I sent this to php.dev and it didn't come through, so I'm sending it again to the mailing list this time. Sorry if this is a repeat.
We have a bunch of apache servers here with php compiled in.
A bunch of them have died in two different rounds, both rounds being on
the same weekend. The servers on ports 8080 and 443 are fine; it's only
the port 80 apaches that are dying.
I did a backtrace with gdb, and I get thousands of these:
#60961 0xef7d46e8 in ?? ()
#60962 0xef7dfcc8 in ?? ()
#60963 0xef7d2d8c in ?? ()
#60964 0xef24c980 in _PROCEDURE_LINKAGE_TABLE_ () from /usr/lib/libthread.so.1
#60965 0xef239308 in kill () from /usr/lib/libthread.so.1
#60966 0x1516dc in sig_coredump ()
#60967 0xef23b928 in __sighndlr () from /usr/lib/libthread.so.1
#60968 <signal handler called>
#60969 dbexit () at dblib.c:561
#60970 0xbb7a8 in zm_shutdown_sybase (type=-282964972,
module_number=-277216068) at php_sybase_db.c:254
It repeats that same pattern again and again.
It seems to me that either there's a weird recursive bug in apache with
php, or someone has found a DOS attack against apache with php by
pelting it with a bad signal. Or maybe there's a gdb bug, but... nah. :)
Any suggestions? Restarting apaches gets old fast. It's been so easy
to take apache's reliability for granted in the past.
TIA.
BTW, thanks for the work you guys do on apache/php.
--
Dan Stromberg UCI/NACS/DCS
msg39918/pgp00000.pgp
Description: PGP signature
