EVA 4400 -- complete drive enclosure failure -- what if

[Thomas Ermann]

Hello,
let's say we have an setup with eight enclosures with a classic loop design.
A consultant on our project claims that we still would be able to continue to access all data if, for example, enclosure #2 went completely dead.
Wouldn't the loops be broken?
Would we be able to access the data on enclosure #3 and following?

We understand that there is no single point of failure in the design and that a whole enclosure going dead would be most unlikely; but the claim seems so bold that we need to have this checked. Is there any reference for this?
I use a reference the "HP StorageWorks EVA4400
M6412 drive enclosureinstallation instructions" document.
Regards,

[marsh_1]

hi,

have a look at the diagrams on page 80 here and trace the paths :-

http://bizsupport1.austin.hp.com/bc/docs/support/SupportManual/c01631857/c01631857.pdf

hope this helps

[Thomas Ermann]

Thanks Mark,

actually this complicates the issue :-)

I only have experience with EMC's CX300 where a dead DEA would mean an interrupted loop (but was told it would be highly unlikely thanks to redundant parts).

Does the EVA handle it differently?

[Rob Buxton]

There's actually a video around of HP shooting an EVA through a drive enclosure and it keeps going.

Not a recommended test, but if you have the RSS Groups set up okay, it should tolerate a complete shelf failure.

[McCready]

The below comment,
"but if you have the RSS Groups set up okay,"
has me really puzzled - how does one "set up" an RSS group to distribute one member per each enclosure?

My understanding was that the EVA's firmware attempts to do the right thing (no more than one member of an RSS group on the same shelf/enclosure) but that over time, there is no guarantee that even in an 8+ enclosure configuration, the RSS groups may not be able to maintain the one member per enclosure optimal distribution (due to drive failures and replacements). And I have never heard of a way to explictly move RSS members around to different enclosures.

And why do you think the latest versions of CommandView no longer show your RSS "status"? HP got tired of getting calls about that "RSS Disk state: None " display when you can't do anyting about it - or can you?

So, Mr. Ermann, your consultant is probably incorrect - there can be single points of failure, especially when using RAID5 directly after a single drive failure until the reconstruct of the RSS group is done.

[Mark...]

Hi

To get back to the original question...
If the EVA is cabled correctly [that is ANY EVA] then on one side of the loop the cables go "top down" and the other side is "bottom up". Therefore, if a shelf should fail both controllers will still be able to access all the remaining shelves - this is regardless of how many shelves are in any of the EVA's.
Obviously a switched system is cabled differently.
Either way, assuming all cabling is as per HP's recommendations then the EVA will continue to function in the event of a SINGLE shelf failure.

OK, as for RSS's - I think enough has already been mentioned in several discussions within the itrc without me going over it again now. Please have a search and read what has been said already!

Mark...