Xentoo

For easier use with Cacti, it is easier to group statistics per type instead of per CPU. So you would have a parent OID for CPU time spent by user with many values (one per CPU).

Put the following in /etc/snmp/percpudetail:

#!/bin/bash
case "$1" in
  "user")
    grep ^cpu /proc/stat | awk '{print 2}' ;;
  "nice")
    grep ^cpu /proc/stat | awk '{print 3}' ;;
  "system")
    grep ^cpu /proc/stat | awk '{print 4}' ;;
  "idle")
    grep ^cpu /proc/stat | awk '{print 5}' ;;
  "iowait")
    grep ^cpu /proc/stat | awk '{print 6}' ;;
  "irq")
    grep ^cpu /proc/stat | awk '{print 7}' ;;
  "softirq")
    grep ^cpu /proc/stat | awk '{print 8}' ;;
esac

Make it executable:

chmod +x /etc/snmp/percpudetail

Then in your /etc/snmp/snmpd.conf file, put the following lines:

extend cpuuser /etc/snmp/percpudetail user
extend cpunice /etc/snmp/percpudetail nice
extend cpusystem /etc/snmp/percpudetail system
extend cpuidle /etc/snmp/percpudetail idle
extend cpuiowait /etc/snmp/percpudetail iowait
extend cpuirq /etc/snmp/percpudetail irq
extend cpusoftirq /etc/snmp/percpudetail softirq

Restart snmpd and you can poll those new OIDs.

I needed to export detailed CPU statistics from Linux servers using SNMP. While UCD-SNMP-MIB export some detailed stats, it only does it for the whole system. I may have missed something easier though :-)

So here is a BASH script, put it in /etc/snmp/percpustats:

#!/bin/bash
/bin/grep "^$1 " /proc/stat | sed -e 's/^cpu[0-9]* *//' | tr ' ' '\n'

Make it executable:

chmod +x /etc/snmp/percpustats

This simple script takes a single argument, which is the cpu id you want to look at. By “cpu id”, I mean the first word of cpu* lines in /proc/stat . So cpu, cpu0, cpu1, etc. Here is a sample content of my /proc/stat file:

# cat /proc/stat
cpu  190769457 225375 18212019 746796473 443933 1361 1240780 0 346283 0
cpu0 98979159 94876 9197600 368299173 282702 1361 1240721 0 164117 0
cpu1 91790298 130498 9014419 378497299 161231 0 58 0 182166 0

In your snmpd config file, you will need to add one extra config line per CPU you want to look at. Let’s say you have 3 CPU in your computer, you would need to add the following lines to get stats for the system in general (cpu) and details for each cpu (cpu0,cpu1,cpu2):

extend cpu /etc/snmp/percpustats cpu
extend cpu0 /etc/snmp/percpustats cpu0
extend cpu1 /etc/snmp/percpustats cpu1
extend cpu2 /etc/snmp/percpustats cpu2

Restart snmpd .

You can check the output using snmpwalk:

snmpwalk -v 2c -c yourcommunity yourip NET-SNMP-EXTEND-MIB::nsExtendObjects

You should get something like this:

NET-SNMP-EXTEND-MIB::nsExtendOutLine."cpu".1 = STRING: 52183124
NET-SNMP-EXTEND-MIB::nsExtendOutLine."cpu".2 = STRING: 2487538
NET-SNMP-EXTEND-MIB::nsExtendOutLine."cpu".3 = STRING: 64930047
NET-SNMP-EXTEND-MIB::nsExtendOutLine."cpu".4 = STRING: 13880645861
NET-SNMP-EXTEND-MIB::nsExtendOutLine."cpu".5 = STRING: 1503768
NET-SNMP-EXTEND-MIB::nsExtendOutLine."cpu".6 = STRING: 5350929
NET-SNMP-EXTEND-MIB::nsExtendOutLine."cpu".7 = STRING: 57971310
NET-SNMP-EXTEND-MIB::nsExtendOutLine."cpu".8 = STRING: 0
NET-SNMP-EXTEND-MIB::nsExtendOutLine."cpu0".1 = STRING: 3557627
NET-SNMP-EXTEND-MIB::nsExtendOutLine."cpu0".2 = STRING: 42604
NET-SNMP-EXTEND-MIB::nsExtendOutLine."cpu0".3 = STRING: 4803282
NET-SNMP-EXTEND-MIB::nsExtendOutLine."cpu0".4 = STRING: 1749637588
NET-SNMP-EXTEND-MIB::nsExtendOutLine."cpu0".5 = STRING: 27945
NET-SNMP-EXTEND-MIB::nsExtendOutLine."cpu0".6 = STRING: 1
NET-SNMP-EXTEND-MIB::nsExtendOutLine."cpu0".7 = STRING: 65021
NET-SNMP-EXTEND-MIB::nsExtendOutLine."cpu0".8 = STRING: 0
NET-SNMP-EXTEND-MIB::nsExtendOutLine."cpu1".1 = STRING: 5861096
NET-SNMP-EXTEND-MIB::nsExtendOutLine."cpu1".2 = STRING: 472784
NET-SNMP-EXTEND-MIB::nsExtendOutLine."cpu1".3 = STRING: 6472817
NET-SNMP-EXTEND-MIB::nsExtendOutLine."cpu1".4 = STRING: 1733427227
NET-SNMP-EXTEND-MIB::nsExtendOutLine."cpu1".5 = STRING: 426662
NET-SNMP-EXTEND-MIB::nsExtendOutLine."cpu1".6 = STRING: 918093
NET-SNMP-EXTEND-MIB::nsExtendOutLine."cpu1".7 = STRING: 10555401
NET-SNMP-EXTEND-MIB::nsExtendOutLine."cpu1".8 = STRING: 0

