On-board RAID controller misbehaviour?

I am not sure I am posting this in the right category, so excuse me if I am not, but here goes:

I have been using a cheap MB with onboard NVidia raid controller connected to mirrored SATA-disks. this went OK for several years although from time to time the mirror broke up resulting in a re-syncing operation that I had to start manually.

This went OK for many times until one time, the primary disk had developed bad sectors.

As usual the mirror broke and I started a re-syncing operation, not knowing that the RAID system was starting to copy the disk with bad sectors to the one without. This resulted - as can be expected - in disaster, both disks had become corrupted.

I probably operated the RAID system in the wrong way and simply had been lucky initially, although I never realised that I was running a strong risk.

Is this recognisable? I am now very much distrusting these hardware based on-board RAID chips since when resynching, they apparantly don't compare both disks and take best readable sectors for copying to the less readable but instead just decide that one disk is master and the other is slave without checking if the presumed master is error-free.

Am I just expecting too much of these solutions? Or was it because it was a cheap version and not 'professional' ?

With software-based mirroring (or volume shadowing) I have a different experience. I trust it more since I have the impression it takes best readable sectors and copies those to the other disk.

I hope for some comments from people who have experienced the same or can recommend a better approach.
2 answers Last reply
More about board raid controller misbehaviour
  1. The nVidia's RAID solution is purely software. It's not hardware. In terms of reliability I would rate it worse than cheap add-in RAID cards (which are also software based).
    Best software RAID solutions would be between Linux, Server 2008 or Intel Matrix RAID. Although not as reliable as hardware solutions, they're reliable enough for home use.

    Look for data recovering software suggested in this forum. I can't remember them all, but there's quite a few around. You should be able to use those to recover data.

    Real hardware RAID is expensive. They're expensive because of onboard processor and cache to do ECC (and parity calculation if using RAID 5 or 6) on the fly. I'm using a Dell PERC 5/i 512MB w/ BBU here which was relatively cheap for a hardware RAID card.
  2. ECC is performed internally, not in the RAID-layer. RAID5 also isn't slow because of parity calculations, those take hardly any CPU time. What takes a lot of CPU time are all the memory copies as a result from combining and splitting virtual I/O requests. It will cost you CPU, but results in much faster operation, which is what you want.

    For the TS, why not create two non-RAID disks and use some utility to syncronise them? To recover your current data, first fix the bad sector using HDD manufacturer utility or a third party utility like Spinrite. Once both disks are free of errors, you can sync the RAID1 again, but personally i would trust a manual recovery more.

    Any linux livecd is able to recover your data.
Ask a new question

Read More

NAS / RAID Controller Storage