Syncmirror detailed description

[KEVINSNOOK]

Hi,

I'm doing some things with Metrocluster at the moment and I'm slightly surprised by what I'm seeing.

(This could be my misunderstanding of the syncmirror stuff in MC)

When I look at the plexes on each aggregate, they all display disks in their RGs that are locally owned. Is this correct? I've always been led to believe that the plexes in MC were "remote copied" i.e. one plex was local and one remote.

Can anyone give a detailed explanation of how syncmirror works in MC?

[vmsjaak13]

Local and remote: geographically yes. But DOT must own both in order to write in a RAID1 fashion.

Simple explanation: Without MC: Imagine site A with aggr0, which consists of 1 shelf.

In MC land, this aggregate needs another plex. So there's a shelf on site B, the disks in that shelf are owned by the same controller.

DOT now writes data to both shelfs (plexes) in RAID1 fashion.

You can also use syncmirror in non MC configs.

Then the 2nd shelf is geographically local to shelf 1.

For more info, read TR: "Best Practices for MetroCluster Design and Implementation"

Regards,

Niek Baakman

[KEVINSNOOK]

That's kind of how I understood it. The plexes here for all aggrs have both been set up in shelves that are all in the same rack (LOCAL). Clearly, the disks have been taken from different pools but the actual disks are NOT in different locations.

I have had cause to doubt my own understanding but it seems that this is misconfiguration! Can you confirm?

[vmsjaak13]

I can't confirm from here.

Is this a stretched or fabric MC ?

I've seen stretched MC in the same rack, or adjoining racks.

Regards,

Niek Baakman

[KEVINSNOOK]

Stretched (on opposite sides of a Data Center)