Prevent nginx 504 Gateway timeout using PHP set_time_limit()

There are several ways in which you can set the timeout for php-fpm. In /etc/php5/fpm/pool.d/www.conf I added this line:

request_terminate_timeout = 180

Also, in /etc/nginx/sites-available/default I added the following line to the location block of the server in question:

fastcgi_read_timeout 180;

The entire location block looks like this:

location ~ \.php$ {
    fastcgi_pass unix:/var/run/php5-fpm.sock;
    fastcgi_index index.php;
    fastcgi_param   SCRIPT_FILENAME $document_root$fastcgi_script_name;
    fastcgi_read_timeout 180;
    include fastcgi_params;
} 

Now just restart php-fpm and nginx and there should be no more timeouts for requests taking less than 180 seconds.

Try this link, it has a better solution on how to fix this. So the steps are:

1. Open your nginx.conf file located in /etc/nginx directory.

2. Add this below piece of code under http { section:

   client_header_timeout 3000;
   client_body_timeout 3000;
   fastcgi_read_timeout 3000;
   client_max_body_size 32m;
   fastcgi_buffers 8 128k;
   fastcgi_buffer_size 128k;
   

   Note: If its already present , change the values according.

3. Reload Nginx and php5-fpm.

   $ service nginx reload
   $ service php5-fpm reload
   

   If the error persists, consider increasing the values.

The correct answer is increasing fastcgi_read_timeout in your Nginx configuration.
Simple as that!

You can't use PHP to prevent a timeout issued by nginx.

To configure nginx to allow more time see the proxy_read_timeout directive.

 sudo nano /etc/nginx/nginx.conf

Add these variables to nginx.conf file:

http {  
  # .....
  proxy_connect_timeout       600;
  proxy_send_timeout          600;
  proxy_read_timeout          600;
  send_timeout                600;
}

And then restart:

service nginx reload

There are three kinds of timeouts which can occur in such a case. It can be seen that each answer is focused on only one aspect of these possibilities. So, I thought to write it down so someone visiting here in future does not need to randomly check each answer and get success without knowing which worked.

1. Timeout the request from requester - Need to set timeout header ( see the header configuration in requesting library)
2. Timeout from nginx while making the request ( before forwarding to the proxied server) eg: Huge file being uploaded
3. Timeout after forwarding to the proxied server, server does not reply back nginx in time. eg: Time consuming scripts running at server

So the fixes for each issue are as follows.

1. set timeout header eg: in ajax

   $.ajax({
       url: "test.html",
       error: function(){
           // will fire when timeout is reached
       },
       success: function(){
           //do something
       },
       timeout: 3000 // sets timeout to 3 seconds
   });
   

2. nginx Client timeout

   http{
        #in seconds
       fastcgi_read_timeout 600;
       client_header_timeout 600;
       client_body_timeout 600;
    }
   

3. nginx proxied server timeout

   http{
     #Time to wait for the replying server
      proxy_read_timeout 600s;
   
   }
   

So use the one that you need. Maybe in some cases, you need all these configurations. I needed.

You need to add extra nginx directive (for ngx_http_proxy_module) in nginx.conf, e.g.:

proxy_read_timeout 300;

Basically the nginx proxy_read_timeout directive changes the proxy timeout, the FcgidIOTimeout is for scripts that are quiet too long, and FcgidBusyTimeout is for scripts that take too long to execute.

Also if you're using FastCGI application, increase these options as well:

FcgidBusyTimeout 300
FcgidIOTimeout 250

Then reload nginx and PHP5-FPM.

Plesk

In Plesk, you can add it in Web Server Settings under Additional nginx directives.

For FastCGI check in Web Server Settings under Additional directives for HTTP.

^{See: How to fix FastCGI timeout issues in Plesk?}

Don't forget to look into your php-fpm logs !

In my case on PHP 7.3 I encountered:

WARNING: [pool www] server reached pm.max_children setting (5), consider raising it

On /etc/php/7.3/fpm/pool.d/www.conf I had to raise the pm.max_children value from 5 up to 50 (I do pretty heavy local stuff sometimes...).

Note: be aware that it may use much more CPU !

Since you're using php-fpm you should take advantage of fastcgi_finish_request() for processing requests you know can take longer.

Using set_time_limit(0) is useless when using php-fpm or similar process manager.

Bottomline is not to use set_time_limit when using php-fpm, to increase your execution timeout, check this tutorial.

After high hours looking for answers with this problem, I see one thing.

There are one more layer into my application:

1. Load balancer (the problem was all the time here)
2. Webserver (nginx)
3. Application (php)

Escaping from make request to LB, I can see the success response on test with "sleep" in php.

For my case, I had all possible TimeOut set correctly around 60s, but still got that 504 Gateway timeout error, then I was wondering why it always timeout around 15s after checking the request in my browser console !

After checking my other Apache2 config files, it turns out that "15s" was just sitting there in the file qos.conf (from the Quality of Service module) in the TimeOut option ! so I just changed it to 30 and my problem was solved (in fact, the request just needed about 19s to finish, so it was too close). I hope this helps someone if nothing else works.

Go into server

sudo nano /etc/nginx/nginx.conf

Add configuration in the HTTP section for the status of 504 (Gateway Time-out) as under:

proxy_read_timeout 3600;

Then restart the server of Nginx and it's working fine

sudo systemctl restart nginx