softraid 2x SATA HDD and Proxmox 4.1 dont boot after RAID failure

[Paulo Pastoriza]

Hello.
I did a default installation of Proxmox 4.1 on a SATA drive (sda), and then follow the documentation (http://kbdone.com/proxmox-ve-3-2-software-raid/) to implement RAID1 by software.

By forcing the failed RAID, removing the cable from one of the disks, both the sda drive as sdb the system will not boot.

The grub menu appears, but the initram can not access the / that is in LVM.

Loading, please wait...
  Volume group "pve" not fount
  Cannot process volume group pve
Unable to find LVM volume pve/root

Has anyone had this same problem?
Can anyone tell if the kernel comes with support for raid and lvm enabled?

Thanks for any help

 

[windinternet]

At least if you followed the article you have the GPT layout. I think if you take a live cd you can verify if one of the disks at least still has that layout.

LVM is already included in the kernel, so I think the problem has more to do with the raid metadata areas being unrecognized. With the right partition layout you can restore them also with the original mdadm commands, and then you could use a pvscan and pvck. Still it is strange that such "demolition" happened.

 

[Erk]

You should really use the ZFS RAID1 option on the 4.1 installer, as it writes the grub boot data to both drives for you.

 

[Paulo Pastoriza]

At least if you followed the article you have the GPT layout. I think if you take a live cd you can verify if one of the disks at least still has that layout.

LVM is already included in the kernel, so I think the problem has more to do with the raid metadata areas being unrecognized. With the right partition layout you can restore them also with the original mdadm commands, and then you could use a pvscan and pvck. Still it is strange that such "demolition" happened.

Hello! Thanks for your help.

I had already done the computer boot with a live CD of clonezilla. The raid partition appears and LVM volumes as well. Volumes appear correctly available in / dev / mapper /

I already have reviewed several documentation, including:
http://www.tldp.org/HOWTO/html_single/Software-RAID-HOWTO/#ss7.4
https://wiki.gentoo.org/wiki/Complete_Handbook/Software_RAID
https://gist.github.com/plepe/52ecc9f18efb32c68d18
http://wiki.tldp.org/LVM-HOWTO#Introduction]

But it seems to lack any modules in the initramfs that does not allow proper assembly of volumes when the RAID failure. I will use the commands you suggested to try to recognize the LVM in initramfs and see what happens.

Again, thank you.

 

[Paulo Pastoriza]

You should really use the ZFS RAID1 option on the 4.1 installer, as it writes the grub boot data to both drives for you.

Hello, thank you for your help.

I read the documentation about using ZFS with Proxmox, which had been recommended in another post. But I believe the server I'm using Dell PowerEdge R220 32 GB RAM and 2 HD 3 TB, do not have the hardware required to implement the ZFS.

https://pve.proxmox.com/wiki/Storage:_ZFS

 

[Erk]

If Proxmox 4.1 runs on the PowerEdge R220 than ZFS is already loaded and should be able to run. Test it at the command line make a couple of files into a small zpool

What you would need is a small SSD drive as the read cache, it would have to go in a PCIe slot. The SSD is just a performance enhancement, it is not essential for running ZFS, I have 2 Proxmox servers which run ZFS RAID 1 on just decent hard drives no SSD and they work fine.

 

[pizza]

@Erk Did you change any zfs settings for performance or memory usage? With sync=disabled, zfs is speedy but to prevent dataloss, sync=standard is highy recommended?

 

[Paulo Pastoriza]

@Erk Did you change any zfs settings for performance or memory usage? With sync=disabled, zfs is speedy but to prevent dataloss, sync=standard is highy recommended?

Hi, thank you for your help.

But the main problem is that the server, it seems, can not mount the LVM after the failure of RAID. As I explained earlier, I did not get to install the ZFS option, believing that the server hardware would not have a good performance with this option.

If I can not find a possible solution for software RAID and LVM, I try to use the documentation to install with ZFS.

Tks!

 

[windinternet]

RAID is not a magical bullet. Whether you use raid with LVM, mdadm, dmraid, zfs, fakeraid, hardware raid, or whatever, there will always be instances that it is not going to help. A powersupply delivering a voltage spike internally, a fan that resonates the enclosures, heat buildup, sooner or later you will experience a malfunction that hits two disks simultaneously.

So, regardless, to recover more easily you are always going to need a copy of your partition layouts and of the different layer setup and versions of metadata that you are using until you arrive at the block device that is actually seen by the userspace.

 

[Erk]

Hi, thank you for your help.

But the main problem is that the server, it seems, can not mount the LVM after the failure of RAID. As I explained earlier, I did not get to install the ZFS option, believing that the server hardware would not have a good performance with this option.

If I can not find a possible solution for software RAID and LVM, I try to use the documentation to install with ZFS.

Tks!

Just boot off something else, mount the LVM, recover the data, and reinstall from scratch using a ZFS mirror instead of the soft RAID.