Directadmin ON Centos 6.4 Couldnot take more than 1 CPU CORE !

R

rnc

Verified User
Joined
Jun 8, 2013
Messages
72
hello
we have a ESXi server with 1240v2 CPU and 32GB of ram
i created a single VM and instaled a Centos6.4 64bit and after that Directadmin with custom Build 2 on this VM


now this VM cant take more tan i CPU power in normal operation and IO wait and load averages goes high

but in other opration ( for example ./build all d ) this VM can go as high as Server CPu power ( 4*3.2Gz )

this VM is configured with nginx with 4 Worker process and some optimization on mysql

please help
 
So what is the problem? You are not explaining it well. Directadmin has nothing to do with the load on your server.
 
rnc said:
please help
Click to expand...
As chatwizrd points out, DirectAdmin has no direct control over other programs running on your server. The CustomBuild ./build command is a script running various compile and other programs, and whether or not they use multiple cores is up to them.

You don't say anything about your VM configuration. Are you using multiple physical cores? Or multiple logical cores? If multiple logical cores it's possible they're all running on the same physical core, in which case your problem could easily be that compilers and other tools are simply putting too much load on the physical core actually assigned to you.

Jeff
 
nobaloney said:
As chatwizrd points out, DirectAdmin has no direct control over other programs running on your server. The CustomBuild ./build command is a script running various compile and other programs, and whether or not they use multiple cores is up to them.

You don't say anything about your VM configuration. Are you using multiple physical cores? Or multiple logical cores? If multiple logical cores it's possible they're all running on the same physical core, in which case your problem could easily be that compilers and other tools are simply putting too much load on the physical core actually assigned to you.

Jeff
Click to expand...

Hello and thanks
1st
the problem is in normal operation i get high load vrage and cpu cant reach more than 1000 Mhz upto 2500Mhz !


2nd
i know about Vcpu and HT
server had 1230V2 (Quad Core (8Core HT )

for this reason i assigned just 3 Core (cuurent) or even 4 Core to Server ! no difference

there is no difference even when i assigned 8 CORE !

i set up 2 cpu 2 core , 4 cpu 1 core and and even 2 cpu 4 core !

but just the same Result!

and because of i checked hard disk in random access nad it can reach much more data rate than normal opration
the IO wait proble this could not be a IOPS issue

top - 23:56:23 up 49 min, 1 user, load average: 0.39, 0.59, 1.11
Tasks: 189 total, 1 running, 188 sleeping, 0 stopped, 0 zombie
Cpu(s): 0.7%us, 0.5%sy, 0.0%ni, 91.0%id, 6.1%wa, 0.2%hi, 1.6%si, 0.0%st
Mem: 16334472k total, 10755564k used, 5578908k free, 2184928k buffers
Swap: 12287992k total, 0k used, 12287992k free, 7615852k cached





i have sam situation in other servers but not this problem

even on a 2*6Core (24core HT) server
assigned 24Core to a single VM and it works without problem and could get as high as CPU core clocks permit ( i limited total Mhz to about 1Ghz below max Available ) to prevent ESX stability issue )

so what do you thins

is this possible somthing in directadmin or kernel prevent Vm to get more CPU clock or CPU core ?
 
and about the low load avrage on top post
the server is on lowest load time ( off peak )

this server used to install picture sharing scripts with more than 400Gb of picture files ( users upload pictures and get direct link to these pictures )
 
Do you use any raid on your hard drives at all? Iowait usually means that your drives are slow for some reason. I dont think its related to memory or cpu at all.
 
chatwizrd said:
Do you use any raid on your hard drives at all? Iowait usually means that your drives are slow for some reason. I dont think its related to memory or cpu at all.
Click to expand...

hello and thanks for response
no there is no raid and alsothe had good performance on normal and random rad/write benchmark as i described in first post
 
maybe I am reading it wrong but are you saying that even though you have restricted the VM to 1 core when the sites are under load its using all cores?
that sounds like a host/vm issue if I am understanding the issue right.
 
dmacleo said:
maybe I am reading it wrong but are you saying that even though you have restricted the VM to 1 core when the sites are under load its using all cores?
that sounds like a host/vm issue if I am understanding the issue right.
Click to expand...

thanks for response
not
even if i assign the VM all cores ! it cant use more than 1 Core
in our monitoring we see 4(8 HT) core of the Processor that just 1 of them are always at 100% utilization and other cores are idle ! there is no difference if i assign two core just 1 core of two core get used and if i assign 4 or even 8 core the same situation just 1 core used and we have load average above 3-5 all the time ( most of the time load average of 4-6 ) and server utilize just 1 core at 100%
 
rnc said:
thanks for response
not
even if i assign the VM all cores ! it cant use more than 1 Core
in our monitoring we see 4(8 HT) core of the Processor that just 1 of them are always at 100% utilization and other cores are idle ! there is no difference if i assign two core just 1 core of two core get used and if i assign 4 or even 8 core the same situation just 1 core used and we have load average above 3-5 all the time ( most of the time load average of 4-6 ) and server utilize just 1 core at 100%
Click to expand...

I say VM /server issue then.
how can it be anything else?
I am on dedicated with twin quad core xeons (so 16 with HT) and I show equal cpu usage across the cores, so DA itself works fine with multicore.
this has to be underlying server setup issue.
 
Hello,

If you're referring to the dataskq during a tally or backup, it's only 1 single process, so it won't use multiple CPUs.
For example, it does not fork off to run a thread for the tally of each User... as this would hog the CPU for other live services, which are more important than the background tally/backups, etc..

The "directadmin" process has many forked child processes, so each one will use one cpu (or core)... but each single request will only use 1.

John
 
You must log in or register to reply here.
Back
Top