Sun Ray in OpenSolaris 2009.06 on Amazon EC2

While running Sun Ray on OpenSolaris might not be supported as of today, it most certainly does work! My latest project was getting a Sun Ray server set up as an Amazon Machine Image (AMI) in the Amazon Elastic Compute Cloud (EC2). Doing this is fairly simple, as there is already an OpenSolaris 2009.06 AMI (two, actually – one 32bit, the other 64bit – depending on the size of the instance you want) in the list that you can choose from.

After launching the new instance, all you really need to do is to download Sun Ray Software 5 EA1 and then follow this configuration guide. If you’re attempting to enable NSCM, RHA and the other components that need Motif libraries, follow the OpenSolaris 2009.06 appendix. Otherwise, just read and follow the page.

Notes for EC2 Instances

\*For EC2 instances, there is no need to play with NWAM – it is disabled by default, which makes everything just that much nicer.

\*When running utadm, I used the private IP that was assigned to the instance, instead of it’s public address. You should also use the private IP between Sun Ray servers if setting up a FOG.

ex. “utadm –A” if your IP is in this range. //Edit 2009-08-15, 0030 - changed "-a" to "-A"

One requirement that I was playing with was how to allow a Sun Ray to use multiple Sun Ray servers without needing to configure VPN, although this is not a hard step (as demonstrated here). The easiest way I came up with was to instead of put a single server in the “Servers” menu of the Sun Ray firmware, put in a DNS entry that had multiple entries for Sun Ray servers. I created a subdomain “”, which points to the two Sun Ray servers that I set up so that both are available to be used by the Sun Rays.

Multiple DNS records for Sun Ray servers

Downsides of Multiple A Record method:

Now there is a downside to this method of foregoing VPN and simply using multiple A records in DNS. With this, you’re not guaranteed to be connected to the same Sun Ray server again if your Ray disconnects for any reason. In addition, you might need to reset your Sun Ray multiple times if a server goes down; nothing prevents the Sun Ray from skipping the offline server.


If the Sun Rays were VPN’ed into your Amazon “network”, the full Fail Over Group (FOG) experience would be there, and it would automatically fail you over to a different server if the one you were on were become unavailable.

If you just want a simple set-up and to be able to use a Sun Ray anywhere in the world with a 1-line configuration in the Ray’s firmware, you now know what to do. Happy Sun Ray using!

Edit 2009-08-14 1223 EST - If your machine is running headless, you may need to append '0=inactive' to /etc/X11/gdm/custom.conf to silence GDM errors that may show up in your logs.


Is that utadm -a or -A? Also, I think FOGs still work even with multiple A records in DNS. If you connect to server 1, but have a session on server 2, server 1 will send you to server 2, by definition of having a FOG :) Although, getting the servers to recognize each other as being in the same FOG might take some more work since I don't know how Amazon EC2 would handle multicast, which is used by default for Sun Ray failover group communication.

Posted by William Yang on August 14, 2009 at 10:37 AM EDT #

Thanks for catching that, William. Yes, it's -A instead of -a.

Perhaps you're right about the FOG setup as shown above, but it doesn't work like that for me :) The DNS pointer in the Servers menu is simply a round-robin setup and will choose the next server in the list every time.

Posted by Stephen Repetski on August 14, 2009 at 05:34 PM EDT #

Doesn't this get pretty expensive to run? I'm not too familiar with Amazons E2 product, but as I remember they charge an hourly rate right?

Posted by Troy Truchon on August 16, 2009 at 08:09 AM EDT #

I think that the default IPs of Amazon EC2 instances are assigned by DHCP.
So did you configure Sun Ray Server as a dhcp client ?
Or you use Elastic IP ?

Posted by Kenji Funasaki on August 16, 2009 at 02:46 PM EDT #

@Troy: Yes, it can be fairly expensive. If you were to implement this for yourself, you'd have to take into consideration the number of people using it, when they'd be using it, and the other variables that would determine the price of what you paid for in the cloud versus buying a server.

If, for example, I ran a company with 200 concurrent Sun Ray users, running this in the cloud would probably cost much more than simply buying a handful of servers for the organization and running it internally. It all depends on the length of time you'd use it for (there are discounts for "reserving" instances for periods of 1-3 years), bandwidth, and "instance" size

Posted by Stephen Repetski on August 21, 2009 at 08:29 AM EDT #

@Kenji: The Sun Ray server is currently configured to give addresses via DHCP, but that's not necessary since all of the units that would be connecting to this are outside of it's local network. It would make sense that the instances are provided IPs by Amazon's DHCP servers.

Regarding your second point, the Sun Rays connect to the DNS hostname, which points to the Elastic IP I assigned the server.

Posted by Stephen Repetski on August 21, 2009 at 08:57 AM EDT #

Post a Comment:
  • HTML Syntax: NOT allowed

Welcome to my blog. My name is Stephen (Trey) Repetski, and I'm working at Sun for the summer. I graduated from TJHSST, and am headed to Rochester Institute of Technology (RIT) in the fall, studying Network Security and Systems Administration (NSSA)


« July 2016