Interpreting MDADM Reports

[Jared]

Is there anyone here who's a pro at understanding the mdadm --examine reports?

Here's what I'm getting from the members of the array (each page is a different drive)
View attachment MDADM Readouts.odt

Some things seem obvious:
RAID 5
64Kb Stripe Size
It looks to me like /dev/sdh went offline first.

But, what's confusing to me is where it shows:

Raid Devices[0]: (--@0K 0@0K 1@0K 2@0K 3@0K 4@0K)

Does the "--" part indicate a hotspare?
I also notice in the report it mentions: "Redundant hdr: yes"

I'm sort of a noob when it comes to using mdadm, so I haven't looked through a lot of these. Just seems that something is amiss here.

 

[lcoughey]

Don't forget, this is mdadm's attempt to interpret Dell's perc raid meta data.

 

[Jared]

Yes, and the XOR results point me to a single RAID 5 set of all drives. I'm starting to think that the user created a new array and or volume over the existing one. Which is why the partition doesn't make sense to what I'm finding.... gonna be a lot of scanning and waiting to get this one done I suspect.

 

[lcoughey]

It is two RAID 5 sets across all the drives, for sure. But, just like you can use LVM to make separate software RAIDs based on partitions, this controller does the same, but with a hardware RAID controller to manage it.

I'm sure that if you were to jump to the offset 377487360K on all the drives, you will find another MBR and the start of a second RAID 5 set.

 

[Jared]

I figured it all out. Seems you can't trust the results of mdadm too much on Dell Perc 6/i controller. At least it had the right stripe size and got me the right starting point of the second virtual disk. Drive order on the other hand was way off, so I guess you can't use it.

 

[Jared]

OK, so just to follow up post the final drive order turned out to be:

• 1
• 4
• 2
• 5
• 3
• 6

Compared with what mdadm seemed to report of:

• 1 or 4
• 2
• 5
• 3
• 4 or 6
• 4 or 6

At least it had drives 2 5 3 in the right rotation in relation to each other. Just shifted in the overall location. I think having taken two drives offline had changed the metadata. Probably would have been right had it been read when the array was healthy.