Hi Mark, Thank you very much for replying! I did open a bug for this last night after I e-mailed:
http://bugs.mysql.com/?id=25460 As for reproducing, we're working on that at the moment. This happened on a production system, so we worked first to stop the hemorrhaging. Currently, we moved the high volume tables to InnoDB as a stop gap. I suppose an important piece of information is that we're running MySQL on Solaris 10. We moved the DB from a SPARC box to Solaris 10 X64 box and it happened again. We're running on top of ZFS, and that appears to be sucking down a majority of the system RAM, so we're wondering if that is causing the problem. Though at the time of crash there's typically 500MB or so free out of 8GB. I was reading that whereas Linux will allow overcommittment of RAM, Solaris returns a NULL to a malloc when there's no more memory. Could this cause the behavior? We will turn on the core file. The funny thing is this happened again, and we didn't get any debugging information outside of the usual friendly message that MySQL had crashed and the output of a few memory-related my.cnf settings. This was despite having --with-debug on. We did run --with-debug=full on and had some serious performance issues. So we're going to try and repro on a dev system by exhausting its RAM. This database is stuffing e-mails into itself, so I'm wondering if it could be a strange character that's not properly escaped. We'll get y'all a core and a stack trace. Thank you again. Any help is very much appreciated. Best Regards, Jason On 1/8/07, Mark Leith <[EMAIL PROTECTED]> wrote:
Hi Jason, Jason J. W. Williams wrote: > Hello, > > We've been getting random crashes on our MySQL servers running MyISAM > tables for the last month, its gotten very bad in the last two weeks. > This has occurred on both 5.0.27, 5.1.11 and 5.1.15-nightly20070103. > > It crashes the tables with high queries per second. We've fixed the > issue on one of the servers by changing its tables to InnoDB. We can't > do that however on another server, which we turned debugging on > instead. It appears to be an assertion failure, the error message from > the MySQL debugging code is: > > Assertion failed: fixed == 1, file item.h, line 1601 > > > Any help is greatly appreciated. Should we report this as a bug? Any crashing is most certainly a bug, so if you could gather as much information on this as possible and report a bug that would be great. Please include: o The full section of the error log for the time of the crash o If there is a stacktrace reported, the resolved trace following: o http://dev.mysql.com/doc/refman/5.0/en/using-stack-trace.html o If you could turn on core files and upload the core file, and mysqld binary used to create it, as tar.gz to: o ftp://ftp.mysql.com/pub/mysql/upload o Link this in the bug report as well Do you have any way to reproduce this as yet? Cheers, Mark -- Mark Leith, Support Engineer MySQL AB, Worcester, England, www.mysql.com Are you MySQL certified? www.mysql.com/certification
-- MySQL General Mailing List For list archives: http://lists.mysql.com/mysql To unsubscribe: http://lists.mysql.com/[EMAIL PROTECTED]