Error creating Raid 5

[Jocko]

ps instead of raid5 you could try if you succeed to build a raid1 array of these 4 disks

Sorry Fvdw, I'd rather that b1gnas get first a clean raid5 and add one by one the Hitachi disk in the raid array after running the fsck command.

(with raid1 you can get a synchronize raid even if a disk has a failure status)

About fsck the flag "-f" should be also used to force scan also on the declared clean blocks

[fvdw]

Hi Jocko,
Understand, but making raid 1 is only to test if he can format a device bigger then 2 TB.
It seems he can format the sdc8 and sdd8 partitions on the disks itself, but it fails when putting them in the array.

Further I find it strange that md mentions it creates an array of 730105968 4k blocks = 2785 GB, you would expect sometihing like 3.6TB. probably this is because it is a raid setup ? I am not a raid expert.. :scratch

from dmesg output it seems formatting the raid fails around block 314572942 which is at 1200 GB so on disk sdc

@b1gnas
what is the output of

Code: Select all

 mke2fs -V

[Jocko]

In the md detail output you have

Raid Level : raid5
Array Size : 2920423872 (2785.13 GiB 2990.51 GB)
Device Size : 973474624 (928.38 GiB 996.84 GB)

Each disk has a size 928.38 GiB. So with four disks, the raid size is (4-1)*928.38=2785,14

It seems he can format the sdc8 and sdd8 partitions on the disks itself, but it fails when putting them in the array.

It is normal because no metadata is set for the md device and so you can not format the md device and you should not do it.

from dmesg output it seems formatting the raid fails around block 314572942 which is at 1200 GB so on disk sdc

Yes I think that sdc and sdd have some bad blocks. ( and I don't like the 7K1000 disks: fast but not robust)

[b1gnas]

@b1gnas
what is the output of

Code: Select all

 mke2fs -V

Code: Select all

root@fvdwsl-base:/ # mke2fs -V
mke2fs 1.42.11 (09-Jul-2014)
   Using EXT2FS Library version 1.42.11


So...i'm a little frustrated... what should i do exactly?

[Jocko]

So you have the right version for e2fsprog tools.

You have the choice between two tests :
- fvdw suggestion:
1/format sd[cd]8 partition

Code: Select all

mke2fs -t ext3 -j -m 1 sd[cd]8

2/ check these partitions

Code: Select all

fsck -y -f sd[cd]8

3/ after scanning both disk, try to make a raid1

Note: anyhow, later you should destroy this raid

- my suggestion (if you have another free disk even if it is small)
1/ replace sdc or sdd by this new disk
2/ create a new partition table on it with only one partition,
3/ build a raid5 with the 2 Seagate disks (because I assume that they are clean) (sdb8 - sde8) and with the newdisk
4/ try to format this raid

if it is ok, then we can try to repair the Hitachi disk and include them in the raid5

[Jocko]

Hi b1gnas,

Did you make something else ?

[b1gnas]

Hi Jocko,

yes, after a short break i'm here again :P

I solved the problem at the root, replacing the two hitachi disks with two other seagate, and finally i think the raid is working

in this moment it's resyncing

Code: Select all

root@fvdwsl-base:/ # cat /proc/mdstat
Personalities : [raid0] [raid1] [raid6] [raid5] [raid4] [raid10]
md0 : active raid5 sdb8[0] sde8[3] sdd8[2] sdc8[1]
      2920423872 blocks level 5, 64k chunk, algorithm 2 [4/4] [UUUU]
      [>....................]  resync =  1.3% (12859648/973474624) finish=1679.1min speed=9534K/sec
     
unused devices: <none>


then when it will finish, how could i add the fifth disk?

[Jocko]

Hi b1gnas,

So that confims the issue was the 2 Hitachi disks.

To add the 5th disk, after making the raid5 with 4 devices, in a root terminal do :
*Include the device in the raid array

Code: Select all

mdadm --manage /dev/md0 --add /dev/sda8

(sda8 is then a spare device)
* Increase the raid size

Code: Select all

mdadm --grow /dev/md0 --raid-devices=5

*
* extend the file system

Code: Select all

resize2fs /dev/md0

* save the new raid configuration

Code: Select all

cat /etc/mdadm.0 > /etc/mdadm.conf
mdadm --detail --scan >> /etc/mdadm.conf

Note if you included previously sda8 in a raid you need to clear its superblok before adding it.

Code: Select all

mdadm --zero-superblock /dev/sda8

Then after rebooting you should have a raid5 with 5 members

[b1gnas]

Ok, i have to wait that the resyng in progress is done?

[Jocko]

I don't have a clear opinion about it

But maybe is more safe to leave mdadm running only one job by step