From Bugzilla Helper: User-Agent: Mozilla/5.0 (Windows; U; Windows NT 5.0; en-US; rv:1.2b) Gecko/20020828 Description of problem: I installed Kernel 2.4.18-10 on a Redhat 7.3 system serving a large website. Apache 1.3.23-14, php-4.1.2-7.3.4. The system isn't able to server more than 5 requests per seconds. Tried to tune many kernel parameters but it didn't help. Going back to 2.4.18-3bigmem will serve about 1000 requests per second. I saw many timeouts to select() in strac'ing the apache root process. Version-Release number of selected component (if applicable): How reproducible: Always Steps to Reproduce: 1. Install 2.4.18-10 on a i686 system 2. 3. Additional info:
what kind of network card is this ?
[root@www ~]# lsmod Module Size Used by Not tainted ipchains 46184 0 autofs 12740 0 (autoclean) (unused) e100 77524 1 usb-ohci 22688 0 (unused) usbcore 77024 1 [usb-ohci] ext3 70720 6 jbd 53504 6 [ext3] aic7xxx 125728 7 sd_mod 12896 14 scsi_mod 115120 2 [aic7xxx sd_mod]
This now also happens with the 'old' kernel: Here are the results of # http_load -parallel 50 -rate 50 -seconds 60 urls where it fetches the index.html page. 1977 fetches, 1038 max parallel, 1.07245e+07 bytes, in 60.0051 seconds 5424.66 mean bytes/connection 32.9472 fetches/sec, 178727 bytes/sec msecs/connect: 6857.95 mean, 45008.7 max, 3.437 min msecs/first-response: 5345.73 mean, 32055 max, 5.173 min 1111 bad byte counts HTTP response codes: code 200 -- 866 Only 32 fetches per second, is this normal? [root@www ~]# uname -a Linux www.edonkey2000.com 2.4.18-3bigmem #1 SMP Thu Apr 18 07:17:10 EDT 2002 i686 unknown Is there anything I can do?
forgot to mention that the system the http_load is running on is connected via 100Mbit switch and using the eepro100 driver.
[root@www ~]# strace -p 846 select(0, NULL, NULL, NULL, {0, 280000}) = 0 (Timeout) time(NULL) = 1033070689 kill(2253, SIGALRM) = 0 kill(2320, SIGALRM) = 0 kill(2341, SIGALRM) = 0 kill(2367, SIGALRM) = 0 wait4(-1, 0xbffff6d8, WNOHANG, NULL) = 0 select(0, NULL, NULL, NULL, {1, 0}) = 0 (Timeout) time(NULL) = 1033070690 kill(2215, SIGALRM) = 0 kill(2219, SIGALRM) = 0 kill(2246, SIGALRM) = 0 wait4(-1, 0xbffff6d8, WNOHANG, NULL) = 0 select(0, NULL, NULL, NULL, {1, 0}) = 0 (Timeout) time(NULL) = 1033070691 kill(2131, SIGALRM) = 0 kill(2214, SIGALRM) = 0 kill(2366, SIGALRM) = 0 wait4(-1, 0xbffff6d8, WNOHANG, NULL) = 0 select(0, NULL, NULL, NULL, {1, 0}) = 0 (Timeout) time(NULL) = 1033070692 kill(2302, SIGALRM) = 0 kill(2310, SIGALRM) = 0 kill(2312, SIGALRM) = 0 kill(2374, SIGALRM) = 0 kill(2390, SIGALRM) = 0 wait4(-1, 0xbffff6d8, WNOHANG, NULL) = 0 select(0, NULL, NULL, NULL, {1, 0}) = 0 (Timeout) time(NULL) = 1033070693 kill(2129, SIGALRM) = 0 kill(2154, SIGALRM) = 0 kill(2193, SIGALRM) = 0 wait4(-1, 0xbffff6d8, WNOHANG, NULL) = 0 select(0, NULL, NULL, NULL, {1, 0}
any idea if this gets fixed when using the "e100" driver instead?
misunderstanding: www is running the e100 driver sda is running the eepro100 driver www is SLOW sda is normal (running Redhat 7.2 - 2.4.7-10) So the www server is suffering due to the usage of e100? Should we try to switch to eepro100?
could this be related to https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=73185
not really; the card in that bug is new and not quite on the market yet
Thanks for the bug report. However, Red Hat no longer maintains this version of the product. Please upgrade to the latest version and open a new bug if the problem persists. The Fedora Legacy project (http://fedoralegacy.org/) maintains some older releases, and if you believe this bug is interesting to them, please report the problem in the bug tracker at: http://bugzilla.fedora.us/