Re: Perl FCGI process and Perl's garbage collection

Yes, I concur that you probably have a real leak.  If the memory was
being freed for use by Perl then it should be used in preference to the
other stuff.  Try the tools - Devel::Cycle, etc.

Another option I've used in the past is to periodically check
/proc/self/stat (on Linux) to see how big the current process is and
exit if it goes over a threshold.  A slightly more efficient and more
portable option would be to write an XS wrapper for getrusage(2).

Sam.

ps Top poster!! :P

David Birnbaum wrote:
> Galen,
>
> I thought that, as of Perl 5.6, the system malloc() was the default on
> most implementations, which returns memory back to the system.  Maybe
> you need to see how your Perl was compiled?  Or perhaps you're not
> freeing memory when you think you are.
>
> We have, in our applications, had a counter that caused it to exit
> after X iterations, just to clear up other memory bugs.  It's not a
> big problem, as long as your counter is sufficiently high.
>
> David.
>
> ------------------------------------------------------------------------
>
> Guorui Wu wrote:
>> HI,
>>
>> On most system, Garbage collection of a perl process does Not return
>> the memory freed to system, instead, it reserves it for its (the perl
>> process) future use.
>>
>> I am developing a Perl FCGI application. ab stress test with 30,000
>> requests result in a huge accumulated memory footprint for the FCGI
>> perl process.
>>
>> My question is: even if the code is clean, but due to perl's garbage
>> collection behavior as stated above, a long run perl FCGI process will
>> bound to accumulate memory footprint. so we have to spawn a new
>> process once over some time and exit the parent?
>>
>> Is my understanding correct? or any other alternative solution to this
>> problem so that perl FCGI process Does return memory to the OS?
>>
>> Many thanks
>>
>> Galen
>> ___________________________________
>> fastcgi-developers mailing list
>> http://fastcgi.com/fastcgi-developers/
>>   
> ------------------------------------------------------------------------
>
> ___________________________________
> fastcgi-developers mailing list
> http://fastcgi.com/fastcgi-developers/
>   

___________________________________
fastcgi-developers mailing list
http://fastcgi.com/fastcgi-developers/