Red Hat Bugzilla – Bug 53581
Stronghold / Apache secure apache/stronghold servers bring network service down under load
Last modified: 2007-04-18 12:37:03 EDT
From Bugzilla Helper: User-Agent: Mozilla/4.0 (compatible; MSIE 5.5; Windows NT 5.0) Description of problem: When I use a secure Strongohld web server on a 4 CPU IA64 machine under a peak load, eventually the machines network fxnality stops working. i.e. No ping, ftp, telnet, http, ..... I generate the load on the web server using Radviews web load application (which does not run on the server itself, it simulates web traffic). The problem can be fixed by simply restarting the network interface (ifdown/up), until it happens again. Version-Release number of selected component (if applicable): How reproducible: Always Steps to Reproduce: 1.Start a secure Stronghold web server. 2.Set up a load (I used webload) that drives the server at peak load. Peak load should be around 600 tps on a 4 CPU 800Mhz IA64 (Lion) system. My webload setup was requesting 20 byte web pages over ssl (https). Actual Results: The servers' network functionality will crash, but I've seen it take between 2 and 10 hours. Using the ifdown and ifup network scripts will restore network activity until the next time.... Expected Results: The network services not to stop working! Additional info: Stronghold I used is a prerelease of the next release of SH, but I've been informed that it is stable. Reference : Mark Cox of Redhat. Email : mjc@redhat.com. Please contact Mark in relation to this. My email is diarmuid@aep.ie. Distribution is RedHat 7.1. Here is the output from lsmod on the server machine... Module Size Used by autofs 29680 1 (autoclean) eepro100 45048 1 (autoclean) ipchains 98496 0 (unused) nls_iso8859-1 4672 1 (autoclean) nls_cp437 6176 1 (autoclean) vfat 25944 1 (autoclean) fat 81880 0 (autoclean) [vfat] usb-uhci 65184 0 (unused) usbcore 139680 1 [usb-uhci] qla1280 121816 3
I have verified that this problem also occurs when using a non secure web server (http requests) with the load also generated using the WebLoad application. It seems to manifest itself quicker like this (about 100 minutes to fail when I tested it). This is probably because there is higher network throughput when in non secure mode. (The test was doing around 3000 SSL transactions per second). I ran the "dmesg" command for the first time (after we lost network fxnality) and there was the following interesting messages there... eth0 card reports no resources eth0 : 0 multicast blocks dropped Again bringing down and up eth0 fixes the problem.
Which kernel is this, and which network card ?
The kernel is the 2.4.3-12 SMP kernel. I'm not sure what the NIC card is because it's built into the machine we're using. The machine is a 4cpu lion(?) which we received from Intel. Should I try and find details on the nic interface?
Thanks for the report. This is a mass bug update; since this release of Red Hat Linux is no longer supported, please either: a) try and reproduce the bug with a supported version of Red Hat Enterprise Linux or Fedora Core, and re-open this bug as appropriate after changing the Product field, or, b) if relevant, try and reproduce this bug using the current version of the upstream package, and report the bug upstream.