Problems with SmartArray 5i RAID migration... Help?

[CKalisiak]

Hello all,

I migrated from a Proliant ML370 G2 to a DL360 G2 to reduce power consumption, and part of that migration was to move the RAID-1 group that was installed in the ML370 G2.

When I started the migration to the DL360, I moved half of the mirror, so I would still have the other half in the ML370 just in case. The RAID group was discovered, and I told the 5i to ignore the one drive missing from the RAID group.

I am now comfortable with the DL360's stability, so I am ready to finish the migration. I dropped the second half of the mirror into the DL360, but it didn't automatically start rebuilding. I pulled up the Array Configuration GUI, but wasn't really able to learn much. So I switched to the CLI, and now I see what the problem is:

=> ctrl all show config
Smart Array 5i in Slot 0 ()
array A (Parallel SCSI, Unused Space: 0 MB)
logicaldrive 2 (279.4 GB, RAID 1+0, Interim Recovery Mode)
physicaldrive 1:0 (port 1:id 0 , Parallel SCSI, ??? GB, Failed)
physicaldrive 2:1 (port 2:id 1 , Parallel SCSI, 300 GB, OK)
unassigned
physicaldrive 2:0 (port 2:id 0 , Parallel SCSI, 300 GB, OK)

Does anyone have any idea how to remove 1:0 from logicaldrive 2, and configure 2:0 as a spare to rebuild onto? I tried a few different things, including adding 2:0 as a spare, and changing the RAID level to zero, to try to break the mirror, but those commands resulted in an "error" response. It might be that I was not typing the commands correctly, though.


Thanks in advance,
Chris

[KarloChacon]

hi

well it looks like controller card messed up SCSI IDs

you can try but no guarantee is going to work
- remove HDD and run erase utility and insert HDD again

- update Firmware for controller

- also you can try putting HDDs back just to check HDDs and if you get the same issue in the previous server

- if that does not work you must backup you half RAID 1 ASAP and recreate it all

bye

[gregersenj]

Looking at:
=> ctrl all show config
Smart Array 5i in Slot 0 ()
array A (Parallel SCSI, Unused Space: 0 MB)
logicaldrive 2 (279.4 GB, RAID 1+0, Interim Recovery Mode)
physicaldrive 1:0 (port 1:id 0 , Parallel SCSI, ??? GB, Failed)
physicaldrive 2:1 (port 2:id 1 , Parallel SCSI, 300 GB, OK)
unassigned
physicaldrive 2:0 (port 2:id 0 , Parallel SCSI, 300 GB, OK)

It appears that the ML370 is using scsi port 1 and the DL360 is using port 2.
I haven't been able to verify that.

You must understand that the metadata say that the mirror originally was located at Port 1 ID 1 and 2.

You have moved HDD 1:1 to 2:1 or 2:0 (I'm not sure wich one is the unassagined)
But it is still waiting for the other Mirror member to be installed on port 1 ID 0.

I see 3 options to choose from.
Take good backup, and prepare an emergency restore plan.

1. Shutdown and power off the server.
Move the disk the ML (Wich must be powered off and all drives removed)
Insert the drive in Port 1 ID 1.
Power up the server and leave the smart array in interim recovery mode.
You don't need to boot the O/S.
Insert a disk in port 1 ID 0.
Let the mirror rebuild.
power down the ML
move both drives to the DL and power it up.
The DL will tell you that the drives appear to has changed places.
I'll guess, that the drive did that when you moved it in the first place.

2. Assign a Hot spare, the mirror and for ever use the hotspare as a part of the mirror. That will leave the mirror redundant, but forever degraded, since it's waiting for the missing disk to be installed.

3. Start all over from scratch = delete array, create new array, and use your emergency recovery plan and backup.

good luck
/jag

[gregersenj]

Here's the link how to break mirror's

http://h20000.www2.hp.com/bc/docs/support/SupportManual/c00378986/c00378986.pdf?jumpid=reg_R1002_USEN

But that won't help you.

[CKalisiak]

Thanks for the replies. I was hoping for an elegant solution, but yeah, it looks like the only option that will help is the brute-force sledgehammer approach.

I'll use Norton Ghost to make a full system image of the drive I'm using now, call it A, and will nuke drives B and C and create a whole new mirror. Then I'll restore Ghost to B and C, and make sure it's ok. I'll wait for the rebuild to finish, shut back down, pull out B and C, reinstall A, power back up, and nuke drive A. Then shut back down again, install drive B, then as the system is booting, reinstall drive A.

Hopefully at that point my RAID problems will be gone, and B+C and A+C will be valid combinations again.

(I swap A and B around to take snapshots of my system, for storage in the firesafe, in addition to backing up to SDLT220.)

Thanks,
Chris

[gregersenj]

Well you don't need to do the sledgehammer.
If it's possible for you to shutdown the ML for 1 -2 hours

Read solution option 1.

but first backup backup backup.

[CKalisiak]

Yes, that's true, option #1 that you offered might take less time, but I don't have 100% confidence in Norton Ghost's ability to full restore an image.

I am pretty close to 99% confident, but I am very risk averse when it comes to my data, which is why I have a mirror with two drives that I swap on a monthly basis, in addition to SDLT backup.

What I don't want to have happen is for the Ghost backup not to work for some reason, and only to find this out because I had to restore after the 5i decided to rebuild onto the wrong drive.

No problems, though, thanks for your time.

Chris

[gregersenj · Accepted solution]

True, there's alway a risk, that why I encurage for a backup, and an emergency restore plan.

If only you put the drive, that you want to maintain in the ML, in position 1, or any other than 0.
Power up the server.
When the SA prompts you that the drive has changed position - Accept (F1)
When it ask you if it has to be in interim recovery mode - accept (F2 or nothing)
As soon as the array is in interim recovery mode, insert a drive in position 0, and it will rebuild.

As soon as rebuild has finished, you can power down and move the drives to the DL

You can create a spare, let it rebuild.
When you move the drive (With the DL powered off) put the spare aside.
A little extra safety.

what ever you do, good luck.

[CKalisiak]

Thank you for following up. Ok, I'll give that a try.

Thanks,
Chris

[CKalisiak]

The rebuild took like 5 or 6 hours, but, yes, rebuilding the mirror in the ML370 then moving the drives to the DL360 worked!

(Worked even better than my idea of using Norton Ghost, because Ghost made a mess of the partitions when I tried to restore the backup...)

Thanks,
Chris