Hi, All
A few weeks ago, I tried to use mod_file_cache which tested under our
pressure test, however I met some strange behaviors. Now I can use ab
to reproduce it.
I'm using Apache2.2.0 on 4P Xeon(EM64T) and EL4U2 ( 2.6.9-22.ELsmp, 64bit).
Apache configuration:
<IfModule mod_file_cache.c>
mmapfile /usr/local/apache2/htdocs/index.html #just cache one file
</IfModule>
ThreadLimit 600
<IfModule worker.c>
StartServers 10
MaxClients 500000
ServerLimit 5000
MinSpareThreads 1000
MaxSpareThreads 2000
ThreadsPerChild 100
MaxRequestsPerChild 0
ListenBackLog 3000
</IfModule>
MaxRequestsPerChild 0
Now I test with "/usr/local/apache2/bin/ab -n 100000 -c 10
localhost/index.html", during the tests, some threads segmentation
fault.
I used gdb to get the backtrace, however looks like these segmentation
fault will happen in different places. Most are:
#0 0x0000003d26e2e989 in kill () from /lib64/tls/libc.so.6
#1 <signal handler called>
#2 0x00000000006fe818 in ?? ()
#3 0x000000000042d5fa in ap_rgetline_core (s=0x45007058, n=8192,
read=0x45007060, r=0x2ac4e2fd28, fold=0, bb=0x2ac4e31248) at
protocol.c:222
#4 0x000000000042dbd8 in ap_get_mime_headers_core (r=0x2ac4e2fd28,
bb=0x2ac4e31248) at protocol.c:681
#5 0x000000000042e4b2 in ap_read_request (conn=0x6e86c0) at protocol.c:909
#6 0x0000000000462540 in ap_process_http_connection (c=0x6e86c0) at
http_core.c:189
#7 0x0000000000440ac3 in ap_run_process_connection (c=0x6e86c0) at
connection.c:43
#8 0x000000000046d877 in worker_thread (thd=0x63cea8, dummy=Variable
"dummy" is not available.
) at worker.c:531
#9 0x0000003d279060aa in start_thread () from /lib64/tls/libpthread.so.0
#10 0x0000003d26ec5b43 in clone () from /lib64/tls/libc.so.6
#11 0x0000000000000000 in ?? ()
#0 0x0000003d26e2e989 in kill () from /lib64/tls/libc.so.6
#1 <signal handler called>
#2 0x0000000000443668 in add_any_filter_handle (frec=0x5b61a8, ctx=0x0,
r=0x7257e8, c=0x715990, r_filters=0x725a68, p_filters=0x725a78,
c_filters=0x715a08)
at util_filter.c:330
#3 0x000000000044390a in ap_add_input_filter_handle (f=Variable "f" is not
available.
) at util_filter.c:415
#4 0x000000000042e639 in ap_read_request (conn=0x715990) at protocol.c:1022
#5 0x0000000000462540 in ap_process_http_connection (c=0x715990) at
http_core.c:189
#6 0x0000000000440ac3 in ap_run_process_connection (c=0x715990) at
connection.c:43
#7 0x000000000046d877 in worker_thread (thd=0x63d208, dummy=Variable "dummy"
is not available.
) at worker.c:531
#8 0x0000003d279060aa in start_thread () from /lib64/tls/libpthread.so.0
#9 0x0000003d26ec5b43 in clone () from /lib64/tls/libc.so.6
#10 0x0000000000000000 in ?? ()
#0 0x0000003d26e2e989 in kill () from /lib64/tls/libc.so.6
#1 <signal handler called>
#2 ap_invoke_filter_init (filters=0x2c) at config.c:312
#3 0x000000000043a6f3 in ap_invoke_handler (r=0x2ac4e02978) at config.c:344
#4 0x0000000000464bb0 in ap_process_request (r=0x2ac4e02978) at
http_request.c:258
#5 0x00000000004625cd in ap_process_http_connection (c=0x7138b0) at
http_core.c:171
#6 0x0000000000440ac3 in ap_run_process_connection (c=0x7138b0) at
connection.c:43
#7 0x000000000046d877 in worker_thread (thd=0x63d598, dummy=Variable "dummy"
is not available.
) at worker.c:531
#8 0x0000003d279060aa in start_thread () from /lib64/tls/libpthread.so.0
#9 0x0000003d26ec5b43 in clone () from /lib64/tls/libc.so.6
#10 0x0000000000000000 in ?? ()
Are these threads segmentation fault due to over pressure? I also
tested cachefile directive and without mod_file_cache, both works OK.
Any ideas? Should I submit a bug request?
Thx, Xuekun
