9cb54690e9
This patch set makes the following changes: * using the new standards * Update doc/source/index.rst with new links * Move content of install-guide to the doc/source/install Change-Id: Ic16671191832d949d2ea0626dcbf334ba60c56ef
205 lines
5.4 KiB
ReStructuredText
205 lines
5.4 KiB
ReStructuredText
========================
|
||
Prepare a lab for murano
|
||
========================
|
||
This section provides basic information about lab's system requirements.
|
||
It also contains a description of a test which you may use to check if
|
||
your hardware fits the requirements. To do this, run the test and
|
||
compare the results with baseline data provided.
|
||
|
||
.. _system_prerequisites:
|
||
|
||
System prerequisites
|
||
~~~~~~~~~~~~~~~~~~~~
|
||
|
||
Supported operating systems
|
||
---------------------------
|
||
|
||
* Ubuntu Server 12.04 LTS
|
||
* RHEL/CentOS 6.4
|
||
|
||
**System packages are required for Murano**
|
||
|
||
*Ubuntu*
|
||
|
||
* gcc
|
||
|
||
* python-pip
|
||
|
||
* python-dev
|
||
|
||
* libxml2-dev
|
||
|
||
* libxslt-dev
|
||
|
||
* libffi-dev
|
||
|
||
* libpq-dev
|
||
|
||
* python-openssl
|
||
|
||
* mysql-client
|
||
|
||
Install all the requirements on Ubuntu by running::
|
||
|
||
sudo apt-get install gcc python-pip python-dev \
|
||
libxml2-dev libxslt-dev libffi-dev \
|
||
libpq-dev python-openssl mysql-client
|
||
|
||
*CentOS*
|
||
|
||
* gcc
|
||
|
||
* python-pip
|
||
|
||
* python-devel
|
||
|
||
* libxml2-devel
|
||
|
||
* libxslt-devel
|
||
|
||
* libffi-devel
|
||
|
||
* postgresql-devel
|
||
|
||
* pyOpenSSL
|
||
|
||
* mysql
|
||
|
||
Install all the requirements on CentOS by running::
|
||
|
||
sudo yum install gcc python-pip python-devel libxml2-devel \
|
||
libxslt-devel libffi-devel postgresql-devel pyOpenSSL \
|
||
mysql
|
||
|
||
.. _lab_requirements:
|
||
|
||
Lab requirements
|
||
----------------
|
||
|
||
+------------+--------------------------------+-----------------------+
|
||
| Criteria | Minimal | Recommended |
|
||
+============+================================+=======================+
|
||
| CPU | 4 core @ 2.4 GHz | 24 core @ 2.67 GHz |
|
||
+------------+--------------------------------+-----------------------+
|
||
| RAM | 8 GB | 24 GB or more |
|
||
+------------+--------------------------------+-----------------------+
|
||
| HDD | 2 x 500 GB (7200 rpm) | 4 x 500 GB (7200 rpm) |
|
||
+------------+--------------------------------+-----------------------+
|
||
| RAID | Software RAID-1 (use mdadm as | Hardware RAID-10 |
|
||
| | it will improve read | |
|
||
| | performance almost two times) | |
|
||
+------------+--------------------------------+-----------------------+
|
||
|
||
`Table: Hardware requirements`
|
||
|
||
There are a few possible storage configurations except the shown above.
|
||
All of them were tested and were working well.
|
||
|
||
* 1x SSD 500+ GB
|
||
|
||
* 1x HDD (7200 rpm) 500+ GB and 1x SSD 250+ GB (install the system onto
|
||
the HDD and mount the SSD drive to folder where VM images are)
|
||
|
||
* 1x HDD (15000 rpm) 500+ GB
|
||
|
||
|
||
Test your lab host performance
|
||
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
|
||
|
||
We have measured time required to boot 1 to 5 instances of Windows
|
||
system simultaneously. You can use this data as the baseline to check if
|
||
your system is fast enough.
|
||
|
||
You should use sysprepped images for this test, to simulate VM first
|
||
boot.
|
||
|
||
Steps to reproduce test:
|
||
|
||
#. Prepare Windows 2012 Standard (with GUI) image in QCOW2 format. Let's
|
||
assume that its name is ws-2012-std.qcow2
|
||
|
||
#. Ensure that there is NO KVM PROCESSES on the host. To do this, run
|
||
command:
|
||
|
||
.. code-block:: console
|
||
|
||
ps aux | grep kvm
|
||
|
||
#. Make 5 copies of Windows image file:
|
||
|
||
.. code-block:: console
|
||
|
||
for i in $(seq 5); do \
|
||
cp ws-2012-std.qcow2 ws-2012-std-$i.qcow2; done
|
||
|
||
#. Create script start-vm.sh in the folder with .qcow2 files:
|
||
|
||
.. code-block:: console
|
||
|
||
#!/bin/bash
|
||
[ -z $1 ] || echo "VM count not provided!"; exit 1
|
||
for i in $(seq $1); do
|
||
echo "Starting VM $i ..."
|
||
kvm -m 1024 -drive file=ws-2012-std-$i.qcow2,if=virtio -net user -net nic,model=virtio -nographic -usbdevice tablet -vnc :$i & done
|
||
|
||
#. Start ONE instance with command below (as root) and measure time
|
||
between VM’s launch and the moment when Server Manager window
|
||
appears. To view VM’s desktop, connect with VNC viewer to your host
|
||
to VNC screen :1 (port 5901):
|
||
|
||
.. code-block:: console
|
||
|
||
sudo ./start-vm.sh 1
|
||
|
||
#. Turn VM off. You may simply kill all KVM processes by
|
||
|
||
.. code-block:: console
|
||
|
||
sudo killall kvm
|
||
|
||
#. Start FIVE instances with command below (as root) and measure time
|
||
interval between ALL VM’s launch and the moment when LAST Server Manager
|
||
window appears. To view VM’s desktops, connect with VNC viewer to your
|
||
host to VNC screens :1 thru :5 (ports 5901-5905):
|
||
|
||
.. code-block:: console
|
||
|
||
sudo ./start-vm.sh 5
|
||
|
||
#. Turn VMs off. You may simply kill all KVM processes by
|
||
|
||
.. code-block:: console
|
||
|
||
sudo killall kvm
|
||
|
||
|
||
Baseline data
|
||
~~~~~~~~~~~~~
|
||
|
||
The table below provides baseline data which we've got in our
|
||
environment.
|
||
|
||
+----------------+--------------------------+---------------------+
|
||
| | Boot 1 instance | Boot 5 instances |
|
||
+================+==========================+=====================+
|
||
| Avg. Time | 3m:40s | 8m |
|
||
+----------------+--------------------------+---------------------+
|
||
| Max. Time | 5m | 20m |
|
||
+----------------+--------------------------+---------------------+
|
||
|
||
``Avg. Time`` refers to the lab with recommended hardware configuration,
|
||
while ``Max. Time`` refers to minimal hardware configuration.
|
||
|
||
|
||
Host optimizations
|
||
~~~~~~~~~~~~~~~~~~
|
||
|
||
Default KVM installation could be improved to provide better
|
||
performance.
|
||
|
||
The following optimizations may improve host performance up to 30%:
|
||
|
||
* change default scheduler from ``CFQ`` to ``Deadline``
|
||
* use ``ksm``
|
||
* use ``vhost-net``
|