Smartarray controller P812 firmware 3.52

[chuckk281]

Daniel was looking for some customer info:

*********************

Hi Experts,

I've a customer who has updated his firmware to 3.52 after reading the

advisory below.

http://h20000.www2.hp.com/bizsupport/TechSupport/Document.jsp?objectID=c0254

7655

The customer is worried and wants to know he can know if he was already hit

by the issues described:

CUSTOMER ACTION REQUIRED to Avoid Potential for Incomplete or Problematic

Data Transfer Scenarios During Periods of Heavy I/O Involving Discontiguous

Write Operations. The customer wants to know how this issues would manifest themselves and

what the remediation steps would be (apart from the firmware upgrade of course). Consistency

checks of the hard disks?

************************

Mark had some info for Daniel's customer:

****************************

Consistency Checks of the Hard Drives are always being run in the background for Smartarrays. Depending on the size of the drive and the IO placed upon it, it may take a while to loop through one consistency check, but they're always on.

Unfortunately I don't know how this Data transfer scenario pans out. "Incomplete" or "Problematic" can address such wide ranging scenarios.

So regarding incomplete, if it means it's incomplete for the whole stripe, then a consistency check isn't going to call it out, but if its incomplete for a disk in the stripe then it most likely will as the XOR data will be wrong. I'm not sure you'll get a good answer on this. These advisories seem to be worded vaguely so they appear more toned down, and it's hard to get the real story sometimes.

Also, the other thing you've got to think about is generally, if it's not picked up by a consistency check, e.g the stripe is OK, then the next port of call is to check the backup logs on a full backup. Of course the file may have a block as an incomplete transfer, so it's not a 100% fool proof, but it gives you a warm and fuzzy on the file system itself, and the master file table.

****************************

Other comments or questions?

[Kretzer]

Hello,

i think i am the customer which is talked about here, and i really need something more than a "warm fuzzy feeling"!

The system is a DataCore SAN which is build from 336 disks in MSA 60 & MSA 2600 connected to 6 P812 controllers. All our company data is on that system. We had a similar problem in an old P800 firmware version with raid 5 (we had inconsistencies in our databases which was most critical) so that is the reason why we switched to raid 1 for the more important data (and so we have wasted much money on the additional disks).

I really cannot understand how something like this is taken so lightly by HP that we do not have an answer until now!

Markus Kretzer

Buhl Data Service GmbH