Page 1 of 1

Use RAID1 to try new system on one disk

Posted: 2020/01/22 21:08:03
by PaulR
Hello

I have a workstation at home working as RAID1 virtual disk (not RAID1+0, but just two mirrored disks)
I'd like to test a new nvidia driver installation (using nouveau now) but last time i tryed i spent days just to recover my GNOME.
So, I'd like to use one of my physical disks to perform all operations on it, and leave the other one just as "backup" so I can rebuild my RAID1 virtual disk in case a failure happens (It will happen,I'm sure... :? )

My RAID controller is a PERC H310 on a Dell T7600. I have the OS tools (MegaCli and perccli,and also the BIOS configuration tool at boot) for this controller, but my doubt is how to proceed.
I think,I should take one physical disk A from the RAID so only disk B is left, do the tests on A, and then rebuild B using A disk if all goes well (reverse if not). But i don't know how to do that exactly, and how to say which disk must be used as "master" for the rebuild process, so its data are written to the other disk and not to the wrong one.
-In case of failure in new system on A disk: Will the RAID be automatically rebuilt if after taking out a disk A I just add it again?
-In case of success: How can i say RAID now use A disk to "refresh" the left B disk?

Any help is welcome.

Re: Use RAID1 to try new system on one disk

Posted: 2020/01/23 18:36:00
by aks
Uh? (probably missing something) Don't use RAID, use JBOD (Just Bunch Of Disks). Surely?

Re: Use RAID1 to try new system on one disk

Posted: 2020/01/23 18:50:58
by hunter86_bg
You have another option called 'BOOM boot Manager' which allows you to create an LVM snapshot and if necessary to boot from that and revert your system.
The only requirements are:
Everything to be under the '/' (/var is an exception)
To have free space for the system VG and of course to be using LVM.

Here is a nice blog:
https://www.redhat.com/en/blog/boom-boo ... -snapshots

Edit: It's available for RHEL/CentOS 7.5+

Re: Use RAID1 to try new system on one disk

Posted: 2020/02/13 18:39:27
by ron7000
I've done this, or had it done, on a work server that had two disks as raid-1 with the linux operating system on it.
There is a way to split the raid-1 array.
that will make it so your existing system will not squawk every time booting saying disk failure when the 2nd disk is removed to use on other things.

Otherwise you could simply do that, just yank the disk, reformat, and use elsewhere. And put up with the warning every boot of a faulted raid-1 array. When your done with that second disk using it wherever, simply delete the partition table on it to make it look like a new blank disk. When you plug it back in the raid controller should simply rebuild the faulted raid-1 array, if it doesn't automatically recognize it as the original 2nd disk which might mean you'll need to manually tell the raid controller to rebuild the raid-1 array based on the existing disk.

try web searching "split raid 1 array"; I don't have my notes handy from 5+ years ago when I did that.

be forewarned that this is risky, and have any data you care about saved elsewhere that is recoverable should you nuke a disk or completely mess up your raid array. Which is why i personally don't do raid-1 for OS disks since it's easy enough to reinstall from scratch and any "data" I keep on a seperate disk(s).

Re: Use RAID1 to try new system on one disk

Posted: 2020/02/13 20:13:57
by ronnie9ball
I have two additional suggestions:

1. use a cloning utility like Clonezilla (open source) or Acronis (paid) to image your entire filesystem to a USB drive or a network share. Then play to your hearts content. This way you can restore your image very quickly in case you scrub your system. On USB3 or 1GB networking, you can count on about a minute per GB transfer speeds. IMHO, this is far less risky than messing with the RAID system. If you have a reasonably small amount of data, you can probably fit your entire image on a 16GB flash drive! Of course most flash drives are slower than external HDD or external SSD.

2. Use rpmfusion's website to install the latest NVIDIA driver from their prepackaged rpms. They have all of the processes documented on their site. Using their packages and instructions, I never have to reinstall NVIDIA drivers because of KERNEL updates. This is FAR more simple and stable than years past. Plus their method has worked better than any other system I have used for installing NVIDIA drivers. I think I have tried them all for EL7 distributions.

Good luck!