Litespeed + cPanel + Cloudlinux

#22
Looks like is working now !

Is easy to get the top ten cpu eaters:

# cat /proc/lve/list|sort -r -n -k5|head -10
1485 3 100 100 1356112822722
824 2 100 100 230866736639
526 3 100 100 150523405592
1231 2 100 100 90540145514
927 2 100 100 89125989547
719 2 100 100 86946923203
633 3 100 100 69499024119
1218 2 100 100 24445937978
1411 2 100 100 19176211971
872 2 100 100 18148676133

The big number is cpu usage but I don't know in what units is that, I guess is not cpu time in seconds :p
The big numbers are jiffies (Within the Linux 2.6 operating system kernel, since release 2.6.13, on the Intel i386 platform a jiffy is by default 4 ms, or 1/250 of a second. The jiffy values for other Linux versions and platforms have typically varied between about 1 ms and 10 ms.) spent by CPU on all the processes within LVE.

lveps -d would give you usage in percentage on per LVE bases.
 

optize

Well-Known Member
#23
I'm not convinced this is working correctly.

Perhaps this is just the way CL works, since I haven't used it on Apache, it's hard to tell.

Should it be rate limiting these processes so they don't appear over the x value of CPU? My value is set at 25%, yet top still shows people at 60%

PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND
26834 rixxopco 16 0 129m 24m 6508 S 60.5 0.3 0:02.17 lsphp5
 
#24
I'm not convinced this is working correctly.

Perhaps this is just the way CL works, since I haven't used it on Apache, it's hard to tell.

Should it be rate limiting these processes so they don't appear over the x value of CPU? My value is set at 25%, yet top still shows people at 60%

PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND
26834 rixxopco 16 0 129m 24m 6508 S 60.5 0.3 0:02.17 lsphp5
We are not limiting individual processes, but all of them together, for a site.
Could you show me what lveps -d shows?
Also, what are you settings in /etc/container/ve.cfg and how many cores does your server have?
 

optize

Well-Known Member
#25
PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND
27330 elvisman 15 0 123m 24m 6048 S 36.9 0.3 0:00.19 lsphp5

root@cp08 [~]# lveps -d | grep elvis
652 1 1 1elvismanelvisman 0% 18556 N/A
root@cp08 [~]#

top says 36%, lveps says 0% ?
 
#26
PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND
27330 elvisman 15 0 123m 24m 6048 S 36.9 0.3 0:00.19 lsphp5

root@cp08 [~]# lveps -d | grep elvis
652 1 1 1elvismanelvisman 0% 18556 N/A
root@cp08 [~]#

top says 36%, lveps says 0% ?
What does this script do? How are you testing it? Are you hitting some PHP script all the time?
TOP shows CPU usage since process starts
lveps shows the usage during the last few seconds
 
#28
This is a live web server, just going off of what's being served.
It is possible that the site is not serving dynamic content right now -- and that is why you see 0 for cpu usage
if you run lveps -- it will show you the number of jiffies (cpu ticks) spent by lve to serve that site.
Also, could you do lveps -p
and see if that process (isphp5) is in LVE? It is possible that only its children are.
 

tina

Well-Known Member
#32
The Version Management page of the LiteSpeed web admin does not show an upgrade link to 4.0.14, only a Force Reinstall link. We are running 4.0.13. Any ideas?
 

DanEZPZ

Well-Known Member
#37
To those of you that properly tested this, how did you get on?

We fired up a new server and put a few users on to test it, within a few minutes they had taken it offline.

The idea behind it is great, but it's not ready for real world use IMO.
 
#38
To those of you that properly tested this, how did you get on?

We fired up a new server and put a few users on to test it, within a few minutes they had taken it offline.

The idea behind it is great, but it's not ready for real world use IMO.
Did you change CPU and maxEntryProc limits, or did you run with default one?
What did you see in lveps -d?
 

DanEZPZ

Well-Known Member
#39
Hi Igor,

We exchanged a few emails a week or so ago, I'm Dan from EZPZ Hosting (as my forum name suggests).

The CPU limit is at 25 and procs is at 30.

A client ran a script designed to stress the server which just killed disk I/O. Obviously clients wouldn't be doing this in a "real-world" environment but it was enough to show us the protection isn't quite what we expected.

If you want to have a look at the setup and run the script you're more than welcome to have the root details.

Dan
 
#40
Dan,

Thanks a lot. Yes, I remember that.
We just figured out that Lite Speed doesn't take into account maxEntryProcs that LVE has.
So, if you site runs more PHP scripts then there is maxEntryProcs -- all extra scripts will run outside LVE -- dragging down the server.
You can control number of concurrent PHP scripts via LiteSpeed, on per user bases by setting PHP suEXEC Max Conns option.
Make sure this value is lower then maxEntryProcs.

In the future versions LiteSpeed will correctly reject (or queue requests) to server dynamic content once the site is over maxEntryProcs.
 
Top