Hi Norbert, can we keep the users@ mailing list CC-ed on this since it may be relevant for others too? Moreover people with a better knowledge of mod_perl than me could chime in...
On Wed, Sep 13, 2017 at 1:33 PM, Norbert <ngrun...@uni-muenster.de> wrote: > > thanks - but it does not help - The case seems to be mod_perl specific (I know very few about it), "normally" preforked processes have the lifetime given by these directives... Could it be a crash of children processes somehow, or a special handling of processes in mod_perl? The creation of multiple root processes at startup looks special to me, usually httpd has a single root process to maintain children (www:www) processes, is a root process still alive after the load? For others on the list, your results below: > have a look, I use top in 1 second intervals > and extracted the httpd processes which have the user www. > The minus represent the 1 second gap: > > - > - > 30830 www 1 24 0 326M 31884K select 0 0:00 0.00% httpd > 30832 www 1 27 0 326M 31636K select 4 0:00 0.00% httpd > 30833 www 1 27 0 326M 31636K select 5 0:00 0.00% httpd > 30834 www 1 27 0 326M 31636K select 1 0:00 0.00% httpd > 30831 www 1 27 0 326M 31644K select 6 0:00 0.00% httpd > 30835 www 1 27 0 326M 31632K select 0 0:00 0.00% httpd > - > 30830 www 1 24 0 326M 31884K select 0 0:00 0.49% httpd > 30832 www 1 27 0 326M 31636K select 4 0:00 0.49% httpd > 30833 www 1 27 0 326M 31636K select 5 0:00 0.49% httpd > 30834 www 1 27 0 326M 31636K select 1 0:00 0.49% httpd > 30831 www 1 27 0 326M 31644K select 6 0:00 0.49% httpd > 30835 www 1 27 0 326M 31632K select 0 0:00 0.49% httpd > - > 30830 www 1 21 0 326M 31928K select 3 0:00 0.39% httpd > 30832 www 1 27 0 326M 31636K select 4 0:00 0.39% httpd > 30833 www 1 27 0 326M 31636K select 5 0:00 0.39% httpd > 30834 www 1 27 0 326M 31636K select 1 0:00 0.39% httpd > 30831 www 1 27 0 326M 31644K select 6 0:00 0.39% httpd > 30835 www 1 27 0 326M 31632K select 0 0:00 0.39% httpd > - > 30830 www 1 21 0 330M 35880K select 3 0:00 0.59% httpd > 30832 www 1 27 0 326M 31636K select 4 0:00 0.29% httpd > 30833 www 1 27 0 326M 31636K select 5 0:00 0.29% httpd > 30834 www 1 27 0 326M 31636K select 1 0:00 0.29% httpd > 30831 www 1 27 0 326M 31644K select 6 0:00 0.29% httpd > 30835 www 1 27 0 326M 31632K select 0 0:00 0.29% httpd > - > 30830 www 1 21 0 330M 35880K select 3 0:00 0.49% httpd > 30832 www 1 27 0 326M 31636K select 4 0:00 0.20% httpd > 30833 www 1 27 0 326M 31636K select 5 0:00 0.20% httpd > 30834 www 1 27 0 326M 31636K select 1 0:00 0.20% httpd > 30831 www 1 27 0 326M 31644K select 6 0:00 0.20% httpd > 30835 www 1 27 0 326M 31632K select 0 0:00 0.20% httpd > - > 30830 www 1 21 0 330M 35900K select 4 0:00 0.39% httpd > 30831 www 1 20 0 326M 31652K select 4 0:00 0.20% httpd > 30835 www 1 20 0 326M 31636K select 6 0:00 0.10% httpd > - > 30830 www 1 20 0 330M 35900K select 1 0:00 0.29% httpd > 30831 www 1 20 0 326M 31876K select 0 0:00 0.20% httpd > 30835 www 1 20 0 326M 31636K select 5 0:00 0.10% httpd > - > 30831 www 1 20 0 326M 31932K select 7 0:00 0.29% httpd > 30830 www 1 20 0 330M 35900K select 0 0:00 0.20% httpd > 30835 www 1 20 0 326M 31636K select 5 0:00 0.00% httpd > - > 30831 www 1 20 0 326M 31932K select 7 0:00 0.20% httpd > 30830 www 1 20 0 330M 35900K select 0 0:00 0.10% httpd > 30835 www 1 20 0 326M 31636K select 5 0:00 0.00% httpd > - > 30831 www 1 20 0 326M 31932K select 7 0:00 0.10% httpd > 30830 www 1 20 0 330M 35900K select 0 0:00 0.00% httpd > 30835 www 1 20 0 326M 31636K select 5 0:00 0.00% httpd > - > 30830 www 1 20 0 330M 35900K select 0 0:00 0.00% httpd > 30831 www 1 20 0 326M 31932K select 7 0:00 0.00% httpd > 30835 www 1 20 0 326M 31636K select 5 0:00 0.00% httpd > - > 30830 www 1 20 0 330M 35900K select 0 0:00 0.00% httpd > 30831 www 1 20 0 326M 31932K select 7 0:00 0.00% httpd > 30835 www 1 20 0 326M 31636K select 5 0:00 0.00% httpd > - > - > - > > and then no www user is activae any more... Note: the test machine is only > used by myself. > > Regards, Norbert > > > > On 09/13/17 13:08, Yann Ylavic wrote: >> >> On Wed, Sep 13, 2017 at 10:10 AM, Norbert <ngrun...@uni-muenster.de> >> wrote: >>> >>> Thanks, I set (on my testmachine): >>> >>> <IfModule mpm_prefork_module> >>> MinSpareServers 4 >>> MaxSpareServers 4 >>> MaxConnectionsPerChild 100000 >>> </IfModule> >>> >>> but nothing changes - the httpd processes with user www:www appear and >>> disappear in top after seconds... >> >> We the above, there should be at least 4 processes that don't >> disappear before 100000 handling >> 100000 connections. >> >> But if more/other processes are created because of the load, they >> would be recycled as soon as the load decreases. >> >> >> What do you observe with the below ? >> >> <IfModule mpm_prefork_module> >> StartServers 4 >> ServerLimit 80 >> MinSpareServers 4 >> MaxSpareServers 40 >> MaxRequestWorkers 80 >> MaxConnectionsPerChild 0 >> </IfModule> >> >> >> Regards, >> Yann. > > > -- > Norbert Grundmann > Inst. of Bioinformatics Muenster > Tel. 0251 83 53 007 > (Use *BSD, because Linux is a patch for Linux) > --------------------------------------------------------------------- To unsubscribe, e-mail: users-unsubscr...@httpd.apache.org For additional commands, e-mail: users-h...@httpd.apache.org
