On Wed, 27 Jun 2001, Justin Erenkrantz wrote:
| -1 (yeah, yeah, non-binding).
|
| Just compiled it with a Mandrake 8.0 box with GCC 3.0. The threaded MPM
| looks completely hosed.
|
| It will start up all of the children, but they are all dead within a
| few seconds. (Only two processes remain...)
I'm seeing the same on Solaris 8 now.
the children start, stick around for a few seconds, and then all but one
die from receiving SIGTERM. Any HTTP requests made to the server connect,
but hang after "GET /blah ..." is issued.
A quick truss on a child process shows that it seems to be
(gracefully?) killing itself:
/4: getpid() = 21853 [21848]
/2: signotifywait() = 15
/4: kill(21853, SIGTERM) = 0
/2: lwp_mutex_lock(0xFF1555D8) = 0
/2: lwp_mutex_wakeup(0xFF1555D8) = 0
/4: lwp_mutex_lock(0xFF1555D8) = 0
/2: lwp_mutex_lock(0xFF1555D8) = 0
/4: lwp_mutex_wakeup(0xFF1555D8) = 0
/4: lwp_mutex_lock(0xFF1555D8) = 0
/2: lwp_mutex_lock(0xFF1555D8) = 0
/4: lwp_mutex_wakeup(0xFF1555D8) = 0
/2: lwp_cond_signal(0xFF1555C8) = 0
/2: lwp_mutex_wakeup(0xFF1555D8) = 0
/3: lwp_cond_wait(0xFF1555C8, 0xFF1555D8, 0xFF105CA8) = 0
/3: lwp_mutex_wakeup(0xFF1555D8) = 0
/4: lwp_mutex_lock(0xFF1555D8) = 0
/4: time() = 993709078
/3: lwp_mutex_lock(0xFF1555D8) = 0
/3: Received signal #15, SIGTERM, in lwp_sigtimedwait() [caught]
/3: siginfo: SIGTERM pid=21853 uid=60001
/3: lwp_sigtimedwait(0xFFBEF7F4, 0x00000000, 0xFF14F860, 0xFFBEF824) =
15
/3: munmap(0xFDC00000, 1056768) = 0
/3: munmap(0xFD70E000, 1056768) = 0
/3: munmap(0xFD60C000, 1056768) = 0
/3: munmap(0xFD50A000, 1056768) = 0
/3: munmap(0xFD408000, 1056768) = 0
/3: munmap(0xFD306000, 1056768) = 0
/3: munmap(0xFD204000, 1056768) = 0
/3: munmap(0xFD102000, 1056768) = 0
/3: munmap(0xFD000000, 1056768) = 0
/3: munmap(0xFCB0E000, 1056768) = 0
/3: munmap(0xFCA0C000, 1056768) = 0
/3: munmap(0xFC90A000, 1056768) = 0
/3: munmap(0xFC808000, 1056768) = 0
/3: llseek(0, 0, SEEK_CUR) = 0
/3: _exit(0)
HTH,
/dale
