maria-developers team mailing list archive

[Patrick Crews <gleebix@xxxxxxxxx>]

Hello,
I have attached a proposed patch file for the MariaDB test suite.  These are
related to pool of threads.

pool_of_threads.test - Nothing major here.  Added some extra comments,
output to .result file (for easier tying .test to .result), and disconnects
of the test connections (We have found in MySQL that while the connections
will go away, there is a little bit of time where the leftover connections
can cause subsequent tests to fail - they show up in a 'connection audit'
and throw off counts, etc).

connect.test - Added a test to ensure that we can only have
max-extra-connections + 1 SUPER user.  Previously, the test used only the
'root' account and did not ensure it was +1 SUPER.  I created a
limited-access user and tried to connect via extra-port after
max-extra-connections were used.  As expected, this connection is not
successful, but the root account CAN connect.  Just wanted this in place to
ensure the proper behavior is tested and won't change (test will fail if it
does).

variables.test - Added some tests of the pool of threads related variables.
 Ensure that the read-only ones can't be changed via SQL and that the others
can be.

Two questions:
1)  I performed a test where I tried to set the main port and extra-port to
the same value.  It seems extra-port was able to claim the port# first and
the server died as the value was already in-use.  Is this correct behavior?
 I know nobody *should* do this, but it raises the following extra
questions:
   a. Should we let 'port' take precedence on claims over 'extra-port' (ie,
the server starts as expected, but throws a warning about extra-port being
an in-use value)?
   b. Should we throw a more informative error message?  It is pretty easy
to troubleshoot, but I just wanted to raise the question directly and not
make assumptions

2)  Do you guys want a test of the behavior described in #1?

3)  Finally, I tried setting extra-max-connections to 0.  It defaults to 1.
 Since the behavior of extra-port is <extra-max-connections> regular
connections + 1 SUPER user, should we let 0 be an acceptable value?  A super
user would/should still be able to connect, but an admin might not want to
allow other connections via extra-port.

Thanks,
Patrick

Attachment: pcrews_patch.dat
Description: Binary data