502 error with one user online

k i think i can reproduce the error if you create a dummp 503.php file in web root i.e. public/503.php with just

Code (Text):

<?php http_response_code(503) ?>

and refresh the page at domain.com/503.php - gives 502 bad gateway. Seems it's because php-fpm looks at 503 as php being down which is what 502 bad gateway reports

---

 

well haven't figured out the solution but maybe this work around might work, basically rewrite the 502 bad gateways to 503 maintenance page so whenever php is down with 503 error, visits get a nginx maintenance page instead

you can take advantage of existing include file /usr/local/nginx/conf/503include-main.conf and uncomment it in nginx domain vhost file

and change it to look like this

Code (Text):

    # only uncomment if you do not need to exclude images or js
    # from 503 redirect rewrites
     #include /usr/local/nginx/conf/503include-only.conf;
     error_page 503 @maintenance;
     error_page 502 @maintenance502;
     location @maintenance {
          rewrite ^ /maintenance.html break;
     }
     location @maintenance502 {
          rewrite ^ /maintenance502.html break;
     }

then go into public/ directory for domain and make a copy of maintenance page as maintenance502.html

Code (Text):

cd /home/nginx/domains/newdomain.com/public
cp maintenance.html maintenance502.html

so when site and php go into 503 error site unavailable status, static pages returns the maintenance.html page for 503 and maintenance502.html for 502 bad gateway pages.

Not really a permanent solution but for this issue with multiple php-fpm pools enabled, might do for the time being but that would put entire site into maintenance mode page unaccessible to even admin users heh - unless admins logged in do not get 503 status from php which means admins would still have access

FYI, the 502 bad gateway comes from multiple php-fpm pools setting for fastcgi_next_upstream which defines which php-fpm error codes will trigger rotating to another php-fpm pool if one is down, i.e. if php-fpm pool2 is down and sees http 503 error returned by forum closure, rotate to pool3 and so forth

Code (Text):

fastcgi_next_upstream error timeout http_500 http_503;

if you changed to remove http_503 then it won't rotate on 503 php-fpm errors but that would defeat purpose of multiple php-fpm pools for scaling

Code (Text):

fastcgi_next_upstream error timeout http_500;

@Matt might have some ideas ?

 

503 status code shown in forum close will be seen by php as site not available and the
fastcgi_next_upstream setting i mentioned in previous post triggers

 

yeah when i have more free time will take a closer look at it

 

looks like pretty much what i suspected Re: multiple php-fpm pool upstream alternating 503 & 502 errors

 

FYI, i've implemented the workaround in post #23 on this forum itself as well as i am using multiple php-fpm pools. Basically, whenever 502 bad gateway errors occur, visitor gets the maintenance page message

 

if it's same issue discussed in this thread then it only happens with multiple php-fpm pools and the work around in this thread maybe used. It's what i use on this forum since it's using multiple php-fpm pools too.

use this solution quoted link above or just go back to using non-multiple php-fpm pool default setup for the upgrades only or 24/7

 

no as it's due to how multiple php-fpm pools need to work see https://community.centminmod.com/posts/39826/

 

you didn't revert php.conf include file changes as that is where the 502 comes from with regards to fastcgi_next_upstream

 

no to revert though as you edit php.conf include will not reference upstream = phpbackend

 

that would defeat purpose of multiple php-fpm pools as when php pool is unavilable it won't be able to use the next available php-fpm pool without http_503 defined - so you'd get 502 error anyway on the first unavailable php-fpm pool access