I've checked using a combination of the exact same thing Stonehenge::Throttle does (in fact, that module would have saved me a bunch of development time!), plus the apache server-status directive, command line ps on the server, and a test script that uses lots of cpu....
| [reply] |
It just seems odd to me because there's no real communication that I can see about this between apache and the OS. The OS has no concept of a request, so why would it be waiting for the request to finish?
| [reply] |
There is no direct communication between apache and the OS, but there is some point in the lifespan of the process that the OS is told to update the /proc/$pid/stats file with the cpu usage of the process and any child processes... Whatever this trigger may be, it does not occur until the apache process is reset and waiting for the next request to come in. I think if I could find out what triggers that update, I could force an update of the stats before the end of the request, and see the cpu usage.
| [reply] |