RSS Disk state

[steveo_3]

Hi Guys,

I have 2 disk groups on my EVA3000. One has 16 disks, the other 8. The RSS disk state for the 8 disk group is mirrored, however for the 16 disk group, command view is reporting it as NONE. Is this correct, surely it should be mirrored too.
Also, going in to a particular disk in the 16 disk group, the RSS ID is 3, and the RSS index is 2. What does this mean?
I have just purchased 5 more disks, how best should I distribute them among the 2 groups to keep the RSS state optimun
Cheers for any help
steve

[Uwe Zessin · Accepted solution]

Each RSS has an RSS ID number (1...). The RSS INDEX (0...) is used to address disk drives within an RSS.

RSS state NONE happens when a disk pair (even and odd RSS index, e.g. 0+1, 2+3) is placed within a single disk drive enclosure.

A disk group should be configured with an even number of disk drives, because the mirrored chunks from a VRAID-1 virtual disk are always placed on disk pairs.

To fix the RSS state, you have to find out which pair of disks is involved.

[steveo_3]

Thanks Uwe,

I'm a bit new to EVA's. How would i go about determining which pair of disks are invloved in the RSS, and how would i rectify this once i've isolated them?

[Uwe Zessin]

I can find it out for you. Fire up a SSSU session and capture the output of either:

> set options display = 255
> show disk full

or (preferred):

> ls disk full XML

Put the output in a .TXT file and attach it to a new message - do not use cut&paste into the reply field.

[steveo_3]

thanks alot Uwe

txt attached

[Uwe Zessin]

Here we are - please see attachment.

[Uwe Zessin]

> and how would i rectify this once i've isolated them?

Sorry, I forgot:
I would swap the disk drives in bays 01:06 and 02:05 - this will keep the quorum disk 01:05 on enclosure 1.

[steveo_3]

Great!! Thanks

Appreciate all the help

[steveo_3]

Hi Uwe,

Just as a matter of interest, how did you go about analysing that output? Is there some sort of tool, in case I have the same issue again

thanks
steve

[Tom O'Toole]

I have some scripts that I might be able to post. XML SSSU output is processed by some VMS dcl command procedures and then displayed with an XSLT stylesheet. If you don't have VMS, this would at least kind of show you how to do it on other systems.

I'll let you know.

[Uwe Zessin]

Hello Steve,
I am sorry, but it is not a ready tool - it is just a set of library routines.

[steveo_3]

Thanks Tom, that would be great

No worries Uwe, cheers anyway

One more thing. When I swop those 2 disks around, would it be a good idea to shut down the eva first

[Uwe Zessin]

Yes, it is definitely a good idea, because both disks belong to the same RSS and one of them is even a quorum disk. Loosing both disks will very likely result in loss of data!

Another thing one might try is:
- increase the disk timeout
- move one of the disks to a different free bay
- wait for the system to settle
- move the second disk to the bay the first disk was in
- wait for the system to settle
- move the first disk to the bay the second disk was in
- set the timeout back to 1 minute

[steveo_3]

hi Uwe,

How would I increase the disk timeout?

cheers

[Uwe Zessin]

- make sure you have a good backup

- start Command View EVA

- click the EVA name on the left side to
-- see its "Initialized Storage System Poperties"

- look for "disk replacement delay" in the lower right corner
- change the value and press [Save changes]

[Tom O'Toole]

OK,

Here's the XSLT sheet you need to format the XML. The OpenVMS DCL command procedure which creates the XML follows. The instructions to run this are in the procedure. Basically you get the sssu output of show disk full and do a search to pick out the lines of interest before running this procedure:

[Tom O'Toole]

Yeeze, this interface is not the best..Anyway here is the dcl procedure. I forgot, you also need the CSS stylesheet, here it is inline.

td {border-width:1px; border-color:black; color:black; background-color:white; font-family: sans-serif;}
td.sml {font-size: 75%;}
td.c1 {background-color:#006666; color:white;}
td.c2 {background-color:#ff1493}
td.c3 {background-color:#ffff00}
td.c4 {background-color:#ffd700}
td.c5 {background-color:#00ffff}
td.c6 {background-color:#ffffd4}
td.c7 {background-color:#48d1cc}
td.c8 {background-color:#afeeee}
td.c9 {background-color:#000fff; color:white;}
td.c10 {background-color:#e0ffff}
td.c11 {background-color:#00bfff}
td.c12 {background-color:#ff6347}
td.c13 {background-color:#00cc7f}
td.c14 {background-color:#90ee90}
td.c15 {background-color:#00ff00}
td.c16 {background-color:#32cd32}
td.c17 {background-color:#dcdcdc}
td.c18 {background-color:#a9a9a9}
td.c19 {background-color:#ff00ff}
td.c20 {background-color:#d8bfd8}
td.c21 {background-color:#ff8c00}
td.c22 {background-color:#ba55d3}
td.c23 {background-color:#d2b48c}
td.c24 {background-color:#8faa8f}
td.c25 {background-color:#1e908f; color:white;}
td.c26 {background-color:#add8e6}
td.c27 {background-color:#ccff00}
td.c28 {background-color:#cc6633}
td.c29 {background-color:#333366; color:white;}
td.c30 {background-color:#666666; color:white;}
td.c31 {background-color:#ff9966;}
td.c32 {background-color:#ff9900;}
td.c33 {background-color:#990099; color:white;}
td.c34 {background-color:#009900; color:white;}
td.cxx {background-color:#ff0000;}
td.cvacant {background-color:white;}
td.ungrouped {background-color:black; color:white;}
span.alert {font: bold;}

[steveo_3]

Thanks Tom, I think that's gonna take me a while to figure out :)

Uwe,
forgive my ignorance, but when you refer to the quorum disk, does this relate to a cluster quorum or something specific to the eva RSS?

[Uwe Zessin]

Hello Steve,
I was talking about EVA's internal so-called "quorum disks". They contain special meta data.

[Tom O'Toole]

Steve, do you have VMS? If not and you want to create something similar, the main tricky thing that this procedure does is the calculation of the drive bay and enclosure from the loop id and looppair (I can't promise this calculation always works, but it has worked for several different eva configurations for me). Most of the other code is just for XML generation to create a pretty picture.

If you can get RSSid, RSSindex, enclosure and bay number from the SSSU output, you've done the important part - then just figure out if there any disks with the same same RSSid on the same shelf (this means you don't have parity, you have at most, mirrored). It sounds like you have a small number of shelves which may not be able to achieve an rss state of parity, but you should be able to get mirrored.

If there are any drives with the same RSSid and with consecutive odd, even rssindexes on the same shelf, you have an rss state of NONE - meaning both vraid5 and vraid1 are vulnerable to a shelf failure.

[steveo_3]

Thanks guys,

It seems odd to me how little information HP provide on this subject given how important it is

[steveo_3]

sorry, so just to reiterate.The fact that I have 2 disks next to each other on the same shelf with the same RSSID and RSS indexes which are numerically sequential is preventing the mirror state