RAID5 Question - Drives got mixed up - Any solution here?

[tbird2340]

A customer of ours had a server die (Dell PowerEdge 2850).. He had another 2850 so he decided to take the drives out of the dead and put it into the working one..

That didn't work.. When that didn't work he started moving drives all around.. So that's when I got brought in..

I have a dead server and a working server.. However, I have 9 HDD's and he doesn't know where / which slots / server they went in to..

Is there ANY way to figure this out.. I told him he's screwed but wanted 100% verification..

The array config shows 1 drive online out of 4 (RAID5) and the other 3 as failed..

Thanks

 

[VirtualLarry]

I would find out if the RAID card writes metadata to the drive, find out where on the drive the metadata is located, and use a disk hex editor (read-only!) to manually decode the metadata, determine which drive goes where, and then plug them all in correctly and see if it recognizes them.

 

[Phaetos]

I would find out if the RAID card writes metadata to the drive, find out where on the drive the metadata is located, and use a disk hex editor (read-only!) to manually decode the metadata, determine which drive goes where, and then plug them all in correctly and see if it recognizes them.

Seriously? Wow ....

 

[Gunbuster]

Is it a discrete RAID card (PERC)? if so I would try moving it over to the working server.

 

[tbird2340]

?? Basically I have 9 drives.. I have no idea what drives go to what server and what slots go to what..

I have the working server but without knowing the above I'm stuck..

 

[tbird2340]

And yes, it's a PERC.

 

[Quad5Ny]

Does the RAID controller store any metadata about previously connected drives (like the serial #)?

--

Assuming all the drives were new when they were put in the server, use smartmontools to find the 'Power On Hours' and 'Power Cycle Count' (maybe POR_Recovery_Count also) of all the drives.

Then take drives that have seem to have matching values for both of those (or close to matching) and see it they work.

You might even try matching drive batch/serial numbers.


P.S. I know on Intel fakeRAID, it stores the array info in the second to last sector of the drive. Its easily viewable using a hex editor (I like HxD personally).

 

[Gunbuster]

Look for the revision or firmware numbers on the drives. Barring hdd replacements they might be different if servers were purchased at different times.

 

[XavierMace]

If memory serves, when I've moved drives between PERC controllers, I get a message that the drive appears to be from a different array rather than just missing drives. Toss a drive in, see if you get a message that drives are missing or of you get the moved message. By process of elimination you can determine which drives were originally in that server. MAY be easier than trying to pull the metadata.

 

[mfenn]

A customer of ours had a server die (Dell PowerEdge 2850).. He had another 2850 so he decided to take the drives out of the dead and put it into the working one..

That didn't work.. When that didn't work he started moving drives all around.. So that's when I got brought in..

I have a dead server and a working server.. However, I have 9 HDD's and he doesn't know where / which slots / server they went in to..

Is there ANY way to figure this out.. I told him he's screwed but wanted 100% verification..

The array config shows 1 drive online out of 4 (RAID5) and the other 3 as failed..

Thanks

Depending on what he did, this could either be fairly easy to fix or he could be 100% screwed. If he force assembled the array (i.e. overwrote the metadata), then the data is probably gone.

If he didn't do that, you can take the card from the old server, put in the new server, and then swap drives in until you find the ones that are recognized by the card. Right now you likely have 1 correct disk and 3 incorrect ones.

It doesn't matter what physical slots the drives are in, the card identifies them by reading the metadata off the disk.