Hi all, Thank you all very much for your contributions.
I tried to monitor the network with Wireshark: there is only one request from my browser to the server, and not any answer (redirect or otherwise). This means the problem is definitely not with unexpected browser requests. Calling die() at the end of the script and removing the redirect did not change the behavior in any way. I like to think my code is good, and that the server calls the script in an unexpected way. The main reason for this belief is that I do not use multithreading at all, while the logs report the script is running several times in parallel and the network monitor reports a single browser request. I could not find in the docs any server configuration parameter which would re-invoke a script without killing its currently running instance. Unfortunately I cannot post the code here, as everything is spread in several files. If stuck with this issue for too long, I'll consider reducing the script to the root cause of the problem and posting it, but it's toomuch overhead at this stage... Cheers, Marc. On Wed, Mar 11, 2009 at 4:40 PM, haliphax <halip...@gmail.com> wrote: > On Wed, Mar 11, 2009 at 10:30 AM, Martin Zvarík <mzva...@gmail.com> wrote: > > Marc Venturini napsal(a): > >> > >> Hi all, > >> > >> I wrote a PHP script running in Apache which takes more than 30 seconds > to > >> complete. It uses set_time_limit() to extend the time it is allowed to > >> run. > >> The script generates thumbnails from a list of images. Upon completion, > >> the > >> script redirects the browser to another page using HTTP headers. > > > > If you die() at the end of the script and don't redirect does it continue > > this auto-30-seconds execution? > > > >> > >> On my local machine (Windows + EasyPHP), the script runs as expected and > >> completes after a few minutes. > >> > >> I observe an unexpected behavior on my production web server: > >> - The script runs as expected for the first 30 seconds. > >> - After 30 seconds, the same script with the same parameters starts > again > >> in > >> a new thread/process. The initial thread/process is *not* interrupted, > so > >> 2 > >> threads/processes run in parallel, executing the same sequence of > >> operations > >> with a 30 time shift. > >> - The same scenario happens every 30 seconds (i.e.: at 0"30, 1"00, 1"30, > >> and > >> so on), multiplying the parallel threads/processes. > > > >> - The browser keeps on loading while the above happens. > > > >> - After some time, the browser displays a blank page and all the > >> threads/processes stop. I assume this is due to resources exhaustion, > but > >> I > >> have no means to check this assumption. > >> > >> I deduced the above reading a text file in which I log the sequence of > >> called functions. > > > > It all seems as a redirection / unclosed loop problem. > > > >> > >> Unfortunately I have no access *at all* to my production web server > >> configuration (shared hosting, no documentation). I cannot even read the > >> configuration settings. While I'm considering moving to another host, > I'd > >> be > >> extremely pleased to have an explanation of the observed behavior. > >> > >> I have browsed the mailing list archives and looked for an explanation > in > >> other forums to no avail. This thread may deal with the same issue but > >> does > >> not include any explanation or solution: > >> http://www.networkedmediatank.com/showthread.php?tid=17140 > >> > >> Thanks for reading, and please do not hesitate to ask for further > >> explanations if what I'm trying to achieve was not clear! > > > > Why it works on your local server is probably caused by different > > versions/settings, but I bet there's an error somewhere in your script. > > > > Consider sending it here, I'll take a look. > > A blank URL does not redirect to the directory index, IIRC... it > refreshes the current page (such as a FORM tag with ACTION=""). This > may very well still be your problem. > > My 2c, > > > -- > // Todd > > -- > PHP General Mailing List (http://www.php.net/) > To unsubscribe, visit: http://www.php.net/unsub.php > >