Bugzilla – Full Text Bug Listing
| Summary: | CDO downloads page calls the webserver, loops and consumes many resources | ||||||
|---|---|---|---|---|---|---|---|
| Product: | [Modeling] EMF | Reporter: | Denis Roy <denis.roy> | ||||
| Component: | cdo.docs | Assignee: | Project Inbox <emf.cdo-inbox> | ||||
| Status: | CLOSED FIXED | QA Contact: | |||||
| Severity: | critical | ||||||
| Priority: | P3 | CC: | stepper | ||||
| Version: | 4.0 | ||||||
| Target Milestone: | --- | ||||||
| Hardware: | PC | ||||||
| OS: | Linux | ||||||
| Whiteboard: | |||||||
| Attachments: |
|
||||||
http://www.eclipse.org/cdo/downloads/ seems to be the culprit. I've commented out this line, since it seems to be the culprit: # printDownloads("http://www.eclipse.org/modeling/emf/downloads/index.php?project=cdo&showAll=0&showMax=5"); Actually, http://www.eclipse.org/modeling/emf/downloads/index.php?project=cdo&showAll=0&showMax=5 seems to redirect to /downloads/cdo/, which is a big part of the problem. Investigating... I've found the loop and removed it. Sorry for the hazzle! Denis, am I right that you've entirely "blanked" the URL http://www.eclipse.org/cdo/downloads/ ? I can not test whether my changes aolve the issue but I'm 100% sure that they do. Please remove your extra filter and try it... I've removed my code and updated from CVS. Looks like it works now, thanks. Closing. |
Created attachment 182643 [details] Access log On www.eclipse.org, I am seeing frequent exhaustion of available sockets for http clients. It appears there is something broken with the CDO downloads page. Since this page calls the webserver itself (instead of opening files locally), each nested call occupies one client 'slot' until all 800 are consumed and never released. Please: a) avoid opening "local" URLs such as www.eclipse.org -- this consumes 2x the memory for one page hit b) test for endless loops in your PHP code Attached is a very small subset of my Apache log. The IP address you see is the webserver calling itself. I'm seeing hits in excess of 30/second.