Matthew Darwin wrote:
> On Sat, 26 May 2001, Stas Bekman wrote:
>
>
>> You right. It seems that mod_proxy won't abort on SIGPIPE.
>
>
> So how can I get it to without breaking everything else? Any ideas?
>
> I have a process that takes 2-3 minutes to run before the user gets any
> any output. I tell them to wait, but they get impatient. If they hit
> STOP I really want to know about it so I can free up CPU resources for
> when they hit RELOAD to try agin.
>
> Any helpful hints would be most appreciated.
Run the process in the background. Have the browser loop on a reporting
URL w/request ID. That will allow you to show progress:
* The initial request generates a processing job. Please see previous
threads regarding the wisdom of forking mod_perl processes. You may want
to consider writing a lightweight daemon to take processing requests
from mod_perl via a unix domain socket.
* Return a "starting" page to the client which will refresh to a
progress reporting URL, which may be the same URL as the original
request (be sure to include a UUID or some other identifier for the job).
* Each time the progress URL is called, check on the status of the
processing job. Return the data if available, or return another refresh.
Also:
* The progress URL can use the UUID to generate updated status info.
If you are clever, you can even offer a status bar.
* If you keep the results of your processing jobs indexed by, say, the
MD5 of the request parameters that created them, you can skip redundant
processing runs.
-jh
