Hello there list! I'm having some trouble with my web server running Apache and MySQL. I'm accessing the DB through the latest DBI:DBD and Perl. I was running 2.22.32 quite successfully for about a year until usage starting becoming a problem. I upgraded to 2.23.36 and the problem stayed the same. Basically, the server will cruise along at 50 queries a second without problems. Top's load level hovers between 2.xx and 3.xx, so it's not overly loaded. When the number of users increases until we hit about 100 queries per second, I get a kernel panic. It claims that it was unable to handle a kernel paging request and I've got to do a hard reboot. All I'm running on the server is Apache, Perl and MySQL. It's an SMP P-III 550x2 with 512MB ECC SDRAM. The memory usage is very low. It never goes above 160MB used -- even at the point when the kernel dies. I've replaced the memory and motherboard -- suspecting either a bad chip or bad memory controller, but it had no affect. When the number of queries reaches 100/sec my load level in top reaches between 10.xx and 12.xx, so there's a lot of stuff being done on the server, certainly. I was just writing to see if i was simply over-processing the server. The site doesn't seem slow per-sec, it's definately tolerable. And we only reach these conditions during extreme peak times of the day, so I don't really want to replicate my DB and send the non-updating queries to another system if I don't have to. My question is: Is it possible that these kernel faults are the result of process context-switching in Linux? If so, would an update from the 2.2.12smp kernel to 2.2.16smp help? or maybe 2.4 -- although I don't trust it for production. What about HDD corruption (I'm not page swapping as a result of low memory) Is it possible that all of the crashing has corrupted the HD and now everything is just hosed? In which case a re-install of Linux using the same kernel should fix the problem. I'm just trying to decide what my best option is to solve the problem. Re-install the OS, run Solaris instead of Linux, find a non-crashing MySQL version, get faster processors, replicate the DB, etc.. I guess I could also move to a PHP or JSP solution, as Perl may be causing the problem. But seeing as my CPU's able to route 100 queries per second to MySQL, I think they're being processed fast enough, just not reliably. Sorry about the open ended post. I really don't want to start an OS war or anything. I'm just trying to figure out if the kernel fault is due to a MySQL problem, an OS problem or a hardware problem. I hope I can get some good feedback to direct my next move. Thanks so much Mike InsideCorner.com
