I am going to reply to my own query so that if someone comes across this they will at least see some kind of answers…. I never got confirmation that this particular configuration was correct, incorrect or could be improved. I do know that we decided to go with it one more time with the following modifications (pure guesswork on my part): Threadlimit 3000 ThreadsPerChild 3000 MaxRequestsPerChild 0 KeepAlive On KeepAliveTimeout 15 AcceptFilter https none AcceptFilter http none EnableSendfile Off EnableMMAP off ProxyRequests Off SSLStrictSNIVHostCheck off ProxyBadHeader Ignore HostnameLookups off ExtendedStatus off We experienced no performance problems with this config so we are running in production with it. I still have some reverse proxy questions that I’ll post up shortly but the main question, which was how to have good performance in a reverse proxy environment has been answered by trial and error. Pete From: Pete Helgren [mailto:Pete@xxxxxxxxxxxxxxxxxxxx] We have implemented Apache 2.4 64bit on Windows Server 2008 R2 as a reverse proxy to take advantage of a wildcard cert for our domain. We have been disappointed in the performance with even a light load (like 8-10 users) so I have probably misconfigured something. The web sites served by the proxy (Glassfish 3.0.1) become increasingly slow to the point of being unresponsive. Restarting Apache temporarily fixes the problem but it quickly becomes unresponsive. Bypassing the proxy by going to the IP address of the Glassfish server directly we get very fast response times so it seems that GF isn't the issue. The only thing we are doing is the reverse proxy of the SSL frontend on Apache and passing traffic on the internal network with http to the Glassfish server. When the websites are unresponsive, a look at the Windows Server processes indicates VERY little activity. CPU might be running at 10% (single CPU). There are less than 30 TCP connections. Memory utilization is less than a GB (it is a 4GB machine). Network traffic barely breaks the 100kbps threshold. All indications are that the server is barely breaking a sweat, yet the site is unresponsive. I have set the following properties: SSLSessionCache "shmcb:C:/Apache_direct/logs/ssl_scache(1024000)" SSLSessionCacheTimeout 300 ThreadsPerChild 3000 MaxRequestsPerChild 0 KeepAlive On KeepAliveTimeout 15 ProxyRequests Off SSLStrictSNIVHostCheck off ProxyBadHeader Ignore I have virtual host settings for each http and https port for each server. VHost settings follow this pattern: <VirtualHost *:443> ServerName test.domain.org # ProxyPreserveHost On SSLEngine on ProxyPass / http://192.168.80.196:8080/ ProxyPassReverse / https://192.168.80.196:8080/ </VirtualHost> Have I missed an important setting here or misconfigured a setting? Based on what I have read, Apache should be able to easily handle the traffic we have which is as follows: 800 visits an hour. 13k visits daily. 100 uploads an hour of roughly 60mb MP3 files. 400 downloads an hour of those same MP3 files. 800 downloads of small (100k) pdf files. Glassfish is serving all the file uploads/downloads and we allow those sessions to persist up to an hour. Since most of the files are unique, we are a little uncertain of what cache settings we should use (if any). Perhaps I should omit the SSL cache settings? There are many, many hits on Apache performance tuning on the web but few that deal specifically with reverse proxy, SSL and give recommendations based on type/amount of traffic. Some suggestions would be very helpful. Pete Helgren Developer/Team Lead Bible Study Fellowship 210.493.4133 |