27

I have Apache2 with PHP + PHP-FPM configured according to:

http://wiki.apache.org/httpd/PHP-FPM

I am writing a script that will take a long time to execute on an internal Vhost, but keep getting timed out, everything runs flawlessly if the script executes in under 30 seconds.

My apache log tells me:

[Wed Apr 17 21:57:23.075175 2013] [proxy_fcgi:error] [pid 9263:tid 140530454267648] (70007)The timeout specified has expired: [client 58.169.202.172:49017] AH01075: Error dispatching request to :, referer:

When trying to run the script I am given a 503 Service Unavailable after exactly 30 seconds of execution time. Logically this would mean I have a timeout directive or setting set to 30 seconds, but I have these in my Vhost's config:

Timeout 600
<IfModule proxy_module>
    ProxyPassMatch ^/(.*\.php)$ fcgi://127.0.0.1:9001/home/pyrokinetiq/scripts/$1 timeout=600
    ProxyTimeout 600
</IfModule>

(php-fpm is running on port 9001 for me)

I have also tried placing the Timeout and ProxyTimeout in httpd.conf with no difference.

It seems there's another timeout setting somewhere that's specific to mod_proxy_fcgi, but I can't find it. I installed the Apache2 httpd from the official tarball, none of the mods seem to have come with any configuration files.

If anyone can point me in the right direction it would be much appreciated.

wyqydsyq
  • 385
  • 1
  • 3
  • 9

9 Answers9

34

I finally fixed this problem after testing several configuration parameters. I tested the solution twice, removing all previous changes. Only one parameter was needed for me to fix it.

For the latest versions of httpd and mod_proxy_fcgi you can simply add timeout= to the end of the ProxyPassMatch line, e.g.:

ProxyPassMatch ^/(.+\.php.*)$ fcgi://127.0.0.1:9000/<docroot>/$1 timeout=1800

For older versions it was a little more complicated, e.g.:

<Proxy fcgi://127.0.0.1:9000>
  ProxySet timeout=1800
</Proxy>
ProxyPassMatch ^/(.+\.php.*)$ fcgi://127.0.0.1:9000/<docroot>/$1

I needed to add the Proxy directive to set the timeout to 30 minutes. In some applications, usually when operating database, there are routines that can take more than 10 minutes to execute. I temporary set the timeout to 30 minutes to ensure they finish. Specifically useful when using the installation wizard, which takes too much time (in my humble opinion).

By the way the intial input that helped me to solve this issue was found in the following URL address.

rsaw
  • 253
  • 2
  • 5
Jordi Ferran
  • 474
  • 3
  • 5
  • 1
    It looks like this broke under recent versions of Apache, AH00526: ProxyPass/ and ProxyPassMatch/ can't be used altogether with the same worker name – Stewart Adam Jan 24 '15 at 10:15
  • 4
    I have resolved the above by adding a parameter 'timeout=120' at the end of the ProxyPassMatch line. – Stewart Adam Jan 24 '15 at 10:28
  • @Palantir glad to hear it! [Submitted as answer](http://serverfault.com/a/698154/123656). – Stewart Adam Jun 11 '15 at 00:54
  • 2
    Two more things I needed: First, you -must- set "Timeout" and "ProxyTimeout" in your global apache config file to be longer than the other FPM timeouts. Second, My FPM pool was listening on a unix socket and I use SetHandler like so: [ SetHandler "proxy:unix:/var/run/php/example.com-php7.0-fpm.sock|fcgi://localhost:8000" ] . But matches on the fcgi://localhost part of the SetHandler line (the part AFTER the |... which isn't even actually used !) and NOT the unix:/var/run/ part. so to config timeouts for above use: and not – Professor Falken Jun 14 '18 at 17:13
  • 1
    As of today (2021), the content at the "URL address" is gone. Please read it on https://web.archive.org/web/20200730023817/http://apache-http-server.18135.x6.nabble.com/odd-30-second-timeout-when-using-mod-proxy-fcgi-and-php-fpm-td5005245.html (I cannot suggest it as edit because "suggested edit queue is full") – 4wk_ Jun 21 '21 at 15:49
10

I wanted to point out that although this answer works great for older versions, it breaks under recent versions of Apache 2.4 with error code AH00526. ProxyPass and ProxyPassMatch or <Proxy> and <ProxyMatch> cannot be used together within the same worker name. This used to function just fine, so don't know if that was changed by design or if it's a bug.

