Some people have had problems with the binary of 3.23.5x. We had that problem, and when we went back to a self-compiled 3.23.46 things worked normally again.
The bad behavior looked like normal operation for anywhere from two hours to five days, followed by a CPU meltdown with loads over 200, which would typically settle back down by itself after ten or fifteen minutes of (seemingly) nothing progressing. The MySQL guys have said they think it was a version of glibc that they linked with; compiling your own from source may fix the problem. I'd be interested to know if it does. OTOH, in the five or six years I've used MySQL, this is the first time behavior like that was the fault of the server---it's almost always something in our code instead. InnoDB is supposed to reduce pathalogical behavior in a running system, so you might try that. The above-described problem with the MySQL binaries happened with either table type. Good luck, --Pete On Wed, Sep 25, 2002 at 09:24:02AM -0700, Andrew Maltsev wrote: > On Wed, Sep 25, 2002 at 10:02:52AM -0500, Philip Molter wrote: > > On Tue, Sep 24, 2002 at 08:21:36PM -0700, Andrew Maltsev wrote: > > : Any suggestions how to approach the problem? How and what to test? It > > : happens randomly, can work for a day or two with no problems and then > > : hang three times in one hour. And obviously I can't reproduce it in my > > : test environment however hard I stress test it. > > > > When it happens, what does the system look like? Is CPU pegged? > > Is MySQL using a lot of CPU. What does iostat tell you in terms > > of drive activity. Are the drives actively seeking or does the > > system seem relatively quiet? You'll have to do *some* troubleshooting. > > Sad thing is the system usually gets rebooted by support personel before > I can get my hands on it because long downtime is not acceptable. I've > seen it first hand only once and had just about a minute to look at it. > > It is not swapping, the system has adequate amount of memory and > generally there is no significant disk activity, no disk bound processes > at all. Show processlist responds with one process being in writing to > network state and others sleeping. But killing it does not work and even > killing the MySQL itself with SIGTERM does not work either, it has to be > killed with -9 and then some indexes are corrupted and need myisamchk. > > Well, I got my answer I guess -- it was a long shot, I kind of hoped > that somebody would respond with something like "I get the same damn > thing with 3.23.52 all the time, try x.xx.xx". As it is not the case > I'll have to try to investigate it myself :) > > Andrew. > > --------------------------------------------------------------------- > Before posting, please check: > http://www.mysql.com/manual.php (the manual) > http://lists.mysql.com/ (the list archive) > > To request this thread, e-mail <[EMAIL PROTECTED]> > To unsubscribe, e-mail <[EMAIL PROTECTED]> > Trouble unsubscribing? Try: http://lists.mysql.com/php/unsubscribe.php > --------------------------------------------------------------------- Before posting, please check: http://www.mysql.com/manual.php (the manual) http://lists.mysql.com/ (the list archive) To request this thread, e-mail <[EMAIL PROTECTED]> To unsubscribe, e-mail <[EMAIL PROTECTED]> Trouble unsubscribing? Try: http://lists.mysql.com/php/unsubscribe.php
