1

I have the following VirtualHost configured (Apache 2.2.22)

<VirtualHost *:80> ServerName ************ ProxyRequests Off <Proxy *> Order deny,allow Allow from all </Proxy> ErrorLog /var/log/apache2/dalo-lt_error_log LogLevel warn TransferLog /var/log/apache2/dalo-lt_access_log ProxyPass / ajp://localhost:8009/ ProxyPassReverse / ajp://localhost:8009/ <Location /> Order allow,deny Allow from all </Location> </VirtualHost>

With the following in my Tomcat's server.xml (7.0.39)

<Connector port="8009" protocol="AJP/1.3" redirectPort="8443" connectionTimeout="60000" maxConnections="36864" maxThreads="600" /> <Connector port="8080" protocol="HTTP/1.1" connectionTimeout="20000" maxConnections="36864" maxThreads="600" redirectPort="8443" />

When I run my stress test through the vhost:80 + ajp, I start seeing dropped connections after a certain point. However, if i run the test directly through tomcat's http connector (:8080) I see no dropped connections whatsoever.

What is the best way to determine what is causing this?

Just for completeness, here are some settings from my apache2.conf

<IfModule mpm_prefork_module> StartServers 10 MinSpareServers 10 MaxSpareServers 20 MaxClients 500 MaxRequestsPerChild 0 </IfModule> <IfModule mpm_worker_module> StartServers 4 MinSpareThreads 50 MaxSpareThreads 450 ThreadLimit 256 ThreadsPerChild 150 MaxClients 300 MaxRequestsPerChild 0 </IfModule> <IfModule mpm_event_module> StartServers 4 MinSpareThreads 50 MaxSpareThreads 450 ThreadLimit 256 ThreadsPerChild 150 MaxClients 300 MaxRequestsPerChild 0 </IfModule>

Update: Requesting server-status?auto yields the following while during the phase of the load test when requests are being dropped - so it looks to me like there are enough workers etc. to spare

Improve this question

1 Answer 1

Reset to default
1

Apache web server will start droping connections if it is getting too many request. You could use mod_status in order to look at the scoreboard, if it is full, you can increase the number of workers. (should be investigated if the increased number of workers fits your environment)

Improve this answer
8
  • Thanks for the suggestion. Unfortunately the dropped connections start occurring before the scoreboard fills up - in fact there is a lot of room. Commented Feb 11, 2015 at 13:50
  • haven't you seen any problem in ajp log file? (timeout, etc ..) Commented Feb 11, 2015 at 14:51
  • I don't have a dedicated ajp log file - how can I set that up? Commented Feb 11, 2015 at 15:04
  • I'd suggest mod_log_forensic: httpd.apache.org/docs/2.2/mod/mod_log_forensic.html Commented Feb 11, 2015 at 15:19
  • Thanks - though I've tried log_forensic and didn't find anything in the output specifically related to AJP / proxy requests. Only request headers for everything coming into the system. Commented Feb 11, 2015 at 15:33

You must log in to answer this question.

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.