[sles-beta] SLES Storage / MD RAID feature

[Decker Maximilian]

Hi Lars,

first of all thanks for your answer.

I'm talking about some sort of visually dumping the bitmap including the PVs involved.
So, you are right, I am basically asking for a way to show which regions in the md bitmap are marked dirty.

On HP-UX, this looks something like: (sample taken from host based RAID 1 over two datacenters)

<.....>
   --- Logical extents ---
   LE       PV1                     PE1      Status 1 PV2                     PE2      Status 2
   00000000 /dev/disk/disk84        00000018 current  /dev/disk/disk86        00000018 current
   00000001 /dev/disk/disk85        00000018 current  /dev/disk/disk87        00000018 current
   00000002 /dev/disk/disk84        00000019 current  /dev/disk/disk86        00000019 current
<.....>

This way you can see that e.g. logical extent No. 0 (LE) is on PV disk84 and disk86 and this specific
extent is in sync ("current"). On large volumes you get quite a long list here, but that is on purpose.

It really is a cool way so actually "see" the status and physical location of each extent. Especially
if a LV is spread among lots of PVs this information is quite vital.
Also, you can get a transparent view on how far a re-sync progress already finished. 

IMHO the ability to manually re-sync parts of the RAID or specific extents is not really necessary as with 
a md bitmap in use only the dirty parts are re-synced automatically, providing a decent nice solution here.

Ok and we cannot really compare the two platforms here as they differ quite a lot behind the scenes, but I
think some way of getting such detailed information out of the md bitmap should be possible.


- Maximilian 

-----Ursprüngliche Nachricht-----
Von: sles-beta-bounces at lists.suse.com [mailto:sles-beta-bounces at lists.suse.com] Im Auftrag von Lars Marowsky-Bree
Gesendet: Freitag, 15. März 2013 15:59
An: SUSE Linux Enterprise Server Authorized Beta Program
Betreff: Re: [sles-beta] SLES Storage / MD RAID feature

On 2013-03-15T13:47:42, Decker Maximilian <Maximilian.Decker at cancom.de> wrote:

Hi Maximilian,

> General recommendation in Linux land (AFAIK) sees LVM for just the 
> volume pools and MD as the RAID layer (so no LVM RAID).

Yes. For non-clustered mirrors, this is preferrable.

> So, long story short - something like this already on the SLES roadmap ?
> Or maybe already there and I just missed it ?
> (/proc/mdstat only shows a very generic form of status display here)

I don't have access to a non-Linux system, so I'm not sure what you're asking.

Are you asking for something to show you which regions in the md bitmap are dirty?

(The MD device always syncs as a whole, you don't get to pick which parts to sync exclusively. This will not change.)


Regards,
    Lars

--
Architect Storage/HA
SUSE LINUX Products GmbH, GF: Jeff Hawn, Jennifer Guild, Felix Imendörffer, HRB 21284 (AG Nürnberg) "Experience is the name everyone gives to their mistakes." -- Oscar Wilde

_______________________________________________
sles-beta mailing list
sles-beta at lists.suse.com
http://lists.suse.com/mailman/listinfo/sles-beta