How to check array is mirroring data? Boot from mirror?

[Callum Bennetts]

I'm trying to find a way to reliably check that my raid 1 array is mirroring data.

I don't want to go down the route of pulling out disks as there must be an easier way. I'm fairly new to servers so was wondering if anyone can tell me if its possible to boot from the mirrored drive to check it? I thought HP would have come up with an easier solution (CD or something) is there anything available?

Thanks for your help.

[K240]

If there is not any error indication for the disks and you correctly set up a RAID1 mirror, data should be consistent. There should be possible to boot up from any drive used in mirror.

[Torsten.]

You did not tell what server, what controller and what OS you are using. However, most raid adapters present a logical drive to the OS, the physical drives are just under HBA control only. So you can use tools like ACU to check the status.

[Callum Bennetts]

Sorry I forgot to add my hardware. I'm running a Proliant ML115 and P400 Smart Array.

I've set the disks up as a raid 1 mirror in ACU and it reports that everything is running fine however I don't just want to take the programs word for it I actually want to be able to see what data is on each drive. As far as I know there is no way to do this in ACU so being able to boot from the mirror is one possibility however I have no idea how to do this as am a bit of a newbie.

Can anyone help?

[Callum Bennetts]

just to add also there is data on the drive as i installed a base os (ubuntu server).

[gregersenj]

Callum

There's no black magic, that I ever have heard aboute.

If you don't want to trust the software, you have to do the physical check = Pull disk 0. If you choose to do so: Power down the server!! and pull disk 0. Confirm the error messege from the Smart Array.

Other means of test is to run the ADU - Array Diagnostics Utility.

And the Insight Diagnostics. Insight Diag, will do read test's on the physical drives only, it will not do any RAID test.

Sorry you have to be realistic.

cheers
/jag

[Callum Bennetts]

Hi gregersenj,

Thanks for your reply. I don't think I'm being too unrealistic. If software and hardware report physical disks and array are healthy thats fine but I don't think its too much to ask to be able to check if there is mirrored data on the drives without disconnecting them.

I would have thought that HP have a procedure for checking data on a disk from within an HP utility or from the command prompt. Alternatively having some way of booting from the mirrored physical disk would confirm the array is working.

Pulling disks out seems to me extreme and also means hours worth of rebuilding the array (one rebuild for each disk checked). Surely there is an alternative way?

[TTr]

The OS does not see the individual disks of the array. The controller will provide the logical disk regardless of whether both disks ar up or only one is up. If you don't trust the ACU views or the ADU reports, then pulling disks is the only way of testing.

You are not unrealistic at all, perhaps a little extreme which can be justified by how critical the environment and data that you have to support. Some auditors can require you to do these tests as well.

[Callum Bennetts]

You can probably tell that I'm paranoid about loosing data but I can't really blame myself as data integrity is critical.

I'm sure many people have the same so I find it a little strange that HP haven't built into the ACU/ADU something that allows the administrator to see a simple visual representation of the data held on each physical disk. Even if it told how much data in total was recorded on each disk that would help.

[Matti_Kurkela]

Any visual representation in ACU/ADU is based on what the SmartArray controller tells it. Changing the form of the presentation from text to graphics to singing & dancing animations won't change that fact.

You seem to be asking a confirmation to verify if the controller is in error or lying to you: for that, you'll need to pull a disk and see what happens.

Some versions of SmartArray can identify the parts of the RAID array by using the LEDs on the respective disks: e.g. when you select a logical disk from the ACU, the identification LEDs on appropriate disks will be lit. Again, this is done by the SmartArray controller, so it is not an independent verification.

The idea of a hardware RAID is "use hardware to make two or more disks look like one so that the OS does not need to deal with disk failures at all". The standard way to implement a true hardware RAID is to hide the physical disks and present to the OS only the fault-tolerant (or in case of RAID 0, high-performance) logical disks. A consequence of that is that the OS bootloader does not even notice disk failures: it literally cannot tell the difference between a good RAID set and a fault-degraded one.

Implementing a way to by-pass the hardware RAID controller would usually require adding extra non-RAID SCSI/SAS/SATA controllers to the system, and implementing a switching logic of some sort to prevent the non-RAID controllers for accessing the disks when the RAID controller is in use and vice versa.

In mass-produced commodity servers, these extra parts would add manufacturing costs to every unit for a feature that would not be important to the majority of customers - so the extra parts will be optimized out of the design.

MK