Question Which factors determine the maximumconcurrent users in Plesk?

[Ayki]

Server operating system version

AlmaLinux 9.4

Plesk version and microupdate number

18.0.61

I am trying to find out upper limits of my Plesk server. The sytem specs are as below:

OS: AlmaLinux 9.4 (Seafoam Ocelot)
Productlesk Obsidian
Version 18.0.61
Server -32 cores/64 threads / 256 GB DDR4/NVme
--------
I am trying to configure nginx/apache/mariadb for maximum performance.
However under load testing The site displays nginx 502 errors such:
( jmeter 2x5K user login test over a 600 rump up time) or If the rump up time reduced with less users.

# peer closed connection in SSL handshake (104: Connection reset by peer) while SSL handshaking to upstream
# connect() failed (110: Connection timed out) while connecting to upstream.

Is there a way of increasing the concurency so that Plesk can handle more users as there are plenty of idele resources on the server CPU and RAM?

Many Thanks

 

[Ayki]

I found the information in the blog post below very useful in troubleshooting the issue regarding nginx error.

504 Gateway Timeout Error: Expert Tips to Fix It

If you receive a 504 Gateway Timeout error while browsing, it’s likely to be the result of a communication problem between your device and the server hosting the affected site. That may involve your internet service provider (ISP) too. But while a 504 Gateway Timeout error may disrupt your...

www.plesk.com

Furthermore below article provided more guidance on how fast cgi can be limiting the concurrent connections. I was under the impression if apache FPM is deployed fastCgi is redunadant but may be this is not true.

https://support.plesk.com/hc/en-us/articles/12377652235159-Website-on-Plesk-server-is-not-accessible-504-Gateway-Time-out-The-timeout-specified-has-expired-client-203-0-113-2-54693-AH01075-Error-dispatching-request-to

Once the new directives are added and services restarted the error seem to have disappeared.

The 10 K jmeter test fully sucessful with no abondant connections - The server response time appers to have slowed down when there are excessive number of site traffic. The number of hits are nearly 50K per minute at peak

Just to makes sure the hit count is accurate cross referenced this with site logs. The data is consistent and shows around 40K transactions per minute at peak as per jmeter test instructions.