RSS Disk state - EVA Command View

[Craig Howe_1]

Hi

Under the properties of the disk group in command view it shows any one of these:

RSS Disk state: None
RSS Disk state: Parity
RSS Disk state: Mirrored

Just wondering what the differences are - how they affect RSS?

thanks

Craig

[Steven Clementi]

Craig:

RSS Disk State of None usually means that all disks are on one shelf and unfortunately... has a higher potential for failure.

RSS Disk State of Parity usually means that at least 2 members of an RSS set are located on the same shelf. Think about an EVA 2c4d with 8 disks in a group. 1 RSS, 2 on each shelf. (or a 2c6d with 8 disks, 2 shelves have 2 drives and the others have 1 drive)

RSS Disk State of Mirrored usually means that all disks in an RSS are on separate shelves.


Steven

[Craig Howe_1]

Hi Steven

Thanks for the info. So obviously I could be in a bit of trouble with a group containing 56 disks and RSS Disk state: None, even thought I have a 8-enclosure EVA 5000.

How is it possible to distribute this across the enclosures (online)?

Perhaps this potentially happened after disk grouping/ungrouping? We have ungrouped disks due to either disk replacement or firmware updates.

thanks

[Sameer_Nirmal]

Hi Craig,

As mentioned in my reply to your previous thread
http://forums1.itrc.hp.com/service/forums/questionanswer.do?threadId=1002866
if the SPOF configuration rules are followed as may be in your setup (2C8D), then very first/initial RSS state of a DG would be mirrored. But later on, if join or split operation for RSS occurs (on account of disk failure or adding new disks), then "levelling" would occur and VCS firmware can not ensure the two disks from the same RSS will not reside in the same disk shelf. This would obviously make the RSS state other than "mirrored".

I would say if one has maintained compliance with SPOF configuration then the firmware should take care of keeping RSS state to its most redundant state "mirrored" through the levelling process. I know VCS 3.x firmware has this drawback , but I am not sure if it is taken care of in XCS. I understand that having this done through levelling would make the process complicated and more time bound, but I hope XCS which signifies Xccelerated Controller Software is/maybe smart enough to do the trick?

I didn't see any advisory or workaround by HP to compensate this VCS shortcoming. Maybe you can contact HP and ask for workaround on this.
But I know Uwe Zessin ( Well known Storage Guru here ), has tools/utilities to do the trick if you provide him current RSS setup through SSSU.

To make DG RSS state back to "mirrored", one has to know RSS ID and RSS index and then physical interchange of certain disks would do the trick.

[Uwe Zessin · Accepted solution]

Well, let me add my understanding of what "RSS Disk state" means:

None - (at least) two disk drives which belong to a disk pair are located in the same disk drive enclosure.

Mirrored - all paired disks in a RSS are located in different disk drive enclosures / there is no disk pair that is located in the same disk drive enclosure.

Parity - ALL disks that belong to one RSS are located in different enclosures.


'disk pair' - the EVA "marries" disks in pairs to designate them for storage of VRAID-1 data. That's the reason why a disk group with an odd number of disks (e.g. 11) which is filled with VRAID-1 virtual disks only puts data on 10 disks. And it should explain why the protection capacity is always calculated using *2* or *4* times the size of the largest physical disk drive in the disk group.


As far as I can tell, there is currently no 'repair' mechanism in either VCS/XCS. Of course you can try to do a bit of group/ungroup operations and just 'hope' that it gets fixed, but I'd say it is a waste of time.

The only repliable way is to identify the problem spots and fix them by manually moving the disks.


If you can provide me with the output of
SSSU> ls disk full XML
SSSU> ls group full XML

and attach it as a .TXT file, I'll see what I can do.

[Craig Howe_1]

Thanks Uwe for your help.

I've attached a zip file of the txt outputs.
Currently there is a group with only 8 disks, this is moving to 16 tomorrow.

cheers

Craig

[Uwe Zessin]

Thanks - interesting data - it caused an exception in my code. Looks like I have to investigate a bit, but I could work around the problem. Attached is the report - enjoy!

[Michiel Van Eyken_1]

Uwe,

Which tool are you using to get an output via the SSSU files?

Is it something you wrote yourself?

Is it somewhere public available or only through HP?

Thanks in advance,

Michiel

[Craig Howe_1]

Hi Uwe

I moved all the online drives to ensure they were on seperate enclosures, however I still get the 'parity' state instead of 'mirrored' even after levelling has finished.

Considering I have all vraid1 in this group, this still isn't good.

Any ideas? Do I have 'married' pairs in the same enclosure or something?

thanks

[Uwe Zessin]

Craig,
I am not sure I understand you correctly. A 'parity' state is "better" than a 'mirrored' state.

Can you provide me with an update of the configuration?

[Craig Howe_1]

Hi Uwe

Sorry I got it wrong, I was thinking the mirrored state was better than parity but its the other way around :)

Thanks for all your help, after a decent amount of ungrouping/grouping, I got the main group back to a parity state and now working on the 2nd group.

[eis1953]

Thanks the information here answered just what needed. F.Y.I. HP has a feature that you can install on your appliance that will create a CSV file showing which diks are in which RSS set, it also shows events, Vdisks etc.. It's called SDC support data collector.

Ed