> From: Vivek Khera <khera@stripped>
> "KM" == Ken Menzel <kenm@stripped> writes:
> KM> you guy's only don't really have the resources or want to support BSDi it
> KM> would have been nice to know that up front, before we had a problem. I
> I've used MySQL on BSD/OS 3.1 thru 4.0.1 with ZERO problems for the
> last two years. We had some issues with earlier MySQL releases, but
It works fine indeed - however there's a caveat, and I've been able
to reproduce the problem here. As long as you're making lots of
quick or small queries everything's fine.
Long running queries monopolize 'mysqld' to the point where new
connections can not be made to the server. In essence once a long
query starts it runs to completion and nothing else (new mysql
connections and other queries) goes on.
Create a simple table (just an 'int' field will do), populate it with
4 or 5 million rows. Then while you're running 'avg()' or 'sum()' or
something over that table try to initiate another mysql connection.
It'll sit there until the long running query completes.
I think it's a threads related problem but the test program I came up
with (reading a file in one thread while reading/writing stdin/stdout
in another) stumbled across a different threads bug (busily being worked
on by BSDi). What's needed I think is a better (but still simple)
test program that can be used to debug the problem (rather than
trying to debug something as complex as the entire MySQL server).
> I see no reason to continue paying for support if they will not
> provide it on my production platform. Somehow it seems odd to me that
> they won't fix it, but I guess things change.
I think they might fix it but not instantaneously - at this point I
suspect a threads problem (since the problem does not occur on
FreeBSD 5.0-current). MySQL's big contribution at this point would
be a test case that could be submitted to support@bsdi