RSS and EV5K

[John McF]

We have an EVA 5K 2c8d, our issue is that the RSS numbers and index's aren't optimally configured as per the number of drive shelf's. We have 96 disk's and 12 RSS's, however 8 have 8 members, 2 have 6 members and the last 2 have 10 members. We are running VCS 3028.

Recently we played the drive juggling game to stripe the RSS's accross the shelves, with the idea of improving the resiliancy of the array to shelf failure (highly unlikely but it was something that was bugging us). The plan after shuffling the drives around was to ungroup the members of the 10 member sets and hope that on regrouping they would append to the lessor number RSS groups - which they didn't.

We have since removed an ungrouped disk from one of the larger number RSS's and replaced it with a fresh drive - it too assumed the original RSS number and index of the larger RSS.

We would like to achieve 12 RSS groups of 8 memebers stripped accros the array. Is there a away to achieve this at the current VCS and if not will VCS 4.00x allow for this to occur. Or is there another way of achieving the required configuration?

[Luk Vandenbussche]

Hi John,

It seems to me that you have multiple diskgroups in your eva?

RSS are build at diskgroup level
RSS are made up to date if you add or remove disks in your eva.

So can you confirm how many diskgroups you have and how many disks there are in each diskgroup?

[John McF]

Apologies on that omission. 1 disk group, all disks in.

[Pshedges]

You probably know this already but this is what I recieved from HP on keeping RSS indexes on different shelves.

To help ensure RSS Disk states are not unintentionally altered keep to
the following rules:

When a disk fails in a storage subsystem:

* Replace the failed disk drive before further modifying the disk group
configuration by adding or removing disks to/from it.

* Replace the failed disk with a disk of the same or larger capacity.

* Insert the replacement disk into the same physical drive location from
which the failed disk was removed.

When adding disks to the array:

*Ensure total number of disks is a multiple of 8, and align disks
vertically across the enclosures.

As for re-arranging your current RSS Indexes I would assume that you would want to ungroup a disk that is within an RSS group of 6 - this should invoke a recalculation of the RSS Indexes and hopefully rectify the index numbers - I have not tried this so I would check with HP first.....

[Madis Tuulparg]

Hi,

We have same problem and we have done all disk replacements and adding correctly. Today we'll try with HP support to correct this problem.

PS We have 8 shelves and 1 disk group as described here.

[Tom O'Toole]

One big problem with having 6 member RSSes - if a disk fails in such an RSS, the remaining members will be migrated to other RSSes, and the algorithm that does this doesn't appear to take care that members of these RSSes end up on different shelves - Thus you can go from parity to mirrored (or none) by losing a single disk! It can be very hard to get back to an ideal configuration with all RSSes of 8 members and RSS state of parity. Even with this ideal state, if you are not hasty replacing failed drives before another fails, you can get degradation.