Either way, you can fix this by only using a ProxyPassMatch with parameter 'timeout=120' (or whatever your desired value is), e.g.:

ProxyPassMatch ^/(.*\.php)$ fcgi://127.0.0.1:9001/path/to/webroot/$1 timeout=120
Community
  • 1
Stewart Adam
  • 379
  • 3
  • 6
7

I have Apache 2.4.6, but the patch to fix it is provided in Apache >= 2.4.8. The key here is to start your output immediately so that Apache (mod_proxy_fcgi) thinks the connection is active.

For example, I am using PHP and the DB query for my AJAX call takes > 30 seconds. Because I know that the overall response will be "Content-Type: application/json", I send that header immediately.

#1: Start output immediately
#Note: Sending the header is innocuous
#   it can be changed later using the $replace parameter
#   (see #3)
header( 'Content-Type: application/json' );

#2: Run slow query
mysql_query( "SELECT * FROM giant_table" );

#3: Change header as needed
header( 'Content-Type: application/csv', true );

#output content
Chris
  • 71
  • 1
  • 2
2

Shouldn't that be:

<IfModule mod_proxy.c>

Be sure the php.ini setting max_execution_time is set to 600 also. (check phpinfo() on the live page to make sure you're seeing the actual value used)

As Jenny said, set the php-fpm setting 

request_terminate_timeout 610s

(note the s at the end)

There's not much to configure with mod_proxy_fcgi itself, as you can see on the apache page. http://httpd.apache.org/docs/current/mod/mod_proxy_fcgi.html

Turn on php-fpm debug logging also so you can see where it times out there. http://php-fpm.org/wiki/Configuration_File (also turn on catch_workers_output)

And turn on debug level logging for the mod_proxy and mod_proxy_fcgi modules since you're using apache 2.4. Very nice feature, turn on just for the modules you need: http://httpd.apache.org/docs/current/mod/core.html#loglevel

If those don't help, post your php-fpm config file.

As a last resort, maybe some daemon is killing long running process?

troseman
  • 311
  • 1
  • 7
2

I noted that you are using PHP-FPM. I too use it, but with Apache 2.4.6.

Assuming that the issue has existed for some time, it seems to be that the timeout value for mod_proxy_fcgi is hard coded. I wrote up what I found here

Community
  • 1
misterich
  • 41
  • 5
1

Since you've fixed the timout settings in apache, that shouldn't be the issue. The second place to look would be any network equipment, but since you're proxying to your own server, that's also unlikely. So the remaining place to look is at the backend server.

Ih the config file for php-pfm, look for

; This is a hard kill switch on php execution.  It ignores the
; max_execution_time that can be set/changed with php_ini.  Basically
; it avoids timeout issues between apache and php-fpm.
request_terminate_timeout=30

This should be set to the same as, or slightly below, the timeout setting in apache.

Jenny D
  • 27,358
  • 21
  • 74
  • 110
  • 1
    I've set the `request_terminate_timeout` to 400, still no change :( I have a feeling that there's something I need to set for `mod_proxy_fcgi`, but it didn't seem to come with any configuration files. – wyqydsyq Apr 18 '13 at 23:13
0

I also have this issue. I am running on CentOS 7. This is how I fixed it:

vi /etc/apache2/conf.d/includes/pre_main_global.conf and set the following:

TimeOut 120
ProxyTimeout 120
Stuggi
  • 3,366
  • 4
  • 17
  • 34
Jon Tegu
  • 1
0

In addition to the timeout, set enablereuse=off. I found when it was on some requests to long running scripts would work correctly and others would be killed early.

ctlq
  • 131
  • 1
  • 1
  • 5
0

This post changed the whole deal for me.

It looks like the mod_reqtimeout of Apache would not be using the default value.

Add the following lines to your httpd.conf file:

<IfModule reqtimeout_module>
  RequestReadTimeout header=20-40,MinRate=500 body=20,MinRate=500
</IfModule>
Frosty Z
  • 193
  • 1
  • 8