On Thu, 2009-04-23 at 11:25 +0200, André Warnier wrote: > Craig Main wrote: > > On Thu, 2009-04-23 at 10:07 +0200, André Warnier wrote: > > > >> Craig Main wrote: > >>> Thanks Andre, > >>> > >>> I am using apache to connect to 10 tomcat instances using the ajp > >>> connector with load balancing and sticky sessions. > >>> > >>> Apache: > >>> Load [r...@lt-webportal-01:~] # /opt/apache/bin/apachectl -V > >>> Server version: Apache/2.2.9 (Unix) > >>> Server built: Sep 6 2008 15:33:44 > >>> Server's Module Magic Number: 20051115:15 > >>> Server loaded: APR 1.3.0, APR-Util 1.3.0 > >>> Compiled using: APR 1.3.0, APR-Util 1.3.0 > >>> Architecture: 64-bit > >>> Server MPM: Prefork > >>> threaded: no > >>> forked: yes (variable process count) > >>> Server compiled with.... > >>> -D APACHE_MPM_DIR="server/mpm/prefork" > >>> -D APR_HAS_SENDFILE > >>> -D APR_HAS_MMAP > >>> -D APR_HAVE_IPV6 (IPv4-mapped addresses enabled) > >>> -D APR_USE_SYSVSEM_SERIALIZE > >>> -D APR_USE_PTHREAD_SERIALIZE > >>> -D SINGLE_LISTEN_UNSERIALIZED_ACCEPT > >>> -D APR_HAS_OTHER_CHILD > >>> -D AP_HAVE_RELIABLE_PIPED_LOGS > >>> -D DYNAMIC_MODULE_LIMIT=128 > >>> -D HTTPD_ROOT="/opt/http-2.2.9" > >>> -D SUEXEC_BIN="/opt/http-2.2.9/bin/suexec" > >>> -D DEFAULT_PIDLOG="logs/httpd.pid" > >>> -D DEFAULT_SCOREBOARD="logs/apache_runtime_status" > >>> -D DEFAULT_LOCKFILE="logs/accept.lock" > >>> -D DEFAULT_ERRORLOG="logs/error_log" > >>> -D AP_TYPES_CONFIG_FILE="/opt/etc/httpd/mime.types" > >>> -D SERVER_CONFIG_FILE="/opt/etc/httpd/httpd.conf" > >>> > >>> Tomcat: apache-tomcat-6.0.18 > >>> > >>> Platform: CentOS 5.2 64bit > >>> > >>> > >>> The problem points to the ajp connection as we have done the following > >>> tests: > >>> > >>> 1) Load tested http with our static content. - Good Response times > >>> 2) Load tested with the tomcat ajp load balancing to 10 tomcat > >>> instances: > >>> * When we test the site's static content, the speed is fine > >>> * When we test the site's pages that go to tomcat, we see the > >>> same page returning sometimes fast, and other times slow. > >>> * If we test connecting to the tomcat via it's http connector > >>> from the command line, the speed is fine under load. > >>> * If we test connecting to the apache requesting the tomcat > >>> pages, some of the requests are returned quickly, others are slow. > >>> 3) At first we thought it could be an specific instance of Tomcat, but > >>> on investigation, one request to a particular instance was fast, and > >>> others we slow to the same instance. > >>> > >>> Please let me know if you need more information. > >>> > >> I believe that will do fine. > >> ;-) > >> > >> Just another question, because someone is bound to ask : would it be an > >> option for you to test with mod_jk instead of mod_proxy_ajp as a > >> connector on the httpd side, and compare the results ? > >> Not to say that one or the other is better, but mod_jk has been there > >> longer, and is probably more mature. And it would help circumscribing > >> the issue. > >> > > > > This is what we are currently looking into. Will let you know when we > > have results on that. > > > Right. > If you have questions on mod_jk later on, it might be better to move > this discussion to the Tomcat users list. The developers of mod_jk > generally follow this httpd list also, but tend to be more often on the > Tomcat side, because officially mod_jk is under the Tomcat umbrella. > I suggest you then copy this whole message, since it has all the info. >
Once we have some info on this, I will. > One more question still : are you using any automated way to collect > your performance info, or just visual impressions ? > It is always better to have some real numbers, and logfiles to compare, > before changing the configuration. > There are a number of tools available for the purpose, but if you have > perl installed on a system, I can give you a simple tool that would > provide some scripted accesses and timestamped response times. > We are using HP Performance centre to do the load testing, and that gives us some response times. While that is running, we manually access the website either by command line or browser. This is where we see the difference. I'll see what kind of stats I can post from those. > > --------------------------------------------------------------------- > The official User-To-User support forum of the Apache HTTP Server Project. > See <URL:http://httpd.apache.org/userslist.html> for more info. > To unsubscribe, e-mail: users-unsubscr...@httpd.apache.org > " from the digest: users-digest-unsubscr...@httpd.apache.org > For additional commands, e-mail: users-h...@httpd.apache.org > To read FirstRand Bank's Disclaimer for this email click on the following address or copy into your Internet browser: https://www.fnb.co.za/disclaimer.html If you are unable to access the Disclaimer, send a blank e-mail to firstrandbankdisclai...@fnb.co.za and we will send you a copy of the Disclaimer.
