Anuket Project
CPU Utilisation case study Configuration
Host:
OS: Ubuntu 16.04
Disable apparmor
$ /etc/init.d/apparmor stop
$ /etc/init.d/apparmor teardown
$ update-rc.d -f apparmor remove
Isolate 2 cores from the OS by modifying: /etc/default/grub
GRUB_CMDLINE_LINUX_DEFAULT="text isolcpus=10,11"
Update grub:
$ sudo update-grub
Reboot
Install barometer build-essentials:
$ git clone https://gerrit.opnfv.org/gerrit/barometer
$ ./src/install_build_deps.sh
Install virtualization packages and other required packages:
$ sudo apt install qemu-kvm libvirt-bin virtinst virt-manager screen
Add the user to the libvirtd group:
sudo adduser $USER libvirtd
Create a dir to store ISOs
$ sudo mkdir /usr/local/src/images/
$ cd /usr/local/src/images/
$ sudo -E wget http://releases.ubuntu.com/16.04/ubuntu-16.04.2-server-i386.iso
Ceate a Test VM
$ sudo virt-install -n test -r 256 --disk path=/var/lib/libvirt/images/test.img,bus=virtio,size=4 -c /usr/local/src/images/ubuntu-16.04.2-server-i386.iso --network network=default,model=virtio --graphics vnc,listen=0.0.0.0 --noautoconsole -v
Connect to the console to complete the installation process.
Clone the latest libvirt code, build and install collectd:
$ git clone https://github.com/maryamtahhan/collectd.git mt_collectd
$ cd mt_collectd
$ ./build.sh
$ ./configure
Check that the virt plugin is configured:
virt . . . . . . . . yes
Install collectd:
$ sudo make -j install
Install collectd as a service:
sudo mv systemd.collectd.service collectd.service
sudo chmod +x collectd.service
vi collectd.service
Edit the collectd.service script
[Service]#ExecStart=/usr/sbin/collectd
ExecStart=/opt/collectd/sbin/collectd -C /opt/collectd/etc/collectd.conf
#EnvironmentFile=-/etc/sysconfig/collectd
#EnvironmentFile=-/etc/default/collectd
ProtectSystem=full
ProtectHome=true
Modify the /opt/collectd/etc/collectd.conf file
<LoadPlugin cpu>
Interval 1
</LoadPlugin>
<LoadPlugin virt>
Interval 1
</LoadPlugin>
LoadPlugin csv
LoadPlugin network
<Plugin cpu>
ReportByCpu true
ReportByState false
ValuesPercentage true
</Plugin>
<Plugin csv>
DataDir "/tmp/collectd/csv"
# StoreRates false
</Plugin>
<Plugin network>
# server setup:
Listen "<host ip address>" "28597"
<Listen "<host ip address>" "28597">
</Listen>
</Plugin>
<Plugin virt>
Connection "qemu:///system"
# RefreshInterval 60
# Domain "name"
# BlockDevice "name:device"
# BlockDeviceFormat target
# BlockDeviceFormatBasename false
# InterfaceDevice "name:device"
# IgnoreSelected false
HostnameFormat uuid
# InterfaceFormat name
# PluginInstanceFormat name
Instances 1
ExtraStats "disk pcpu"
</Plugin>
Core 10 and 11 were isolated from the OS using the isolcpus argument to grub command line and using virsh to Pin the vCPUs to those cores (Requires KVM acceleration):
$ virsh vcpupin test 0 10
$ virsh vcpupin test 1 11To check for KVM acceleration see that the response to the command below is > 0:
$ cat /proc/cpuinfo | egrep -c '(vmx|svm)'
without KVM acceleration you will need to taskset the qemu threads to the isolated cores:
$ grep pid /var/run/libvirt/qemu/test.xml
<domstatus state='running' pid='4907'>
$ grep Cpus_allowed_list /proc/4907/task/*/status
/proc/4907/task/4916/status:Cpus_allowed_list: 4
/proc/4907/task/4917/status:Cpus_allowed_list: 5
/proc/4907/task/4918/status:Cpus_allowed_list: 6
/proc/4907/task/4919/status:Cpus_allowed_list: 7
</section>
To taskset the process + tasks repeat the following for the PID and the subtasks:
$ taskset -pc 4 4907...
Guest
OS: Ubuntu 16.04.2 LTS
Install collectd and stress
$ sudo apt install collectd stress stress-ng
configure /etc/collectd/collectd.conf
Hostname "ubuntu_vm"
<LoadPlugin cpu>
Interval 1
</LoadPlugin>
LoadPlugin network
<Plugin cpu>
ReportByCpu true
ReportByState false
ValuesPercentage false
</Plugin>
<Plugin network>
# # client setup:
Server "<HOST IP>" "28597"
<Server "<HOST IP>" "28597">
</Server>
</Plugin>
Start collectd on the host and the guest
$ service collectd start
Run Stress in the VM on a single core
$ taskset 1 stress -c 2 -t 10
stress: info: [1422] dispatching hogs: 2 cpu, 0 io, 0 vm, 0 hdd
stress: info: [1422] successful run completed in 10s