Is there a way to not lose RAID0 array after power loss ?

EVIDEON · 07-05-2022, 10:43 AM

Hello everyone,

I'm owner of an hp ZBook 17 G6 with 3 NVMe setted up in RAID0 with mdadm.
The operating system installed is CentOS 9 Stream.

I had my pc in sleep mode for too much time and it shutted down.
After reboot RAID0 array was lost.

If I shut down or reboot manually no problem occurrs but, instead, on power loss the RAID0 virtual disk disappears.

What I wish to know:

-Is there a way to avoid this or prevent it ?
-Is there a way to let Linux rebuild it easyly and quickly after reboot (from power loss) ?

Thanks for help

smallpond · 07-05-2022, 11:27 AM

Are you trying to boot from RAID? How is grub configured and installed? What's in your mdadm.conf file?

EVIDEON · 07-05-2022, 06:07 PM

Thanks a lot for quick reply.

My system has :

- 1 SSD with CentOS installed and from where the machine boots up (/dev/sda)
- 3 NVMe configured in RAID0 (/dev/md0 mounted on /mnt/md0)

I'd like to say more about grub configuration and installation but I've no experience. (don't know what information to give)

My mdadm.conf file (which isn't located in /etc but in /usr/share/doc/mdadm) :

Code:

ARRAY /dev/md0 metadata=1.2 name=ZERO:0 UUID=d307e76e:384447ad:17f0cd99:7cb32895
ARRAY /dev/md0 metadata=1.2 name=ZERO:0 UUID=430c76f6:daf84a85:ce98bd65:80a6a25b

I don't know if it can be useful but my /etc/fstab file contains :

Code:

# CentOS
/dev/mapper/cs_zero-root                     /                  xfs        defaults                          0 0
UUID=df0563e5-7e57-428e-8962-b51a7d84cff8    /boot              xfs        defaults                          0 0
UUID=AA57-9A10                               /boot/efi          vfat       umask=0077,shortname=winnt        0 2
/dev/mapper/cs_zero-home                     /home              xfs        defaults                          0 0
/dev/mapper/cs_zero-swap                     none               swap       defaults                          0 0
# RAID0
/dev/md0                                     /mnt/md0           ext4       defaults,nofail,discard           0 0

Really appreciate your help

Thanks

syg00 · 07-05-2022, 07:15 PM

Let's see

Code:

cat /proc/mdstat
sudo lsblk -f

lvm_ · 07-05-2022, 11:57 PM

When array is being written to there are moments when some of the drives in the array have been updated while some haven't. If computer is turned off at such a moment, array becomes inconsistent and as a precaution won't be mounted automatically. You can mount inconsistent array manually with --force flag but because RAID0 lacks any means to resolve this inconsistency - checksums, redundancy or journaling, these errors will keep on accumulating. RAID0, especially a software one, is not for systems which can be powered down unexpectedly, even if you forget about its general unreliability. Just isn't.

smallpond · 07-06-2022, 05:03 AM

How does md find your mdadm.conf file if it's not in /etc?

business_kid · 07-06-2022, 05:58 AM

I would go through it with ls -lht, and find the last stripes being written. Avoid that file if you can trace it.

I had dealing with 1 Raid 0 system, which striped data onto 2 drives and the data map was on a third. It seems obvious that whoever set the system up wasn't worried about backups. Until now

.

I personally doubt that a file in /usr/share/doc is the configuration, and personally, I would expect 3 drives in a Raid 0 config; Two at least for data, and a third for housekeeping. Are those your disk UUIDs?

Also, just because it says 'Raid0' in /etc/fstab doesn't necessarily mean the thing uses Raid 0. It could be Raid (device) 0, as opposed to Raid device 1, etc. I'd look for independent verification.

EVIDEON · 07-06-2022, 09:33 AM

Thanks everyone,

mdadm.conf not in /etc sounds strange for me too.
On debian based os the file was there, but here, the system made its folder choices.
I've tried to copy it in /etc but nothing changed.
Anyway without power loss, with manual reboot or manual shut down the array was recognized with no problems.

Code:

[zero@ZERO ~]$ cat /proc/mdstat
Personalities : 
unused devices: <none>
[zero@ZERO ~]$ sudo lsblk -f
[sudo] password for zero: 
NAME FSTYPE FSVER LABEL UUID                                   FSAVAIL FSUSE% MOUNTPOINTS
sda                                                                           
├─sda1
│    vfat   FAT32       AA57-9A10                               591.3M     1% /boot/efi
├─sda2
│    xfs                df0563e5-7e57-428e-8962-b51a7d84cff8    612.7M    40% /boot
└─sda3
     LVM2_m LVM2        rG0PHk-E5OH-o3OT-eVrJ-6CVW-Gd4l-3lAJZJ                
  ├─cs_zero-root
  │  xfs          /     41b7a843-75db-489f-a84a-421df2ac9f9a     45.3G    35% /
  ├─cs_zero-swap
  │  swap   1           b55acf2e-2642-4a54-9b88-ff03acf3fb86                  [SWAP]
  └─cs_zero-home
     xfs          /home 912815cf-60de-4667-a89e-782a65bb5783      1.7T     1% /home
nvme0n1
                                                                              
nvme1n1
                                                                              
nvme2n1

Quote:

I personally doubt that a file in /usr/share/doc is the configuration, and personally, I would expect 3 drives in a Raid 0 config; Two at least for data, and a third for housekeeping. Are those your disk UUIDs?

