EVA:: RSS calculation

[masthan_1]

Hi,
I am having confusions over concepts of RSS. Pls help me with these questions.

- where RSS will play its role and is that mean only 8 to 11 disks will be marked for an RSS creation?
- How can i assure and where can i check that i have not lost any two disks of
same RSS ?
- In an EVA array, Can anybody help me how to ensure that for vraid1, the mirrored partner (Copy) are in diff shelfes?

- what are the important ares of EVA 5k and how to configure vraid5 in the same.

- pls help me, what is the procedure to get the SSSU output for an array?

Thanks
ks

[Víctor Cespón]

The RSS distribution is something that is not intended to be seen by customer.

Internally the disk group is subdivided in smaller groups, 6 to 11 disks, usually 8. Each subgroup contains all the data needed to regenerate the RAID info, so you can lose one disk from each RSS and no RAID 5 or RAID 1 info will be lost.

If you had two disks from the same RSS failed, you'll see a big message on Command View saying "unresolved hardware failure" and many vdisks with "vdisk data loss"

When creating the disk group, the EVA usually choses mirror pairs from different shelves, but this can change as disks fail and are replaced.

You chose the RAID level of a vdisk when creating it.

To collect a SSSU report, use attached txt file.

[Matt Palmer_2]

Hi,

I wrote a custom tool to use SSSU and PHP to draw a color table of the RSS groups on my EVA8000. I would be happy to share it with you if it helps?? It enabled me to be sure that I didnt have uneven RSS layouts.

regards

Matt

[masthan_1]

hi vcespon,

thanks for your answer. but if two disks failed in an RSS, can we able to retrieve the data back despite the failure of lun failure.

and at which scenario, the disk will be detected as going bad, and the levelling happen. again, the reconstruction will happen by the return of that failed disk. pls explain me, reverting, levelling and reconstruction..appreciate your efforts!

Hi Matt palmer,
thanks for your time as well. pls share that tool.

ks

[Matt Palmer_2]

Hi,

please see this post:

http://forums13.itrc.hp.com/service/forums/questionanswer.do?admit=109447627+1280227686621+28353475&threadId=1321854

I have included the programs there for you to grab as attachements, with some basic instructions. Hope it helps. let me know if you find it useful

best regards

Matt

[Oscar_Maimó]

Hi kskat:
<< where RSS will play its role and is that mean only 8 to 11 disks will be marked for an RSS creation? >>

In a Disk Group all redundancy data is inside a RSS. A RSS set is formed by 6-11 disk according the number of disk drives on your group.

<< How can i assure and where can i check that i have not lost any two disks of
same RSS ? >>


If you lose one disk drive, there isn't problems. The reconstruction begins from redundant data and after that the leveling takes place. The mirrored disk becomes empty. Everybody happy.
Now if you lose two disk drives in the same RSS set, then you will lose all VRAID 5
vdisks because of the reconstruction can't take place; but VRAID 1 data remains if the failed disk aren't a pair. The pair is formed by the two disk with the lowest RSSS ID consecutively, eg. 0-1,2-3... So if you lose two disk in the same RSS you will know it :(

<<- In an EVA array, Can anybody help me how to ensure that for vraid1, the mirrored partner (Copy) are in diff shelfes?>>

If your array is empty follow the best practices for joining disk. If you already had configured it, with a map of RRS state (Matt tool )you may ungroup and group the disk, so you would get the desired config. It is time consuming process.take in to account that when you ungroup a disk the RSS map may change drastically.

<< what are the important ares of EVA 5k and how to configure vraid5 in the same >>

You can not configure VRAID levels. The only choice is the selection of the appropriate VRAID level according your enviroment.

<>

To get the output from SSSU in a text file, you must write an script with the commands to use, because SSSU (V6, that i have) doesn't support redirection. Then you must use SSSU file command to execute the script.

If the answer were helpful, please submit points.
Ah, please visit my blog to include it in search engines indexes :)

Greetings
Oscar Maimo

[MillerB]

Hi Matt,
I would be interested in the RSS tool you have written if you dont mind sharing it. This would be really usefull in order to explain RSS concepts to some the the guys here

[McCready]

One important point to stress is that you will only lose data if you are unlucky enough that

- you have two (or three, for raid 6) fail

- Before the array has completed reconstructing data from the first failure and

- the failed disks are all in the same RSS group (for raid 5 or 6 data) or

- the exact mirror within the RSS group of the failed disk fails (for raid 1)

What you report to your management when a disk fails is either reconstruction has finished and all data is protected by a raid level (above 0) or reconstruction is happening and you are at some (small) level of risk until it completes. Reconstruction can take from a few minutes to a few hours (or longer, if you metadata issues like we did once)