I use virt-install in a shell script to build my virtual machines. I want to know how to define multiple qcow2 files so that I can define software raid-1 during installation.
Any Ideas?
Gene

You can use the manual to see how to add an additional virtual disk to the VM. When you then install the OS, you can define raid like it is done when installing onto barebones.

If however, your virtual host is already running raid, it would be pointless having raid running in a raid environment.

Speaking of which, i recommend you let the vm host do the raid and just do a plain install. That way, the base system takes care of the mirroring. Ibwould imagine doing that on qcow2 would be a really big performance hit.

I want to try and raid-1 the entire disk as opposed to raid-1 by partition ( does this mean I'll partition the MD device??).
TIA
Gene

If you created a number of virtual hard drives and if your client has them attached, you could construct a software raid. If you don't want to mess with the client's host drive then just add in two more virtual hard drives to the client and use the clients mdadm tools.

Don't mix up host and client in this.

My host is running great using CentOS 6.8 and software RAID-1. I'll not mess with that.

Here's what I want to determine:
When I installed CentOS 6 on the host I implemented software raid-1 during the installation by doing by partition.
When I had a hard drive failure on /dev/sda the machine would not boot and I remembered that during the install the MBR went on /dev/sda - I had to reinstall

So I want to test if:
I raid-1 /dev/sda and /dev/sdb and call it /dev/md0_d1
When I get to the installation point where I'm installing the MBR and I install it to /dev/md0_d1
And later on /dev/sda dies will my system still boot with a degraded raid-1?

Hence, these questions.
I've added virtual disk drives to guests running on VMware, but the OS is already installed.
I want to have 2 virtual disk drives available before I install the OS

TIA
Gene

You should have (manually) installed to the MBR on /dev/sdb as well. This is well documented for RAID1 /boot.Nope. I doubt anaconda will allow it anyway. If you use the entire devices, there will be no MBR - it will be over-written. If you later run grub-install (or setup) you will over-write the metadata for the mdadm device. Nothing good can come from any of that.

I'm not a expert on this stuff, but isn't that what you do when running RAID-10 (putting entire disks together as raid-0 the taking 2 raid-0 devices making them raid-1)? What I want to do is a software equiv of a hardware raid environment (the MBR is installed on a /dev/sda that is really 2 physical disks). Is the electronics on the hardware raid controller managing the MBR?

Am I overthinking this stuff?

TIA
Gene

I've marked this as solved, but I am still working on it. 'Real' work has raised it's ugly head and it is taking all of my time.