Don't mean to be a nuisance but does anyone have any advice here? File mutex didn't solve the problem.
I'd rather not bug tech@ about it but this is important to me. Some more information... this system runs GENERIC.MP (4 CPU's) and has 4GB RAM. Apache segfaulted chrooted and non-chrooted but it usually only happened when the load was high. Some posts suggested php5 could be to blame as well. Am I alone w.r.t the segfaults (and occasionally Bus Errors) that I get under load? I am building another box with 3.8 and php 4.4.1 so we'll see in a few days if the issues remain. I'm also running memtest86 on both boxes just to rule out memory issues. So far, no errors on the machine in question... Thanks. On 11/8/05, Need Coffee <[EMAIL PROTECTED]> wrote: > Hi. > > I have an amd64/3.7-stable machine here running apache with mod_ssl > and php 5.0.4. SSLMutex is set to 'sem', the default. Intermittently > the httpds start segfaulting. > Of course the parent process remains, and respawns them. There's a > large number of clients > and the machine is kept pretty busy. > > The httpds always end the same way: > > 26510 httpd CALL semop(0x70000,0x7f7ffffeced0,0x1) > 26510 httpd RET semop -1 errno 13 Permission denied > 26510 httpd CALL semop(0x70000,0x7f7ffffeced0,0x1) > 26510 httpd RET semop -1 errno 13 Permission denied > 26510 httpd CALL semop(0x70000,0x7f7ffffeced0,0x1) > 26510 httpd RET semop -1 errno 13 Permission denied > 26510 httpd PSIG SIGSEGV SIG_DFL code 1 addr=0x2f2e2e trapno=6 > > From what I've read, semume and semaem values that are too low can > cause this. I've cranked semume to 100 but haven't messed with > semaem since that's actually hardcoded in sys/sem.h. > > Should I be messing with semaem or is there something else at work > here? I'm tempted to try a file mutex but saw a post (which I only > vaguely remember) where henning@ said it was defaulted to 'sem' for > permissions reasons and that file wouldn't work right. > > I'm at my wits end with this... anyone have a solution? Thanks much in > advance.
