httpd-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Luc Andre <landre3...@gmail.com>
Subject Re: [users@httpd] Random latency in reentrant calls (Bug 57916)
Date Tue, 12 May 2015 10:07:29 GMT
My mistake, apache2.conf is overridden by  /mods-enabled/mpm_prefork.conf

sorry about that...




On Tue, May 12, 2015 at 12:03 PM, Luc Andre <landre3567@gmail.com> wrote:

> php_sapi_name: apache2handler
>
>
> Server MPM:     prefork
>
> I tried
>
>     StartServers          20
>     MinSpareServers       20
>     MaxSpareServers      20
>
> And after restarting apache still have the issue but only after waiting
> about a minute between 2 tries.
>
> Then I did a "ps -ef | grep apache" and I discovered that I have only 11
> apache processes running !
>
> So it seems that MinSpareServers and StartServers  are ignored and the
> problem is really a child spin up issue.
>
> Regards,
>
> Luc
>
>
>
>
>
>
> On Tue, May 12, 2015 at 11:16 AM, Rainer Canavan <
> rainer.canavan@sevenval.com> wrote:
>
>>
>>
>> 2015-05-12 10:03 GMT+02:00 Luc Andre <landre3567@gmail.com>:
>>
>>> Thanks for your reply.
>>>
>>> We did a test on a powerful server with
>>>
>>>     StartServers          20
>>>     MinSpareServers       5
>>>     MaxSpareServers      20
>>>
>>> And we still have the issue...
>>>
>>
>> To ensure that you don't hit the child spin up issue, you'll have to set
>>  MinSpareServers to a value equal to or greater than the number of
>> processes you need for your test, I'd recommend 20 in this case, just to be
>> sure. Are you sure that you're using the prefork mpm, and therefore those
>> settings are actually relevant? If you don't have a good reason to use
>> prefork, you may get better performance with worker or event (but be sure
>> to tune the associated settings, such as ThreadsPerChild appropriately)
>>
>> Additionally, you don't specify how your PHP is configured. If you're
>> using PHP-FPM, you need to ensure that pm.start_servers and pm.min_spare_servers
>> are large enough as well.
>>
>> rainer
>>
>
>

Mime
View raw message