I've searched but the .conf file is only there (/usr/share/doc).
I setted up level=0 for raid during configuration but can not say much about system choices for data. (sorry)

Thanks

business_kid · 07-06-2022, 03:09 PM

A file in /usr/share/doc is a sample config. What's more, the UUIDs in the config don't seem to be the UUIDs in the config file. You can check with

Code:

ls -l /dev/disk/by-uuid

I don't think you can have LVM and RAID operating at the same time. LVM hoovers up all the space and administers it. So directories can grow or shrink, and LVM manages that space. Raid 0 wants very definitely assigned set of spaces.

michaelk · 07-06-2022, 06:13 PM

Post the output of the command:
sudo mdadm --detail --scan

Yes, you can have both RAID and LVM. The interesting part is that none of the NVMs show any information about being a RAID member.

EVIDEON · 07-07-2022, 05:42 PM

Code:

[zero@ZERO ~]$ ls -l /dev/disk/by-uuid
total 0
lrwxrwxrwx 1 root root 10 Jul  6 00:31 41b7a843-75db-489f-a84a-421df2ac9f9a -> ../../dm-0
lrwxrwxrwx 1 root root 15 Jul  6 00:31 5a7cc108-4795-4324-89b7-3d035b4cd8e5 -> ../../mmcblk0p1
lrwxrwxrwx 1 root root 10 Jul  6 00:31 912815cf-60de-4667-a89e-782a65bb5783 -> ../../dm-2
lrwxrwxrwx 1 root root 10 Jul  6 00:31 AA57-9A10 -> ../../sda1
lrwxrwxrwx 1 root root 10 Jul  6 00:31 b55acf2e-2642-4a54-9b88-ff03acf3fb86 -> ../../dm-1
lrwxrwxrwx 1 root root 10 Jul  6 00:31 df0563e5-7e57-428e-8962-b51a7d84cff8 -> ../../sda2

(mmcblk0p1 is my SD removable card)

Code:

[zero@ZERO ~]$ sudo mdadm --detail --scan
[zero@ZERO ~]$

(absolutely nothing)

Thanks

michaelk · 07-07-2022, 06:21 PM

dm-0, dm-1 and dm-2 is the LVM but it appears your RAID has just disappeared. I should of asked you to also post the output of the commands:

mdadm -E /dev/nvme0n1
mdadm -E /dev/nvme1n1
mdadm -E /dev/nvme2n1

You can try to reassemble the array as:
sudo mdadm --assemble /dev/md0 /dev/nvme0n1 /dev/nvme1n1 /dev/nvme2n1

EVIDEON · 07-07-2022, 06:29 PM

Code:

[root@ZERO zero]# mdadm -E /dev/nvme0n1
/dev/nvme0n1:
   MBR Magic : aa55
Partition[0] :   4000797359 sectors at            1 (type ee)
[root@ZERO zero]# mdadm -E /dev/nvme1n1
/dev/nvme1n1:
   MBR Magic : aa55
Partition[0] :   4000797359 sectors at            1 (type ee)
[root@ZERO zero]# mdadm -E /dev/nvme2n1
/dev/nvme2n1:
   MBR Magic : aa55
Partition[0] :   4000797359 sectors at            1 (type ee)
[root@ZERO zero]#

Code:

[root@ZERO zero]# mdadm --assemble /dev/md0 /dev/nvme0n1 /dev/nvme1n1 /dev/nvme2n1
mdadm: Cannot assemble mbr metadata on /dev/nvme0n1
mdadm: /dev/nvme0n1 has no superblock - assembly aborted
[root@ZERO zero]#

michaelk · 07-07-2022, 07:01 PM

The drives do not look like they are partitioned but look like there is something there. Post the output of:

sudo fdisk -l /dev/nvme0n1

Do you have a backup of important data from the RAID? Do you need to recover any data?

EVIDEON · 07-08-2022, 09:58 AM

Code:

[zero@ZERO ~]$ sudo fdisk -l /dev/nvme0n1
[sudo] password for zero: 
Disk /dev/nvme0n1: 1.86 TiB, 2048408248320 bytes, 4000797360 sectors
Disk model: KXG50PNV2T04 KIOXIA                     
Units: sectors of 1 * 512 = 512 bytes
Sector size (logical/physical): 512 bytes / 512 bytes
I/O size (minimum/optimal): 512 bytes / 512 bytes
Disklabel type: gpt
Disk identifier: CBFEE7D0-7424-43D9-A4D3-32D30F9D620C
[zero@ZERO ~]$

Really appreciate the time you gave to my problem.
I don't need backups of data from the old RAID array.
I wish only to know if there is a method to avoid the same problem now that I reset up a brand new RAID configuration.
I need to not lose data in case of power loss.

Thanks everyone for help.

Thank you Michael.