Occasionally a request will fail when connecting to the devenv. [Fri Jun 08 13:54:42 2012] [error] [client 127.0.0.1] Premature end of script headers: console, referer: https://ec2-23-20-107-39.compute-1.amazonaws.com/app/login [ pid=4934 thr=140441917069280 file=ext/apache2/Hooks.cpp:821 time=2012-06-08 13:54:42.314 ]: The backend application (process 3995) did not send a valid HTTP response; instead, it sent nothing at all. It is possible that it has crashed; please check whether there are crashing bugs in this application. I can't consistently tie this down to a certain URL, but it does happen in the automated tests (libra_web is randomly failing because of it). Need an Apache expert to help me figure out what to do.
Passing to Mike to help find an Apache expert. Could this be timeout related?
The bulk of times I can remember seeing this is outside of a 3-5 second window since the last click. It's possible this is related to the keepalive settings in the devenv or the proxy configuration - I don't know if we configure the balancer in the devenv exactly like stage/prod.
Discussed on 2012-06-19 bug review call... We believe this is related to restarting apache. Not seeign in production, lowering severity.
*think* this is related to graceful restarts and the proxy pool being not empty during the graceful. Want to try using proxy-initial-not-pooled in devenv to see if this resolves the issue, if we can reproduce it.
Should be long fixed.