On Fri, 2005-03-18 at 11:15 -0500, Stas Bekman wrote: > Anders, do you by chance use a proxy that may have eaten the response? > > What do you see in logs/access_log, it tells how many bytes were sent to > the client. > > What happens if you replace print($x) with Apache->request->print($x) >
The problem occur when I switch some scripts from mod_cgi to mod_perl. When I run mod_cgi the front end server takes care of the request. This setup doesn't have this problem. I use mod_perl by forwarding (proxying) the request to a back-end mod_perl server. The mod_perl server is listing on a port on 127.0.0.1. The problem situation occurs with this setup after a while (usually after days). To eliminate the effect of the proxy in my tests I used wget on the same machine direct on the mod_perl server. I even created a separate mod_perl server that wasn't used by the front-end. When the problem situation occurred this independent server also had the problem. This make me believe that the problem is related to a resource being depleted. Any ideas on what to monitor? In the access log I only see lines with the number of bytes being correct. Example: 127.0.0.1 - - [16/Mar/2005:10:56:20 +0100] "GET /cgise/env.cgi?x=1598223 HTTP/1.0" 200 1598223 "-" "Wget/1.8.2" 0 However it looks like the number of lines in the access log is too low. This would indicate that some lines are missing. Maybe missing responses didn't result in a line in the access log. Since I can't reproduce the problem I can't be sure of this. If the problem comes back I will look closer at this. I will also replace the print with Apache->request->print and see what happens. Thanks again Stas :-)
