Help: recover raid-6 configuration on LSI controller

[bitgyu]

Can anybody help with a safe procedure to restore the RAID-6 config on an LSI 3008 controller? I’m in over my head, and even after a full day of googling, I’m unsure of the safest way to go.

Configuration:
9 Drives on SAS3 port expander.
Ports 0-7 are RAID-6.
Port-8 was a hot spare.

Steps so far:

1. The drive on port-0 failed. Rebuild started with hot spare (port-8).
2. For “reasons”, decided to recable ports 1-8 directly to the controller and avoid the SAS expander.
   1. Drives were connected in the same order and showed as ports 0-7 instead of 1-8.
3. On startup, multiple errors were observed, so I immediately shutdown
4. Restored cables to the original configuration with the SAS expander. The initially failed drive (port 0) is NOT connected, so only ports 1-8 have drives.
5. On startup, the virtual drive is offline as port-1 now shows as “failed” and port-4 and port-8 are “foreign”.
   1. port-8 was the hot spare that just starting rebuilding to
   2. I don’t know why port-1 is showing as “Failed”.

Nothing has been written to the filesystems, so ideally, bringing port-1 and port-4 online (and not port-8) should have all the data intact, albeit with single parity. That should allow a full rebuild to run.

Worst case is just bringing port-4 back online, but runs the risk of errors during the rebuild.

Any suggestions? Thank you in advance!

 

[alex]

It sounds like a pickle. We run a bunch of RAID5 volumes in remote branches on 3108 controllers, and whenever a drive fails, or when we swap the compute nodes, all it takes is to import foreign configuration - so I've never run into a situation like yours.

What are the options with "Foreign / Unconfigured Bad" drives? Can you mark them "Unconfigured Good"? Once you do, does anything change?

(Hope you figure this out.)

 

[fjstore]

3008 only raid 0 1 5

 

[Micro]

YMMV !
I ran into a similar situation with an R6 on a 9361-8i (caused by a bad power cable).
After changing the power cable, using MegaRaid Storage Manager (LSI Storage Authority also works) I changed the drives to "unconfigured good", selected "replace missing drive", then "make drive online".
The array came back online completely usable, while noting the "failed drive" and when I replaced the "failed" drive, it immediately rebuilt itself.
I also found out (after testing) the "failed" drive wasn't, the bad power cable was the only problem, so now it is the "hot spare".
Again - YMMV !

 

[imprecise_matt]

Oh geez, that's rough. It's been a long time since I had to recover a megaraid configuration, but a few general tips:
- If you've got the space elsewhere, image the drives before you attempt a recovery
- Linux md-raid understands megaraid disk format, and can import megaraid arrays. There are a lot more levers (that I know of, at least) to pull in trying to get the array back with mdadm and related tools, including and especially bringing the array online read-only