ESX Partitioning - Scripted Install - need some clarification

[davismisbehavis]

I am about to embark on upgrading some ESX 3.0.2 hosts to ESX 3.5 U4 using scripted installs via the ESX deployment Appliance.

Below is the section of the script that does the initial partitioning, I have a couple of questions and I'm basically looking for peace of mind.

1. As I understand it c0d0 is the first set of disks local to the server?

2. Will the clearpart command remove all existing partitions on that first disk

3. I am detaching the host from the SAN (Physically) will any other local VMFS partitions on the host be safe, i.e. is not wiped by the installation. We have some hosts that have templates and business continuity clones held locally.

4. In the final line of the script we create a very small VMFS partition, in a manual install you can tell it to use all remaining space, is there a command I can use to do this in a scripted installation?

clearpart --all --drives=cciss/c0d0 --initlabel

part /boot --fstype ext3 --size 250 --ondisk=cciss/c0d0

part swap --size 1600 --ondisk=cciss/c0d0

part / --fstype ext3 --size 5120 --ondisk=cciss/c0d0

part /var --fstype ext3 --size 4096 --ondisk=cciss/c0d0

part /tmp --fstype ext3 --size 2048 --ondisk=cciss/c0d0

part None --fstype vmkcore --size 100 --ondisk=cciss/c0d0

part None --fstype vmfs3 --size 1 --grow --ondisk=cciss/c0d0

Any help you can give would be much appreciated

[davismisbehavis]

I've had a look in HP insight manager and can see the second set of disks are seen as c0d1, so I am presuming I'm safe and that I only need to concern myself with the logical drive that is known as c0d0

[SimonLong]

Your configuration looks find to me, but if you are a little worried about it it might be worth thinking about doing a test install using VMWare Workstation. It's pretty simple and straight forward, but it just allows you to get it wrong without causing any major prolems.

I have writen a post about deploying ESX3.5 with EDA which can be found here: Deploying ESX3.5 using EDA v0.87 inside VMware Workstation 6.5

The only difference is i used sda instead of c0d0, Let me know if you need a hand with anything

Simon

[davismisbehavis]

Hey thanks Simon

I've done quite a lot of testing of it with an HP proliant 380 G3 and happy with the overall config, the 380 G3 was not connected to any shared storage and only had a raid 1 mirror of 2 x 18GB disks. This was one of the reasons I was concerned about what would happen with the other local drive.

I might go and break that mirror and create 2 single drives which should represent sda and sdb or c0d0 and c0d1 and test it again. That would set my mind at ease!

Do you have any idea what the answer could be to question number 4, during testing I ended up with a very small VMFS partition which was no use to man nor beast. the first logical drive is made up of 2 x 72GB drives so there are about 55GB to play with after the ESX build which i'd like to utilise.

does the grow switch allow me to use something to dynamically grow the VMFS? I thought you could only do that in the new vSphere!

[SimonLong]

Without Testing i would say, yes it would grow.

Looking around at some other KS scripts, which is basicly what EDA uses everyone else seems to be using the -grow switch. Have a look here and here

If i get a chance this afternoon i'll have a quick play with Workstation and see what happens

Simon

[dconvery]

You say you are performing an "upgrade". In this situation, all of your existing partitions remain as laid out originally. If you are doing a fresh install and overwriting everything, you should be OK, but the clearpart concerns me all the time. In theory, and according to the docs, it should work as expected. But your production VMs are on c0d1...You may be better off clearing them yourself manually just to be safe.At least make sure youhave a SOLID backup of all VMs and that you know how to restore if things go wrong. OrSVMotion them to the SAN. It is always a good idea to disconnect the SAN, but the theory about clearpart applies here too. The SAN drive should actually appear as sda, sdb, etc. If you are disconnecting the SAN because you do not trust the script, why risk it on the local stuff?

Dave Convery

VMware vExpert 2009

Careful. We don't want to learn from this.

Bill Watterson, "Calvin and Hobbes"

[davismisbehavis]

Hi Guys

Have been testing this out on a HP Proliant 380 G3 that I had spare, the scripted install only affected the c0d0 drive. I created a logical drive that was c0d1 and contained a VM and a template, once the build was complete the c0d1 partition remained and was unaffected and I was able to re-add the VM that was on c0d1 back into the inventory.

I found the best way of determining how the partitions have been named was to look in HP Insight manager, c0d0 and c0d1 were showing in there. When I did a df -k in the service console I could only see c0d0 and when I look at it within the VI Client you couldn't get that information at all.

I've also now found out that the following command, actually does give me the ability to create the final VMFS partiton and utilise all remaining space. The --size 1 is the minimum size of the partition and the --grow switch instructs it to use all remaining space.

part None --fstype vmfs3 --size 1 --grow --ondisk=cciss/c0d0

First development box gets upgraded tomorrow so I'm 99.99% happy that I now have a script that will do what I need it to do

[dconvery]

In order to "see" VMFS partitions, use the vdf command

Dave Convery

VMware vExpert 2009

Careful. We don't want to learn from this.

Bill Watterson, "Calvin and Hobbes"