Invalid offload() log messages

[unbit]

Invalid offload() log messages

[unbit]

fixed (or better, 'improved') in 1.9

[prymitive]

[pid: 16500|app: 0|req: -1/1] 172.16.110.48 () {28 vars in 355 bytes} [Fri Feb 8 21:45:48 2013] GET /bigfile => generated 536870912 bytes in 0 msecs via offload() (HTTP/1.1 200) 2 headers in 92 bytes (0 switches on core 0)

I'm getting proper HTTP status code, but message is logged immediately and shows 0 msecs, shouldn't this be logged by offload thread itself once it finishes sending file?
Not a big issue but request duration is lost in logs this way.

[prymitive]

To have proper time reporting for offloaded request, we would have to:
a) move logging to sendfile functions in core/offload.c
b) add callback support to offload function and add log callbacks if requests are offloaded
c) some other/better way

it's very very unimportant for me

[prymitive]

I've created new milestone "Future Ideas" to track such feature requests

[unbit]

This is an interesting problem, as offloading is becaming present in more and more areas. For proxy offloaders (http, uwsgi) the backend node will log normally, so the information is not lost. Response offloaders (cache, and the new 'offload' transformation) refers only to data transfer time (as well as sendfile()), so we need to evaluate if such information is relevant (considering kernel buffering all over the place) to a level needing a new logging category.

[prymitive]

I don't care for upload time for static files so let's close it