Mailing List Archive

On 2/22/2019 2:15 PM, John Dunlap wrote:
> The Prefork MPM has the following settings:
> <IfModule mpm_prefork_module>
>         StartServers                     5
>         MinSpareServers           5
>         MaxSpareServers          10
>         MaxRequestWorkers         150
>         MaxConnectionsPerChild   0
> </IfModule>
>
> I see these Apache directives but the documentation states that they
> only apply to a threaded mod_perl/Apache which prefork definitely is not:
> PerlInterpStart
> PerlInterpMax
> PerlInterpMinSpare
> PerlInterpMaxSpare
> PerlInterpMaxRequests
>
> How does mod_perl allocate interpreters to the prefork worker processes?
> Is there one perl interpreter for each of preform worker? Is there a
> pool of perl interpreters which is smaller than the pool of prefork
> workers? Are there settings for configuring the size of the perl
> interpreter pool? When a request comes in, does the user have to wait
> for a perl interpreter process to start or is there already one waiting
> for them?

If you're asking about mod_perl and prefork:

Apache handles child process creation. Each Apache child process loads
the Perl interpreter when it starts and the Perl interpreter persists in
the process till the process terminates. So, at any given time, there're
as many Perl interpreters loaded and ready as there're Apache processes.
When a request comes in, if there's an available Apache process to serve
the request, it's served. If not, Apache will create a new process or
queue the request to be served when a process becomes available.

With the numbers above, your server is able to handle 150 simultaneous
requests. If processes are not busy serving requests, then Apache will
kill idle processes based on (Min|Max)SpareServers of 5,10. For an
overly simple example, if you get 150 requests at 8:30 am, the server
could create 145 processes to meet demand. However, if not busy at 8:31
am, when everyone has stepped away for coffee, your process count could
be down near 10. When the boss comes in at 10am and it's back to work,
the server will create 140 new ... at 10:01am, it's down to 10 again.

Generally, with mod_perl, you don't want Apache to kill idle processes
but want them to be around, ready to handle new requests. For maximum
performance, (Min|Max)SpareServers could be as high as your
MaxRequestWorkers. Of course, all of this depends on your RAM, CPU
utilization etc. and you should be able to find a happy medium as you
tweak the numbers.

>
> --
> John Dunlap
> /CTO | Lariat/
> /
> /
> /*Direct:*/
> /john@lariat.co <mailto:john@lariat.co>/
> /
> *Customer Service:*/
> 877.268.6667
> support@lariat.co <mailto:support@lariat.co>

On 24.02.2019 01:50, Gazzali Jaleel wrote:
>
> On 2/22/2019 2:15 PM, John Dunlap wrote:
>> The Prefork MPM has the following settings:
>> <IfModule mpm_prefork_module>
>> StartServers 5
>> MinSpareServers 5
>> MaxSpareServers 10
>> MaxRequestWorkers 150
>> MaxConnectionsPerChild 0
>> </IfModule>
>>
>> I see these Apache directives but the documentation states that they only apply to a
>> threaded mod_perl/Apache which prefork definitely is not:
>> PerlInterpStart
>> PerlInterpMax
>> PerlInterpMinSpare
>> PerlInterpMaxSpare
>> PerlInterpMaxRequests
>>
>> How does mod_perl allocate interpreters to the prefork worker processes? Is there one
>> perl interpreter for each of preform worker? Is there a pool of perl interpreters which
>> is smaller than the pool of prefork workers? Are there settings for configuring the size
>> of the perl interpreter pool? When a request comes in, does the user have to wait for a
>> perl interpreter process to start or is there already one waiting for them?
>
> If you're asking about mod_perl and prefork:
>
> Apache handles child process creation. Each Apache child process loads the Perl
> interpreter when it starts

Actually, I believe that since this is "prefork", it is the main Apache process (which
already contains a perl interpreter) which gets "forked" (in other words duplicated, as it
is) to create a child. So there is not really a perl interpreter "being started" here (at
least not in the sense of "being loaded" etc), it is already loaded and it is part of what
is being duplicated.
(For a more detailed and precise explanation and example, see :
http://perl.apache.org/docs/2.0/user/handlers/server.html)
Correct me if I'm wrong.

