Nginx + php-fpm continuesly "execution timed out" ?

This error [502/504] usually means a Nginx configuration problem.

Means that Nginx is having issues passing events [communicating] to PHP-FPM.

Here is the real 'clue': ""GET /index.php") execution timed out"

But before I used a unix sockets too (fastcgi_pass unix:/tmp/php5-fpm.sock;)

I think there is something with timeouts. Maybe:
request_terminate_timeout = 120s
request_slowlog_timeout = 5s

When I comment those lines and rise pm.max_children = from 6 to 20 (I have 16gb memory) and pm.max_spare_servers = (from 5 to 15) the things going better.

In nginx logs I found:
582 recv() failed (104: Connection reset by peer) while reading response header from upstream,

*580 no live upstreams while connecting to upstream,

I need to check again but I did not have any timeouts anymore when i changed my php.ini max_execution_time max_input_time to a higher number.

But i did not have so much time to test this well.

i now started from scratch on a new machine.. and i still have this issue, so i mean it didn't had to do anything with the machine, so there must be something wrong with my setup or/and the php-fpm/nginx config. I'll run php-fpm in debug log now log_level = debug to see what will happen and i'll report here.

@superfedya did you solve it?

at fist sight that you're POSTing to /index.php. What's the request you're executing exactly? Does the site works with GET requests?

1. Increase max_requests.

2. Set the request_terminate_timeout to 30 minutes.

3. Check to see if all children are taken.

Could be that the FCGI setup permits long lived requests, but the php-fpm don't.

I've been bumping my head for days into this problem, did a lot of googling and fiddling around with different settings and I came across this post a number of times. I somehow got rid of the time-outs and I would like to share my experience. I'm not exactly shure what did the trick. May be my post will help solve your time-out problems.

I recently installed a D6.26 in a Freebsd jail with nginx 1.2.2 with php-fpm 5.2.17 behind a nginx reverse proxy in another jail. My purpose of the site is to test Drupal on a nginx platform. I use memcached and xcache instead of apc. Here are excerpts of php.ini php-fpm.conf and nginx.conf that ihmo are relevant to this problem:

php.ini

php-fpm.conf

nginx.conf

As you can see my time-out settings are short, below average. I'm building a dev environment without load, so I undoubtfully will have to raise limits in a production environment. In any case this setup somehow solved my immediate problems, where mysite.com/admin and mysite,com/admin/reports/status returned time-outs. So far so good.

Like I said, I'm not sure what did the trick, but I doubt it has to do with timeout settings. I think that once I turned mysql.allow_persistent off and mysqli.reconnect on things started working for me. Or maybe it's a combination of more of the above. I'll leave the matter for now, as I already spent a couple of days on cracking this nut, and have to jump one more hurdle in replacing $_SERVER["REMOTE_ADDR"] with a function which understands X-Forwarded-For, due to my reverse proxy setup.

Hopefully this post is of any use to you all.

So far I'm very impressed with nginx's speed and simplicity.

Cheers!

P.

I am unfortunately still experiencing problems. I receive 502 Bad Gateway every few minutes.

It does seem I have some long lived requests. I have tried tuning max_execution_time on php.ini and request_terminate_timeout on the pool config file.

in php.ini
request_terminate_timeout = 30m

^ This did "fix" the errors, but this doesn't really resolve the root of the problem. Can anyone explain how to find what script is running so long? What are the cons of having request_terminate_timeout such a large value (scripts are never terminated when they probably should be).

I have examined the nginx error log and have the following errors (multiple):

recv() failed (104: Connection reset by peer) while reading response header from upstream
upstream timed out (110: Connection timed out) while reading response header from upstream

None of my php5-fpm logs contain any relevant errors.

yes i use elysia cron and regular crontabs (every minute) to execute all crons; everything is spread evenly to avoid excessive load at a single time.

again, if i just increase:

fastcgi_read_timeout 14400;

won't that just be "hiding" the problem too? with these high timeouts, the problem isn't as obvious because 502 bad gateway errors don't occur. However, if i leave and return after 30 minutes in my case (when php.ini request_terminate_timeout = 30m), refreshing the page returns a 502 bad gateway error too!

It seems I have a script that is always running. Would google analytics possibly be the culprit? I'm just looking for any guidance or help for HOW to troubleshoot this. The logs just seem so inconclusive.