Resolved 503 issue on website

[stefanoostwegel]

Hi all, i have a very common problem i cannot seem to resolve.
My website returns a 503 error.

I have a wordpress 5.3 (latest version) on my webhost running:
OS‪CentOS 6.10 (Final)‬ProductPlesk Obsidian
Version 18.0.23, last updated on Jan 22, 2020 02:33 PM

Things i have tried to resolve the issue:

Wordpress (no results)

• Disable all wordpress plugins
• Enable all plugins
• Change theme without plugins
• Change theme wih all plugins
• Suspend the website (resolved the issue)
• Activated the domain (503 again)

Hosting settings (no results)

• Switchted from php 71, to 7.2 to 7.3
• Switchted from CGI, to FastCGI to FPM, on all PHP versions
• Restarted nGinx
• Restarted PHP all version
• Toggle proxy settings of nginx

Tried the following KB articles from plesk, no match, no result (none of the KB articles applied here)

• WordPress site shows 503 Service Temporarily Unavailable error
• WordPress website shows 503 Service Temporarily Unavailable on Plesk for Linux: AH01075: Error dispatching request to
• WordPress site shows 503 Error with LSPHP enabled
• Website shows 503 Service Unavailable: Unable to allocate bytes in shared memory

Furthermore this would be good to know:

• The problems do not occur on any other website that runs the same wordpress and PHP settings.
• I can visit the website on my computer and update it, after i press ctrl+f5
• I have Cloudflare currently disabled, but enabling it (or development mode) doesn't change
• In none of the logfiles (plesk, php, domain, ngninx) there is any error listed at the time fo the 503 error
• The Apache & php-fpm memory usage sometimes spikes, but nog more than 50% of the complete memory
  • No memory shortage on the server
• I have restarted the server, services and components as often as i can

I don't know where to look at and what to look for anymore.

Please, any suggestions??

 

[Bitpalast]

Please provide the lines from /logs/proxy_error_log that correspond to the error, e.g. shortly after you tried to open the page.

 

[stefanoostwegel]

Hi Peter,

im completly taken by suprise that the problem is suddenly resolved.
After i posted this message i haven't done anything and it got away
For now, i mark this as read and complete.
Whenever i get back into the error i will try to find the logs you mention when the error occured.

Thank you

 

[stefanoostwegel]

Hi peter,

To my surprise the error comes and goes as it will.
I have checked the errors in the prose error log and found out that no errors are made on that specific time.
It seems like there is no traffic to be logged.
I verrified:

• nginx access
• apache ssl/tls access
• nginx error
• apache access
• apache error
• nginx ssl/tls access

To my surpise i do see quite a lot of access logs, from my very own server.
In the process list i see quite a lot of processeses on this domain, on this subscription where the error occurs.
The 503 error does not occur on any other domain or subscription.
I have killed all PHP-CGI processes using command

kill $(ps aux | grep '[p]hp-cgi' | awk '{print $2}')

as soon as i visit the website again, all these processes come up. 15 at once. I'm not quite sure how this happens.

Maybe it has something to do with the Fcgid Max settings i have:

• FcgidMaxProcesses 50
• FcgidMaxProcessesPerClass 15

But, that doesn't explain why i have this error, especially when im fully runing on php7.3 FPM.
I cannot see increased traffic to my website externally.
Rebooting the server, restarting the HTTPD, PHP and nGinx servers doesn't seem to have effect.

Do you have any clue?
Maybe im on track by seeing a lot of traffic comming from my own servers in the access logs?

[edit]
I have found a KB that results in the display of own IP address in stead of the clients.
I followed the instructions but it didn't work. A lots of entires are visible under the local servers IP address:

Server IP instead of client's is shown in Apache access log for website hosted on Plesk server due to disabled "remoteip" Apache module

Applicable to: Plesk Onyx for Linux Symptoms The domain's access log in Domains > example.com > Logs > Apache access (the file is /var/www/vhosts/example.com/logs/access_log) contains...

support.plesk.com

[/edit]

[edit2]
I have the httpd version of 2.2, not 2.4.

httpd

2.2.15-69.el6.centos

Wondering why plesk doesn't update this.
Would it be save to update this over yum?
Could be the resolution to my error?
[/edit2]

 

[dopeboy]

afaik FcgidMaxProcessesPerClass 15 means 15 req/s increase that and let's hear the result

 

[stefanoostwegel]

Eventually i have had it resolved by installing the FPM agent, which wasn't installed eventhjough it was used by PHP.
Furthermore i was able to resolve it by whitelisting my servers IP addresses in fail2ban.
Eventually i extended the RAM with 1 more GB and everything works smooth.

Thank you for all your help!