Abort request processing

[Mr_Parham]

Hi

My website is behind an nginx proxy and every now and then I get "upstream prematurely closed connection while reading response header from upstream". I checked the log file on my server and I found a lot of the similar looking thing to this in my log files

Abort request processing by PID:5486, kill: 1, begin time: 0, sent time: 0, req processed: 0

Can you please advise on what might have caused this to happen?

Regards
Parham

 

[Pong]

Did you enable some Anti-DDOS feature on LiteSpeed? It may block your Nginx proxy if it is over limit.

You may try to add nginx proxy IP to allowed list.

https://www.litespeedtech.com/support/wiki/doku.php/litespeed_wiki:config:cloudflare-ips-or-subnets

 

[Mr_Parham]

I already have all of those IP addresses in the whitelist so I don't think it's from that

https://www.screencast.com/t/k3n9G8YpZ

Regards
Parham

 

[Unique_Eric]

Hi Parham,

When issue happens, could you capture screenshot of real-time stats in lsws web admin, and top -c result. so that we have an idea where the bottleneck might be.

Here's some tip about how to improve PHP performance
https://www.litespeedtech.com/support/wiki/doku.php/litespeed_wiki:php:improve-performance

Best,
Eric

 

[Mr_Parham]

My server is pretty stable I would say, here is a screenshot of the results

https://www.screencast.com/t/v47YjVoSMSS

Regards
Parham

 

[NiteWave]

1st, it's [INFO] message, may not help you to identify issue of "upstream prematurely closed connection while reading response header from upstream"
2nd, you can try following

You can go to server->General tab, under General Setting table, there is an "External application abort", set it to "no abort".

to see if number of "Abort request processing" messages will decrease a lot.

 

[Mr_Parham]

Is there an option where I can choose not to abort only for trusted IP addresses? if now is that something that you might consider implementing?

Regards
Parham

 

[Pong]

whitelisted IP should not have any connection issue to reverse proxy. aborting requests are normal in general. I would suggest you bypass nginx to proxy when errors happen to see if the domain is running well by accessing to it directly to backend server. If so, then try to figure out why reserver proxy being blocked?

 

[Mr_Parham]

I have already tried that, when the proxy is getting blocked normal connections work perfectly fine and the server is stable. Keep in mind that the reserve proxy IP addresses are already whitelisted!

 

[Pong]

You may log a ticket with us with tmp root ssh access and detailed steps to test and we can check for you.