ProLiant Servers (ML,DL,SL)
cancel
Showing results for 
Search instead for 
Did you mean: 

About the the RAIDs portability and its rebuilding process

Daniel Bareiro
Occasional Advisor

About the the RAIDs portability and its rebuilding process

Hi all!

I have two new equipment Proliant DL380 G5. In one of them I was testing
leaving an operating system installed on a RAID 5 with a spare disk and soon I
created a RAID 1+0 without data with remaining disks.

The other equipment was offline by days, and to test the arrays portability I
remove the disks in RAID 5 on which is the installation of the OS and when
putting them in the other Proliant it seems that the controller detected
without problems the structure of the Array with the spare disk nevertheless
the OS showed an I/O error. I suppose that this is because was
not present any disk of RAID 1 and for that reason it was broken. Is it
correct?

This takes to me to think that the arrays structure is not stored in some
memory of the controller but in metadata of disks. This is really handled of
that way? Still being thus, I do not have left absolutely clear as it can be
that a broken RAID 1 has been detected without the disks of this Array were
present.

Then I re-enable RAID 1 pressing F2 after bootstrapping assuming the loss of
the data (is there some way to re-enable an array without having to reboot the
OS using hpacucli or would have to erase the array and to create it again
using this tool?) and both arrays was consistent.

The following test that I did after the disks migration was to remove one of
the disks of both arrays to verify that spares responded and I could see using
hpacucli that a auto-rebuild in RAID 5 began, but for RAID 1 I obtained the
message ' ready for rebuild' delaying the rebuilding by some time.

Can it be related to the following message that I then obtained with the
check_cciss Nagios plugin?:

HP Smart Array Problem: Smart Array E200 in Slot 3 Controller Status: OK
Cache Status: Temporarily Disabled Battery Status: Recharging



Thanks in advance.

Regards,
Daniel
2 REPLIES
Cederberg
Honored Contributor

Re: About the the RAIDs portability and its rebuilding process

Hi!

Wow thats a lot of questions..

1) Yes i belive you are correct as the Operating system is expexting to have a Harddrive (Your Raid1+0) thats not there it belives it has crashed.

2) The information is stored on the Harddrives aswell as the Array controller to give you as much flexability as possible. If you take your disks and move them to a identical system then the array controller can read the config from the disks. And when you just need to exchange one drive because its broken the array controller use its config to insert the drive int to the RAID and rebuild it.

3) I have no idea as you often move every disk in a server if replacing the hardware i have never done it..

4) If i understand you correctly you had one global spare? and you removed two disks? well is i understand correctly then the answere is simple. One disk cannot replace two disks.

And if you hade two online spares i'm guessing that it rebuilds on Raid at a time..
Daniel Bareiro
Occasional Advisor

Re: About the the RAIDs portability and its rebuilding process

Hi, Cederberg!

1) I had supposed that. I believe that if I had moved _all_ disks of a server to the other, I would not have had this problem. The OS continues working without problems, but with certain frequency it appears in console that error of I/O that I mentioned.

2) Ok. Then, if I understood correctly, the controller would be taking the information from the metadata of the disks of RAID 5 to have in its hardware a copy of that configuration. This would be meaning that in the metadata of RAID 5 would be information indicating the existence of a RAID 1+0?

3) In this point I referred to if exists any form to remake a broken RAID without reboot the server and using ORCA.

4) No. I have a spare for each RAID and I had removed a disk of each RAID. By the tests that I was doing, it seems that the SmartArray E200 controller does not support the simultaneous reconstruction of RAIDs, since when finishes a reconstruction, the other begins.

Thanks for your reply, Cederberg.

Regards,
Daniel