[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: [Linux-cluster] distributed cluster



I'm doing something similar...

I'm using:
 - 4 IBM x3550 (four nodes)
 - 2 DS4700 storage
 - 4 Fiber Channel Switches (2 on each site)

What I do:
 I bought storage ERM (Enhanced Remote Mirror) feature and Brocade
Full Fabric feature (to comunicate between sites over switches)

I have LUN's active on site1 and backup on site2 and vice-versa.
You will need to use/configure a multipath I/O driver (IBM uses RDAC,
EMC powerpath... go on)
When both storages are configured to mirror LUN1 I see 4 paths to access LUN1:
 - Site1 -> Controller_A -> LUN1 (Priority: 0)
 - Site1 -> Controller_B -> LUN1 (Priority: 1)
 - Site2 -> Controller_A -> LUN1 (Priority: 2)
 - Site2 -> Controller_B -> LUN1 (Priority: 3)

In really, I can only use one path each time, but if one path goes
down, I follow to next one by priority. I'm using a mirrored qdisk
device too, in the same way.
But, if site1 loose comunication with site2 (and vice-versa), local
qdisk device can establish the quorum and provide service.
In my case, I'll never write data on both sites if loose contact each other.

The project is not complete, but I did some tests and dont see any big
problem with this setup.

You can see a draft of scenario here:
http://www.uploadimagens.com/upload/2b58a80ce8c93c205109c9703b6740f5.jpg

--

Flávio do Carmo Júnior aka waKKu

2009/3/11 Andrew Hole <andremailinglist gmail com>:
> Hello!
>
> I am trying to find the best approach for creating a distributed cluster
> based on the following specs:
>
> -          2 datacenters (close to each other and connected by fiber
> channel);
>
> -          2 storages (one on each datacenter);
>
> -          1 server on each data center
>
>
>
> In an event of a failure of a single datacenter, the remaining datacenter
> must be responsible for assuring all services.
>
>
>
> My approach would be to (I don’t exactly know if it is possible):
>
> -          Present one LUN from each storage to both servers:
>
>    server A (DC A) - LUN A (STORAGE A)
>
>                                               LUN B (STORAGE B)
>
>            server B (DC B) - LUN B (STORAGE B)
>
>                                       LUN A (STORAGE A)
>
>
>
> -          Create one logical volume on each of the servers using RAID1:
>
>     server A (DC A) - LVM A /mydata [RAID1 de LUN A e LUN B]
>
>     server B (DC B) - LVM B /mydata [RAID1 de LUN B e LUN A]
>
>
>
> -          Configure a RedHat Cluster;
>
> -          Put /mydata on the cluster;
>
>
>
> Is this a feasible approach? Is there a better option?
>
>  A.
>
> --
> Linux-cluster mailing list
> Linux-cluster redhat com
> https://www.redhat.com/mailman/listinfo/linux-cluster
>


[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]