I have a brand new HP Proliant DL380G5 and Ultrium 920 SAS tape device connected to a P800 HBA in the server.
I want a bare metal backup and restore of a single ESX 3.5 server. I'm NOT concerned with the contents of the VM's.
I only want to be able to restore the HOST server from OBDR or from some sort of bootable tape or dvd and then have the settings for the existing VM's already set up.
I already have a method for booting the VM's and using a system imager to restore the OS within the VM's.
I've been told that a new install or scripted install is easiest but have been asked to give a reason why it's easier than a simple bare metal backup of the entire esx server.
The proliant can see the p800 & the ultrium 920 but the device isn't seen by the Yosemite Backup software I've installed. Are there additional drivers?
IF anyone has taken a different route other than VCB then please chime in.
This is what I do, I install ESX 3.0.2 then run this Post Installation script that sets everything up for me. The only thing it does not do is patch the server, and that is only because patches have to be approved before I apply them. Here is the script I use, if you decide to use it just change out my information for yours.
Hope this helps!
Since you are not too concerned with the VM's them selve's, I would recommend a re-install. The process I explained above has a replacement ESX server in about 20 minutes at most (without the patches).
Message was edited by: espi3030
If you bare metal backup a configured ESX server, when you do the restore you will get the same configuration. This includes the generated MAC Addresses for the Service Console and VMKernel NICS. This works fine if you are just going to restore a single system again in the event of a failure. However, it doesn't work for provisioning new hosts since they will all have the same MAC addresses.
Also, just deleting the vswif and vmknics and recreating them does not work, since it will automatically generate the exact same mac addresses again...
i'm of the opinion that if you lose an ESX server, a rebuild would probably be the best. In reality it only takes about 30 minutes to rebuild an ESX server. This will probably be a lot quicker than a restore, which takes time. Also, assuming the backup media you choose is available or not corrupt, who know's how long it could take.
just my two cents.
A kickstart of a single server takes about 30 minutes and for safety reasons you don't want the installer to see storage volumes with vmfs filesystems on them since it can possibly overwrite the vmfs volumes with the OS.
An image install over the same network takes less than 4 minutes. It's also possible to set up an image install to write to specific luns and not overwrite a VMFS volume with the OS. This is a real win if the server boots from SAN where the same HBA is used to access the OS volume and the VMFS volumes.
On servers that boot from SAN it possible to provision a new ESX server from an unconfigured image in under 2 seconds using cloning on the disk array. Like an image install this also avoids the possiblity of kickstart overwriting a vmfs volume.
vRanger will back up VMs but it will not backup an ESX server. The problem with an ESX server is that if you restore to the exact same hardware there is a good change it will work, but if the hardware is in anyway different then the restore will not work very well and you will end up having to reinstall the main issue is that the system setups up various internal settings based on the hardware footprint. Specifically there is a unique ID for just about every ESX server out there that is used by VC, etc. So a simple restore does not work.
If you do backup /etc, the main configuration store, then it can be used as a guide to redo the configuration but a kickstart file is really the best approach. It is faster, and will work every time across disparate hardware.
Edward L. Haletky
VMware Communities User Moderator
Author of the book 'VMWare ESX Server in the Enterprise: Planning and Securing Virtualization Servers', Copyright 2008 Pearson Education.
CIO Virtualization Blog: http://www.cio.com/blog/index/topic/168354
As well as the Virtualization Wiki at http://www.astroarch.com/wiki/index.php/Virtualization
I agree with what Edward is saying, for restoring a ESX host a kickstart method is the best approach.We also recommend as was said to make a copy of the etc directories so you have them for reference if needed.
I know you've already received some excellant answers. But I wanted to add one more option for those who stumble upon the thread -- partimage. It's a great solution for Unix partition backups -- time tested, reliable & compact.
If you like, give our tool Kleo a try. It's a graphical wrapper around partimage that provides a wizard interface through the process of indetifying which partition to backup, and locating and mounting remote shares to store your backups. Kleo is provided with our LiveCD, the Carroll-Net Server Recovery Kit which inlcudes hundreds of specialized server recovery tool.
Kleo & the Carroll-Net Server Recovery Kit are both free for any use