[Beowulf] Big storage

Bruce Allen ballen at gravity.phys.uwm.edu
Fri Aug 31 03:02:26 PDT 2007


Hi Leif,

>>> I think you should add a parity verification to this - not just a
>>> sector scan.
>>
>> Good idea.  In other words, don't just check that all disk surfaces
>> are readable by the disk device.  Also check that the data is
>> consistent from the RAID perspective.  Did I understand you correctly?
>
> Exactly. A colleague (hi, Peter!) just found that our shiny (and cheap) 
> new p400 controllers don't have this rather basic functionality. So we 
> will probably change our procurement procedures to avoid this in the 
> future.
>
> And even among those controllers that *can* verify the parity, there 
> seem to be many that don't handle a detected inconsistency in any sane 
> fashion. Often they just recompute the parity block and claim that the 
> damage is "repaired". Yeah, right.

Uck.

Would someone on the list volunteer to set up and host a web page with the 
pros and cons of various RAID controllers?  We could then collect 
information like this and put it into one place.

For each controller we could put information like:

  - After encountering an unreadable (UNC) disk sector, does it immediately
    issue a WRITE to force sector reallacation?

  - Does it do a background or periodic scan for unreadable (UNC) disk
    sectors?

  - Does it do a background or periodic scan for inconsistent redundant
    RAID data?  If so, does it compute the correct data and then write
    this consistent data to the disks?

  - Does it give access to SMART data?

  - Does it allow periodic SMART scans for unreadable sectors or other
    problems?

  - Other info about performance, drivers, etc could also be in the table.

Cheers,
 	Bruce




More information about the Beowulf mailing list