Re: fastcgi+php=lots of 500 errors ?
"Gabriel Barazer" <[email protected]>
| Newsgroups | gmane.comp.web.fastcgi.devel |
|---|---|
| Organization | REAGI |
| Message-ID | <016801c51fda$c8d85700$4500a8c0@inspiron> |
On Wednesday, March 02, 2005 10:40 PM CET, Jason Pump <JPump-HbNbuK1M3AseIZ0/[email protected]> wrote: > What OS are you running on? > Linux kernel 2.6.10. > Have you checked that PHP is not running out of memory, or modifying > php.ini to allow more memory for the process to see how that effects > the situation? > Yes, and it's not a memory problem. I finally solved the problem with mod_fcgid. I think the problem was the timeout in the communication between mod_fastcgi and the php process. In the mod_fcgid, i got the IPCCommTimeout variable which has the desired effect (not returning a 500 when php take over 3 seconds to respond). I've note found the same variable in mod_fastcgi. What about adding this directive in fastcgi ? In the next weeks, i will certainly attempt to code a fastcgi module designed for use with php and minimal configuration (but with the tuning stuff), inspired on the fastcgi and mod_fcgid code. I'll let know if this project becomes reality :) So using mod_fcgid (+a suexec patch for my case) seems to work properly with php (with the very useful "wrapper" function): zero "500" errors since i've passed to this module. Hope this maybe helps someone, -- Gabriel Barazer <gabriel-G2fI/[email protected]> REAGI >> -----Original Message----- >> From: >> fastcgi-developers-bounces+jpump=looksmart.net-xGejAJT2w6yOE5Ap4OspEtHuzzzSOjJt@public.gmane.org >> [mailto:fastcgi-developers-bounces+jpump=looksmart.net-xGejAJT2w6yOE5Ap4OspEv8+0UxHXcjY@public.gmane.org >> et]On Behalf Of Gabriel Barazer >> Sent: Monday, February 28, 2005 12:26 PM >> To: Simon; fastcgi-developers-xGejAJT2w6yOE5Ap4OspEtHuzzzSOjJt@public.gmane.org >> Subject: Re: [FASTCGI] fastcgi+php=lots of 500 errors ? >> >> >> On Monday, February 28, 2005 9:17 PM CET, >> Simon <[email protected]> wrote: >> >>> Is that the only error you see in your error log? when you see >>> 200req a sec is that all PHP? is it 200req/sec per server or on all >>> 3? >>> >> >> excepted the normal info messages from fastcgi such as: >> [Mon Feb 28 21:20:09 2005] [warn] FastCGI: (dynamic) server >> "/usr/local/bin/php-fcgi/php.fcgi" (pid 31004) terminated by calling >> exit with status '0' >> >> or >> >> [Mon Feb 28 21:20:09 2005] [warn] FastCGI: (dynamic) server >> "/usr/local/bin/php-fcgi/php.fcgi" (pid 31417) termination signaled >> >> >> or even >> >> [Mon Feb 28 21:20:09 2005] [warn] FastCGI: (dynamic) server >> "/usr/local/bin/php-fcgi/php.fcgi" (uid 522, gid 100) started (pid >> 31533) >> >> which are normal info message about fastcgi activity, yes , this >> is the only >> real errors. >> >> The servers are 3 pentium 4 3.2Ghz and it is 200req/s on all 3. If >> they weren't 500's , the configuration was clean : lesser than load >> average of 1, >> and a maximum of 400req/s capacity on all 3 servers (benchmark >> results, and >> a load not higher than 5). >> >> No other debugging information... I think i'll become crazy soon... >> >> BTW: i tested with the "stable" fastcgi version (2.4.2), and it >> produces the >> same result. >> >> -- >> Gabriel Barazer <gabriel-G2fI/[email protected]> >> REAGI >> >>> >>> On Mon, 28 Feb 2005 19:26:06 +0100, Gabriel Barazer wrote: >>> >>>> Hi, >>>> >>>> When using php with mod_fastcgi , i get a HUGE number of 500 errors >>>> and this line in the error log: >>>> FastCGI: incomplete headers (0 bytes) received from server >>>> "/usr/local/bin/php-fcgi/php.fcgi" >>>> >>>> I have a moderated traffic (~200req per second on 3 servers), and >>>> these 500's occurs 2-3 times per second on each server. >>>> >>>> My actual config is the same on both servers: >>>> --- >>>> FastCgiWrapper On >>>> FastCgiConfig -gainValue 0.8 -killInterval 5 -listen-queue-depth >>>> 512 -maxClassProcesses 96 -maxProcesses 192 -multiThreshold >>>> 60 -singleThreshold 1 -startDelay 1 -updateInterval 5 >>>> ScriptAlias /fcgi-bin/ "/usr/local/bin/php-fcgi/" >>>> <Location /fcgi-bin/> >>>> Options ExecCGI >>>> SetHandler fastcgi-script >>>> </Location> >>>> >>>> AddHandler php-script .php .inc .pinc .phtml .php3 .php4 >>>> Action php-script /fcgi-bin/php.fcgi >>>> --- >>>> >>>> It appears that if i raise the killinterval AND updateinterval, the >>>> number of errors decrease after a restart of apache, but as the >>>> uptime grows, the number of errors grows too... >>>> >>>> i haven't succeed to reproduce this bug as it seems to be very >>>> random (but it happens very more often when the traffic increases) >>>> >>>> Anyone have an idea about this *very* annoying effect ? >>>> >>>> Oh, i forgot to mention the software versions: >>>> apache 2.0.53 >>>> php 4.3.10 >>>> mod_fastcgi-snap-0404142202 >>>> >>>> -- >>>> Gabriel Barazer <gabriel-G2fI/[email protected]> >>>> REAGI >> >> ___________________________________ >> fastcgi-developers mailing list >> http://fastcgi.com/fastcgi-developers/ ___________________________________ fastcgi-developers mailing list http://fastcgi.com/fastcgi-developers/