The OpenStack deployment sometimes can be a bit overwhelming, because it has a lot of components that interact between each other and the deployment can include several servers that will be running the services mentioned in the previous post (cinder, horizon, swift, glance, etc).
Many production environments can be built, by one server to be used as a compute node (running nova and neutron services for example) and a controller node (running cinder and glance for example) and also several storage appliances included (ZFS appliance, iSCSI server or fiber channel storage), as a part of the architecture.
We can find several deployment scenarios and all depends on what we want to achieve and also our budget.
You can find here some interesting links about OpenStack Deployment requirements:
- Hardware requirements of OpenStack
There's a good way to test this virtualization solution without suffering from the implementation, and this is using all-in-one setups. This type of configurations provides an easy way to check this solution, most of the time providing virtual machines, with all the OpenStack services running in the same OS instance.
You can find here a useful link with this type of tools.
There's a Solaris all-in-one solution, provided by Oracle that is distributed as UAR file. UAR files (Unified Archives) are files created by the archiveadm utility that clones a Solaris instance, including any zone running in the OS. This provides an easy way to clone systems and accelerate Solaris deployments.
The UAR file with the Solaris OS, running OpenStack services, can be downloaded from http://www.oracle.com/technetwork/server-storage/solaris11/downloads/unified-archives-2245488.html
We will use the SPARC uar file, because our testing server will be SPARC box.
DEPLOYMENT
As I mentioned before our testing server will be a SPARC box, running Solaris 11.3.
root@t3-2:~# more /etc/release
Oracle Solaris 11.3 SPARC
Copyright (c) 1983, 2016, Oracle and/or its affiliates. All rights reserved.
Assembled 03 August 2016
root@t3-2:~# prtdiag -v | grep 'System'
System Configuration: Oracle Corporation sun4v SPARC T3-2
Sun System Firmware 8.3.11 2015/06/04 07:29
====================== System PROM revisions =======================
root@t3-2:~#
Virtual CPU'S: 256
root@t3-2:~# psrinfo | wc -l
256
root@t3-2:~# prtdiag -v | grep 'Memory'
Memory size: 261632 Megabytes
I was planning to use the kernel zone option, as is described in the following document
But in the moment I was trying to config and install the kernel zone, I was getting some errors
root@t3-2-syd04-b:~# zoneadm -z openstack install -a sol-11_3-openstack-sparc.uar -x install-size=50g
Platform does not support the kernel-zone brand.
zoneadm: zone openstack failed to verify
I realized that this server was not having support for kernel zones
root@t3-2:~# virtinfo
NAME CLASS
logical-domain current
non-global-zone supported
logical-domain supported
root@t3-2-syd04-b:~#
I also confirmed this information from this document:
Hardware and Software Requirements for Oracle Solaris Kernel Zones
The physical machine must meet the following requirements.
SPARC based systems:
A SPARC T4 series server with at least System Firmware 8.8.
A SPARC T5, SPARC M5, or SPARC M6 series server with at least System Firmware 9.5.
A SPARC T7 or SPARC M7 series server. All firmware versions are supported.
To run kernel zones, a Fujitsu M10 or SPARC M10 server with XCP Firmware 2230 or newer and Oracle Solaris 11.3 or newer.
Unfortunately, this is T3 series server (T3-2) and doesn't support kernel zones :( .
After this scenario I was thinking other options:
1) To choose alternative deployment mentioned in the document (the AI alternative). I disregard this option, because is not the easiest way to accomplish our purpose (fast deployment and testing), because I don't have an automated installer server (AI server) configured and this option makes a baremetal installation of the OpenStack solution.
2) I was thinking to create a non-global zone and install all the OpenStack services from the repository, this allows me to use already a server without using some much resources. This looks like a good option, but we have to configure a couple of things in order to have a working deployment.
3) Finally the last option was use a server that support kernel zones and continue with the original plan. I decided to choose this option, for the original scope of this post.
Points 1) and 2) can be worked in other post entries :).
0 comentarios:
Post a Comment