Understanding EVA RSS and protection level

[Patrick Terlisten]

Hello,

I'm trying to understand the RSS and protection level of the EVA.

Where can I found more informations about it. I read many postings here in the forum, but I need more input. :)

I don't really understand how RSS works and why I can lose disks in a RSS, but my data is still available in my disk group.

Will a whole VRAID 1 rebuilded in the sparespace (space that is reserved by the protection level) or only the failed disks. I read that EVA marries disks to pairs and that this pairs are the reason why space is reserved two or four times.

How works it with VRAID 5?

Thanks for advice.

Kind Regards,
Patrick

[Uwe Zessin]

A disk group is devided into multiple RSSes (Redundant Storage Sets). Each RSS forms a complete RAID protection domain.

Disks are 'married' to pairs to define the VRAID-1 mirror members and both members are always contained in the same RSS. If one of the members fails, yes, the EVA picks up the data of the remaining member and stores it on a different member pair. Then it releases the data from the 'widow'.

By the way: you can see this behaviour if you create a disk group with an odd number of disks (e.g. 9) and fill it only with VRAID-1 virtual disks. One of the disk drives will remain empty!
That is why the EVA reserves 2x or 4x the space of the largest disk in a disk group for the "protection space" - the space set aside for rebuilding redundancy although it does not 'protect' anything - an unfortunate naming, if you ask me.


For VRAID-5, the EVA uses a 4D+1P mechanism: 4 data chunks on 4 different physical disk drives are combined to a parity chunk that is stored on a 5th physical disk drive. Those 5 disk drives are all stored in the same RSS.

The VRAID-5 data is equally distributed over all disk drives (6 to 11) within an RSS. It is possible, because a different 4D+1P stripe can be allocated differently to another set of 5 disk drives.

So one disk per RSS can fail without loss of data for virtual disks created in VRAID-1 or VRAID-5, because the redundancy is included in the RSS set of disks.

[Patrick Terlisten]

Hello Uwe,

thanks for you reply. I read a posting that described that a RSS hold between 6 and 11 disks. So if I have an EVA with 12 disks in one disk group there are two RSSes, right? Or an EVA with 16 disks and two disk groups will have one RSS in each disk group, correct? What if the EVA got 15 disks? IMHO there will be two RSSes.

A VRAID5 over 16 disks in one disk group with two RSSes will store the data in both RSS. IMHO I can lose one disk in each RSS, and after rebuild in the spare space (or free space), I can lost two disks again, right?

That so confusing. ;)

Kind regards,
Patrick

[Uwe Zessin · Accepted solution]

No. A single disk group with 12 physical disk drives makes 2 RSSes. Two disk groups each with 12 disk drives makes 4 RSSes, but RSSes from different disk groups are not related.

If you have an EVA with a single disk group with 15 disk drives, you still have 2 RSSes, but there are several possibilities

RSS1_disks-RSS2_disks: 8-7, 6-9, 9-6, 7-8

It is said that the EVA tries to build RSSes with 8 physical disk drives, but it is not always possible and the controller software sometimes failes to create the optimal case.

> A VRAID5 over 16 disks in one disk group with two RSSes...

Yes, a virtual disk's data is always spread across all RSSes in a disk group (modulo the exception for VRAID-1 and an odd number of disk drives in an RSS ;-).

And yes, after the redundancy in an RSSes hase been recoved, you can again loose one disk drive from this RSS.

[Patrick Terlisten]

Thanks again Uwe. :) But one last question:

---
Yes, a virtual disk's data is always spread across all RSSes in a disk group (modulo the exception for VRAID-1 and an odd number of disk drives in an RSS ;-).
---

The disks of a disk pair reside always in the same RSS. And that is why "protection level" reserves space for two or four disks: To rebuild one oder two complete "disk-pairs", right (if VRAID-1 is used)?

Regards,
Patrick

[Uwe Zessin]

Yes. The data is relocated from the remaining member to a new pair of disks and duplicated there.

[Patrick Terlisten]

Ahh, now it's clear. Thank you very much.

Kind regards,
Patrick