Re: [Fwd: Re: How to reload PERL module in all Apache children]

[do...@pobox.com]

if you use Apache::Reload with a threaded MPM and multiple interpreters, 
the modules will be reloaded by each interpreter as they are used, not 
every interpreter all at once.  similar to 1.x where each child has 
its own interpreter, the modules are reloaded as each kid is hit with a 
request.

note that if a module is loaded at startup, the syntax tree of each 
subroutine is shared between interpreters (big win), but each subroutine 
has its own padlist (where lexical my variables are stored).  once 
Apache::Reload reloads a module, this sharing goes away and each 
interpreter will have its own copy of the syntax tree for the given 
subroutines.

Re: [Fwd: Re: How to reload PERL module in all Apache children]

[do...@pobox.com]

On Thu, 11 Apr 2002, Perrin Harkins wrote:
 
> Does it look you'll be able to get the solar variables idea to work for 
> those data types?

i had a simple prototype way back that sorta worked for simple scalars, 
probably won't take it any further now that there is threads::shared in 
5.7.x.  haven't tried it but i believe the concept is the same.

Re: [Fwd: Re: How to reload PERL module in all Apache children]

[Perrin Harkins <pe...@elem.com>]

dougm@pobox.com wrote:
> On Fri, 12 Apr 2002, Stas Bekman wrote:
>  
> 
>>But if talk about futuristic Solar variables (perl globals shared 
>>between threads). what if a solar variable is a reference to CODE? Can 
>>this be shared? If so, will reloading this variable in one interpreter 
>>affect others?
> 
> 
> even if that would work, it would kill performance due to required mutex 
> locking.  and that locking would need to happen in the perl core, unlike
> simple scalars, arrays and hashes.

Does it look you'll be able to get the solar variables idea to work for 
those data types?

- Perrin

Re: [Fwd: Re: How to reload PERL module in all Apache children]

[do...@pobox.com]

On Fri, 12 Apr 2002, Stas Bekman wrote:
 
> But if talk about futuristic Solar variables (perl globals shared 
> between threads). what if a solar variable is a reference to CODE? Can 
> this be shared? If so, will reloading this variable in one interpreter 
> affect others?

even if that would work, it would kill performance due to required mutex 
locking.  and that locking would need to happen in the perl core, unlike
simple scalars, arrays and hashes.

> Also if we put the sharing aside for a moment and assuming that we have 
> a pool of interpreters with idle interpreters in it, there can be a 
> thread that monitors changed modules and update the idle interpreters by 
> making them reload the code and put them in the head of the list. This 
> should save the overhead of reloading during a request. Does this make 
> sense?

there already is a plan to have a low-priority thread that monitors idle 
interpreters.  this would be a pluggable interface, so you can do whatever 
you want with the interpreter via callback hooks.  but that'll all wait 
until well after everything else is solid with ithreads, including and 
most important: perl 5.8.x

Re: [Fwd: Re: How to reload PERL module in all Apache children]

[Stas Bekman <st...@stason.org>]

dougm@pobox.com wrote:
> if you use Apache::Reload with a threaded MPM and multiple interpreters, 
> the modules will be reloaded by each interpreter as they are used, not 
> every interpreter all at once.  similar to 1.x where each child has 
> its own interpreter, the modules are reloaded as each kid is hit with a 
> request.

thanks for correcting me. It's all about interpreters and not threads, 
so it's the same for the code with forked and threaded httpds.

But if talk about futuristic Solar variables (perl globals shared 
between threads). what if a solar variable is a reference to CODE? Can 
this be shared? If so, will reloading this variable in one interpreter 
affect others?

Also if we put the sharing aside for a moment and assuming that we have 
a pool of interpreters with idle interpreters in it, there can be a 
thread that monitors changed modules and update the idle interpreters by 
making them reload the code and put them in the head of the list. This 
should save the overhead of reloading during a request. Does this make 
sense?



__________________________________________________________________
Stas Bekman            JAm_pH ------> Just Another mod_perl Hacker
http://stason.org/     mod_perl Guide ---> http://perl.apache.org
mailto:stas@stason.org http://use.perl.org http://apacheweek.com
http://modperlbook.org http://apache.org   http://ticketmaster.com