Connection refused, restart!

[bettinz]

Hello, this is my VPS:
Cloudlinux 6.5 with CageFS and cPanel. Litespeed 4.2.11

When I enable Process Group in /usr/local/apache/conf/includes/pre_virtualhost_global.conf with this:
<IfModule LiteSpeed>
LSPHP_ProcessGroup on
LSPHP_Workers 10
</IfModule>
I see a lot of random error in error_log like this:
2014-06-05 19:35:23.947 [INFO] [uds://tmp/lshttpd/APVH_xxxx_Suphp.sock] Connection refused, restart!
2014-06-05 19:36:23.949 [INFO] [uds://tmp/lshttpd/APVH_xxxx_Suphp.sock] Connection refused, restart!
2014-06-05 19:37:23.916 [INFO] [uds://tmp/lshttpd/APVH_xxxx_Suphp.sock] Connection refused, restart!
2014-06-05 19:38:24.073 [INFO] [uds://tmp/lshttpd/APVH_xxxx_Suphp.sock] Connection refused, restart!

With different sites, and at random (sometimes they work without errors, sometimes connection refused).
I use zend opcache with 8MB for cache, phpsuexec yes with maxconn = 10.
In external app, max conn = 35, PHP_LSAPI_MAX_REQUESTS=1000, PHP_LSAPI_CHILDREN=35

Autostart: CGI Daemon, Run on startup: Suexec Daemon. If I set LSPHP_ProcessGroup on to off, error disappear.

Thank you

 

[mistwang]

That's normal messages, server detects that there is no PHP process listen on those domain socket, and start a new set.
As long as the site working fine, nothing to worry about.

 

[bettinz]

Thank you, the problem is the browser sometimes show an "unable to connect" during that messages

 

[mistwang]

Please try the latest 4.2.12 build.

/usr/local/lsws/admin/misc/lsup.sh -f -v 4.2.12

 

[bettinz]

Now the message is back: i'll update this post if something doesn't work as expected

 

[bettinz]

Hello, so:
4.1.12 and Process Group enabled, the load was around 1.30 (normally is 0.30, every time below 1). The message is still present, but it appear to work.
4.1.12 without Process Group and Suexec Daemon Enabled, load 0.30-0.50

4.1.11 and Process Group enabled: load between 0.40-0.50 but errors as explained before
4.1.11 without Process Group and Suexec Daemon Enabled: 0.30-0.50

 

[NiteWave]

should be 4.2.12 / 4.2.11 instead of 4.1.12 / 4.1.11

the higher load looks reasonable. it's the price of Process Group mode.
in suEXEC Daemon mode, only one root lsphp process will keep running there
in Process Group mode, each account has parent lsphp process running which is ready to fork child processes(though if it's idle too long, it'll be killed)
usually how many parent lsphp processes are running on your server ?

 

[bettinz]

I'm sorry for the mistake, it's 4.2.x of course.
This is a webserver with 45 accounts, usually less than 10 running at same time.
I've note the difference in load from 4.2.11 and 4.2.12 with process Group: it seem very high with .12

 

[NiteWave]

if you have multiple cores, for example 8 cores, load 8 is normal. in this case, load 1.3 still low.

so far by your tests, 4.2.12 + Process Group load is higher than 4.2.12 + Suexec Daemon, 1.3 vs 0.3, about 4 times higher.
need more load tests, is it always 4 times ? for example when suEXEC Daemon go into 2, will Process Group go into 8 or just 3 ?
it'll help us to understand it's a big problem or small problem or normal.