Ubuntu Server Eucalyptus Testers Needed

[Technoviking]

From: http://fridge.ubuntu.com/node/1925

Koalas love eucalyptus, they spend three hours a day munching away on the sturdy plant. Likewise, Ubuntu 9.10 Karmic Koala loves Eucalyptus, the Open Source system for implementing on-premise private and hybrid clouds using the hardware and software infrastructure that is in place, without modification. This allows you to run your own private cloud on your own hardware and infrastructure. Sound interesting? It really is, and this a rocking new feature in the new Ubuntu Server edition.

As we build to release, we could really use your help to make sure that Karmic Koala’s Eucalyptus support is rock solid. This post outlines how you can test this functionality, and provide some valuable feedback.

What You Will Need

You need two machines, one of which has to be capable of handling [KVM](https://help.ubuntu.com/community/KVM). The following command will check to see if your CPU has the correct VT extensions for running KVM (though you might have to additionally enable VT in your BIOS):

egrep '(vmx|svm)' /proc/cpuinfo

Each machine will also need 1GB of RAM and 80GB of free space. Documentation for all this is provided here.

You will also need to download the latest daily Ubuntu Server ISO image and burn it to CD.

Performing The Tests

Testing the *Ubuntu Enterprise Cloud* support with Eucalyptus involves three steps: setting up the cluster, the nodes and activating the cloud. Let’s look at it in three easy steps:

The Cluster Machine

This machine will control the nodes, it does NOT need KVM support. You can install it via the Ubuntu Enterprise Cloud task in the installer and select Cluster as the type.

Step-by-step instructions are here.

The Nodes

After you have installed a controller you are ready to add *nodes*. This is the machine that needs the KVM support. Install it via the *Ubuntu Enterprise Cloud* task in the installer and select *Node* as the type.

Step-by-step instructions are here.

Activating Your Cloud

After you have got the cluster and the node all installed and ready to go you’re ready for the final steps which are available here.

Testing and Filing Bugs

Each of these steps should be relatively pain free, after that you’re ready to start testing Eucalyptus.

The Eucalyptus Getting Started Guide contains commands you may want to try. Please Note: the Getting Started Guide is for version 1.5.2. Karmic Koala includes version 1.6, so there are some differences involved. You can however take a look to the on going work of the version 1.6 documentation.

Bugs should be filed in Launchpad in the Ubuntu eucalyptus package. You can see this list of bugs at https://bugs.launchpad.net/ubuntu/+source/eucalyptus. Bugs should be reported using the ubuntu-bug tool. This tool is shipped with Ubuntu Server. To file a bug, simply type in:

ubuntu-bug eucalyptus

This tool will send relevant debugging content to Launchpad to help identify and resolve the bug. More details on ubuntu-bug can be found here.

Discussion and Getting Help

Discussion about Eucalyptus can be posted directly to the ubuntu-devel mailing list and you are welcome to join the server development team in the #ubuntu-server IRC channel on irc.freenode.net.

[Mewto]

The thought about spending time on a eucalyptus cloud sounds appealing to me. Do you still need more testers? I guess you do otherwise the note would not be still up here.

I am putting together a system with AMD Phenom II X4 CPU. I am pretty sure it would support KVM. It is not up and running yet so I cannot verify it. Been putting it off for a week, now it is time to get it up. The computer I am on now can be the console node. I don't believe it supports KVM

egrep '(vmx|svm)' /proc/cpuinfo

returns nothing. I'll check the BIOS when I get off but I doubt it.

Guess I'll get an email notice. Have a great day

[grenoml]

There have been a number of bugs found in eucalyptus 1.6 and I think ubuntu needs to upgrade to 1.6.1 or 1.6.2 before getting testers involved. The version released in karmic was a pre-release 1.6bzr... version.

[rek2]

see http://ubuntuforums.org/showthread.p...29#post8378529

[holastickboy]

My Athlon X2 64 3800+ supports VT, so I would imagine your Phenom 2 would do it quite nicely. I was under the impression that practically all cpus from Core 2 Duo to Athlon x2 supported VT or their brand specific version of it?

[munky99999]

My Athlon X2 64 3800+ supports VT, so I would imagine your Phenom 2 would do it quite nicely. I was under the impression that practically all cpus from Core 2 Duo to Athlon x2 supported VT or their brand specific version of it?

Yep. More or less.

Even the intel atoms in these netbooks support hypervisors.

Personally I'm currently trying to hijack a few computers in my class to be able to test out this. However they are occupied atm

PErsonally I just dont have the physical machines in order to test this at home

[IkeLewis]

I am going to be trying this. I have a bunch of computers sitting around that will work, and I have thousands of photos to try to backup, so it might be worth it to try the cloud...

[munky99999]

So I have the cluster running. I just need hardware to run the node. Everything seems to be setup and working well.

edit/
So tempted to build myself a cluster now.

100$ mini-itx mobo
8 gigsram
flash bios allows xeons and higher end core 2 duos.
gigabit nic
8 usb ports, 3 sata

Roughly 400$ per node. If only I really had a need for it.

[n1md4]

Hi all,

I'd like to get involved in testing, on a beginners level.

I've 2 servers that have the required hardware, and the UEC install CD.

I've followed the tutorial, which for Cluster and Node installation is pretty much point and click through the install. I accepted most defaults for the Cluster install, which appeared to run flawlessly. Installing the Node was not as successful.

The Cluster has 2 ethernet ports; eth0 public and eth1 private, eth1 is connected to a 24 port switch. The Node has 1 ethernet port, and is connected to the switch.

The Node breaks an automated install at attempting a DHCP acquisition with the Cluster Controller. Instead I set the Node with a static IP, and can connect to the box from the Controller via SSH, but public internet access is broken from the Node.

Ideally, I'd be looking for an idiot proof install, is this possible? Is anyone else having problems with this?

I'd like to help with testing, I'm keen, but I'm unsure at this point what I can offer. Let me know if there's a way I can assist.

Thanks,

[johnrusso]

I have been working on an Ubuntu Enterprise Cloud Configuration off and on since December.

I have an odd set of parameters, some resources and some time to devote to this. Let me describe the environment and if I can add to your testing effort, I would be delighted to lend a hand.

First, I have a number of older,non-VT enabled multi core Xeon servers. I have used 3 of them to set up separate Cloud (CRC), Walrus (WS3) and Cluster/Storage (CC+EBS) Controllers.

In the process, I discovered a small installation "bug" where a file on the Walrus controller does not install automatically.

Where I am stuck now, is that I have two types of hardware, the aforementioned non-VT machines and some very capable, multi-core VT-enabled machines.

The non-VT machines are abundant and not being used. I am trying to figure a way to utilize them. I have learned that XEN will work with them, and if I do a CentOS installation, install Xen and the NC Controller, I am in business. I am told by the Eucalyptus folks that the UEC environment relies on XVM and XVM requires VT.

I would like to offer them as a platform to configure an Ubuntu/XEN version of an NC Controller package. If this is something you would like to pursue, let me know.

The VT enabled machines are currently running CentOS, and are highly tuned for DBMS testing. I can make them available 80% of the time they are not in demand, but I cannot reconfigure them. Possibly dual-booting them is an option.

Let me know if any of this has any interest.

Kind regards,

John Russo