeds up the testing greatly by removing
unnecessary sleeps from it.
Also reduced the amount of data inserted into the cluster. There's no real
need to test with large amounts of data as it is only a functional test.
When a read-only transaction fails due to a connection error, no message
would be logged. Also added an info level message for the case when a
backend connection would get closed before the session is in the correct
state and a debug assertion that the router session should never be closed
when the handleError method is called.
If the log file is successfully opened, both stdout and stderr are
redirected to it. This helps catch ASAN reports without having to read the
system journal files.
As the output is redirected to a file, some of the output was made visible
only in non-daemon mode. This helps keep the log file clean and readable.
The client connections had the Nagle algorithm enabled which could cause
bad performance with smaller workloads. The common network configuration
code in utils.cc, currently used by the backend connections, sets it
properly.
Cherry-picking the commit from develop since it's also needed for 2.2.
FindASAN did not look for so version 5 and TSAN referred to ASAN. Disable
leak checks for unit tests as some objects always leak, namely servers.
SELECT...FOR UPDATE locks the rows for update, but only if
autocommit==0 or a transaction is active, so in principle even if
it were classified as READ it'd still be sent to master when it
actually matters.
However, even if autocommit==1 and/or no transaction is active, a
slave in read only mode will reject the statement if the user is
subject to the read only restriction (a user with super privileges
is not), which might be considered a server bug. By classifying the
statement as a write, it'll be sent to master and always succeed.
By default ASAN only reports the errors and doesn't cause the program to
crash. By forcing a SIGABRT, the process will generate a core dump which
causes the test to fail.
Although the act of placing environment variables inside the systemd
service file might seem intrusive, they have no effect on non-ASAN
builds. This appears to be the most convenient and straightforward way of
changing ASAN behavior for testing.
The debug assertion is wrong as the code was changed to prioritize hints
over the router target selection. Also removed the superficial check for
master, slave and relay master states as they are implied by the fact that
the connection is in use.
The collection of resultsets needs to be disabled by default when a
response is received to cover the cases where an error is returned.
The collection of results should also not be set for queries that do not
generate any responses.
When the 10.2 users query is executed with a MariaDB server older than
10.2.11, the query will fail due to inadequate grants on the 'users' table
generated as a part of the CTE. To work around the issue, a SELECT grant
on the whole mysql database is required. Logging the server where the
query fails also helps resolve the problem by pointing out where the grant
needs to be added.
Removed unnecessary sections and updated some of the text to be more
specific. Expanded the explanations on where the replication starts and
how the avrorouter needs to be configured. Added example output from the
cdc.py client and removed the relatively useless maxavrocheck section.