Now you can use these in your favourite monitoring/graph tools.

So you have your Unified Networking Lab (UNL) server running and you want to use arbitrary QEMU images with it. Here is how to create one.

All UNL data is stored under /opt/unetlab , images are stored under /opt/unetlab/addons and QEMU images are a folder deeper under /opt/unetlab/addons/qemu .

UNL expects to find images within directories named according to some scheme. The directory names must start with some keywords followed by anything you want. For Linux images, you need to name the directory linux-something, e.g. linux-ubuntu-14.04.5-amd64. You can find the list of supported image “types” in /opt/unetlab/html/templates. The names of the files are the prefixes needed for directory names. If you want to, you can edit those files to change the default parameters of the images (make a backup before you change anything).

UNL expects the disk image file to be named hda.qcow2 . You do not have any choice here.

Now, let’s create a Debian image.

Log on UNL server using ssh (username=root password=unl).

Let’s create a new directory for our image:

cd /opt/unetlab/addons/qemu
mkdir linux-debian-8.3.0-i386
cd linux-debian-8.3.0-i386

Now you have two options, either you create an image on another computer and copy its disk image on UNL server, either you create the image on the UNL server itself. I’ll explain how to do the latter.

Download a Debian ISO file from your favorite mirror:

wget http://cdimage.debian.org/debian-cd/8.3.0/i386/iso-cd/debian-8.3.0-i386-netinst.iso

Create a disk image for the virtual machine:

/opt/qemu/bin/qemu-img create -f qcow2 hda.qcow2 4G

Start the virtual machine manually with a ISO file as virtual CDROM, booting on that ISO file and with a network interface:

/opt/qemu-2.0.2/bin/qemu-system-i386 -m 256 -smp 1 \
-cdrom debian-8.3.0-i386-netinst.iso -boot d \
-hda hda.qcow2 -monitor stdio -vnc 0.0.0.0:10 -k fr-be \
-device e1000,netdev=net0,mac=50:02:12:34:00:00 \
-netdev tap,id=net0,ifname=debian0,script=no -S

You can now attach to the console (virtual screen) of the virtual machine using VNC (use the IP address of the UNL server and the screen number 10, e.g. 10.0.0.1:10). You are going to get a black screen, that is normal because the last parameter of the qemu-system-i386 command is -S which means freeze CPU at startup (basically pause). To unpause the virtual machine, type cont in the QEMU console you got.

QEMU 2.0.2 monitor - type 'help' for more information
(qemu) cont

You should see the Debian install menu.

With the netinst image, the virtual machine will need Internet access. You have multiple options to achieve this: * connect the virtual machine network directly to your physical network * route the traffic between your physical network and the virtual machine * NAT the traffic of the virtual machine on the UNL server

I am going to go with the third option.

You can use one of the default bridge interfaces that UNL creates or use a new one. Let’s use a new one, we will call it brdebian and assign it the IP address 10.0.0.1/24.

brctl addbr brdebian
ip link set brdebian up
ip address add 10.0.0.1/24 dev brdebian

UNL does not come with iptables installed but we need it to enable NAT.

apt-get install iptables

Let’s tell the kernel we want it to NAT everything that comes from the subnet we chose:

iptables -t nat -A POSTROUTING -s 10.0.0.0/24 -j MASQUERADE

and let’s tell the kernel it can forward IP packets:

echo 1 | tee /proc/sys/net/ipv4/ip_forward

Finally, let’s attach the virtual machine network interface to this network

brctl addif brdebian debian0
ip link set debian0 up

Assign an IP address from subnet 10.0.0.0/24 to the virtual machine, set its gateway to 10.0.0.1 and it will get Internet access.

Customize your image as you wish. When you are done, power it off, make a backup of the hda.qcow2 file (just in case).

You can now use that image in UNL.

Note that all the network changes we did are temporary and will disappear when you reboot the UNL server.