I have no qualms with the rest below, sounds correct to me.

and the Perl interpreter persists in the process till the
> process terminates. So, at any given time, there're as many Perl interpreters loaded and
> ready as there're Apache processes. When a request comes in, if there's an available
> Apache process to serve the request, it's served. If not, Apache will create a new process
> or queue the request to be served when a process becomes available.
>
> With the numbers above, your server is able to handle 150 simultaneous requests. If
> processes are not busy serving requests, then Apache will kill idle processes based on
> (Min|Max)SpareServers of 5,10. For an overly simple example, if you get 150 requests at
> 8:30 am, the server could create 145 processes to meet demand. However, if not busy at
> 8:31 am, when everyone has stepped away for coffee, your process count could be down near
> 10. When the boss comes in at 10am and it's back to work, the server will create 140 new
> ... at 10:01am, it's down to 10 again.
>
> Generally, with mod_perl, you don't want Apache to kill idle processes but want them to be
> around, ready to handle new requests. For maximum performance, (Min|Max)SpareServers
> could be as high as your MaxRequestWorkers. Of course, all of this depends on your RAM,
> CPU utilization etc. and you should be able to find a happy medium as you tweak the numbers.
>
>>
>> --
>> John Dunlap
>> /CTO | Lariat/
>> /
>> /
>> /*Direct:*/
>> /john@lariat.co <mailto:john@lariat.co>/
>> /
>> *Customer Service:*/
>> 877.268.6667
>> support@lariat.co <mailto:support@lariat.co>

On 2/24/2019 1:32 PM, André Warnier (tomcat) wrote:
> On 24.02.2019 01:50, Gazzali Jaleel wrote:
>>
>> On 2/22/2019 2:15 PM, John Dunlap wrote:
>>> The Prefork MPM has the following settings:
>>> <IfModule mpm_prefork_module>
>>>          StartServers                     5
>>>          MinSpareServers           5
>>>          MaxSpareServers          10
>>>          MaxRequestWorkers         150
>>>          MaxConnectionsPerChild   0
>>> </IfModule>
>>>
>>> I see these Apache directives but the documentation states that they
>>> only apply to a
>>> threaded mod_perl/Apache which prefork definitely is not:
>>> PerlInterpStart
>>> PerlInterpMax
>>> PerlInterpMinSpare
>>> PerlInterpMaxSpare
>>> PerlInterpMaxRequests
>>>
>>> How does mod_perl allocate interpreters to the prefork worker
>>> processes? Is there one
>>> perl interpreter for each of preform worker? Is there a pool of perl
>>> interpreters which
>>> is smaller than the pool of prefork workers? Are there settings for
>>> configuring the size
>>> of the perl interpreter pool? When a request comes in, does the user
>>> have to wait for a
>>> perl interpreter process to start or is there already one waiting for
>>> them?
>>
>> If you're asking about mod_perl and prefork:
>>
>> Apache handles child process creation. Each Apache child process loads
>> the Perl
>> interpreter when it starts
>
> Actually, I believe that since this is "prefork", it is the main Apache
> process (which already contains a perl interpreter) which gets "forked"
> (in other words duplicated, as it is) to create a child.  So there is
> not really a perl interpreter "being started" here (at least not in the
> sense of "being loaded" etc), it is already loaded and it is part of
> what is being duplicated.
> (For a more detailed and precise explanation and example, see :
> http://perl.apache.org/docs/2.0/user/handlers/server.html)
> Correct me if I'm wrong.

That's right, thanks for catching it. Should've said child processes
will already have the Perl interpreter loaded.

>
> I have no qualms with the rest below, sounds correct to me.
>
>  and the Perl interpreter persists in the process till the
>> process terminates. So, at any given time, there're as many Perl
>> interpreters loaded and
>> ready as there're Apache processes. When a request comes in, if
>> there's an available
>> Apache process to serve the request, it's served. If not, Apache will
>> create a new process
>> or queue the request to be served when a process becomes available.
>>
>> With the numbers above, your server is able to handle 150 simultaneous
>> requests. If
>> processes are not busy serving requests, then Apache will kill idle
>> processes based on
>> (Min|Max)SpareServers of 5,10. For an overly simple example, if you
>> get 150 requests at
>> 8:30 am,  the server could create 145 processes to meet demand.
>> However, if not busy at
>> 8:31 am, when everyone has stepped away for coffee, your process count
>> could be down near
>> 10. When the boss comes in at 10am and it's back to work, the server
>> will create 140 new
>> ... at 10:01am, it's down to 10 again.
>>
>> Generally, with mod_perl, you don't want Apache to kill idle processes
>> but want them to be
>> around, ready to handle new requests. For maximum performance,
>> (Min|Max)SpareServers
>> could be as high as your MaxRequestWorkers. Of course, all of this
>> depends on your RAM,
>> CPU utilization etc. and you should be able to find a happy medium as
>> you tweak the numbers.
>>
>>>
>>> --
>>> John Dunlap
>>> /CTO | Lariat/
>>> /
>>> /
>>> /*Direct:*/
>>> /john@lariat.co <mailto:john@lariat.co>/
>>> /
>>> *Customer Service:*/
>>> 877.268.6667
>>> support@lariat.co <mailto:support@lariat.co>
>
>