Hello,
I'm posting this in the "Unix for Dummies" forum, since it is more of a theoretical question than an exact problem/fix I'm inquiring about.

I have a SunFire T2000 server with 4 internal hard drives, running Solaris 10. (So far so good

The company just purchased a large EMC SAN.

Now the question: Is it "possible" to essentially make those 4 disks be 4 LUNs on the SAN, remove the 4 drives internal to the T2000, and fully run (boot and operate) off of data stored within the SAN??

What the the pros/cons? Any high-level steps to doing this? In theory at least, it seems pretty do-able.

Thanks for any inputs!!!

Booting off the network means that any network problem you have will effect your system at the very core. I don't recommend it for production systems unless you like making things more unreliable.

Many years ago I was working somewhere and someone has configured all the /home user accounts across the network and so every we logged in and there was any slight network problem, it was impossible to work.

Networks are not perfect. They break. Cables get kinked. There are routing problems. There are protocol bugs.

It is best to remove the network unless you absolutely have to, for example, you have a web server front end and a database back on connect on the same LAN segment. Normally, these devices are in the same rack and use very reliable cables.

If you are in the same rack and have very good cable control, then you might consider it.

Othewise, caution is advised.

Thank you for the thoughts. You bring up some great points.

If it's something I want to play with on a test server at some point, I'm guessing it's "possible" then, putting EVERYTHING on the SAN?

See if the attached (legacy) document on Solaris Jumpstart is helpful.

Attached Files

Jumpstart.pdf (183.1 KB, 24 views)

The advantage is that you can have 100% standardization. Boxes running off of local drives get out of sync over time as changes are made (patching, break/fix work, installing/uninstalling software, etc.). If you boot everything off of SAN you can prevent that.

Also, if you're going to have a huge farm of something you can have them all be identical. For example, if you run a huge web site and want 1000 redundant blades all running instances of your webserver you could do it on SAN and have them all pointing to the image you want them to use.

After spending many hours trying to get clients machines backup up and running, I always defer to the KISS principle. "Keep It Simple Stupid". Basically, keep your core OS on a local disk. This may sound a little luddite but when things have gone badly wrong, you may need to take it down to minimum config and start to bring it back from the base. At a time like this your really don't want to have to start by fighting with SAN drives when you probably have a couple of nice local, cheap disks which will be connected directly to your server.