> >> I have not completely figured out what the problem is and I never got a
> >> response from any one else on the list. I can tell you it looks like a
> >> mutex problem between two threads. We have followed it down into
> >> pthreads
> >> where it seems to be stuck. The external symptom we see is that
> >> the MySQL
> >> server does not open any of the databases.
> >>
> >> The really bad news is that it is not a MySQL problem so
> >> reinstalling MySQL
> >> does nothing to fix it. The worse news is that the MySQL
> server does not
> >> give you a clue what the problem is.
> We had a Linux guru look at the version of mysql on this system that we
> compiled with debug. I got information about the debugging session third
> hand and really don't have any details other than the problem appeared to
> be in pthreads. I am not so sure pthreads is at fault other than a
> dead-lock condition exists and it is not detecting it and blowing one of
> the stuck processes away.
What version of Linux are you running? After a lot of reearch and gdb'ing,
I found a bug report in bugzilla on redhat.com that seems to illustrate the
bug that we're having:
http://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=13785
In short, it seems it's a contention problem in pthread_lock. the fix is to
upgrade to libc 2.1.92 (or better). RedHat 6.2 is running libpthreads 0.8,
and I think this problem is fixed in 0.9. Unfortuantely, 0.9 needs libc
2.2, so you're probably better off upgrading to RedHat 7.2.
I know, this bug report applies to SMP kernels, but ieven if you're not
running an SMP kernel, it's probably a good idea to upgrade anyway.
Why it started suddenly and won't go away, I have no idea.
