[Linux-cluster] Disaster recovery setup

[Maciej Bogucki]

brem belguebli pisze:
> Hello,
>
> I need to setup a 4 nodes cluster for hosting a few database instances.
>
> The main constraint being that this cluster has to be equally 
> distributed on 2 different sites (distance less than 100 km), ie 2 
> nodes on each site.
>
> The storage being based on 2 FC disk arrays, one on each site, 
> connected thru 2 SAN fabrics for redundancy and failover.
>
> Each database instance runs on one node, a should be able to switch to 
> another in case of machine crash.
>
> I would like to use CLVM with mirror (or mdadm) to give access  each 
> node to both disk arrays (local and distant). 
>
>
> In case something happens on one site(disaster) , the nodes on the 
> other site should be able to take over the failing nodes by hosting 
> the  databases instances that failed.
>  
> My questions concern :
>
> - Quorum: In case of a disaster, the remaining nodes won't have the 
> majority (2 over 4)
> - If I use qdisk, on wich site to put it ? the disaster could occur on 
> any site
Use qdisk and mirror it over network.
> - CLVM with mirror (or mdadm layer to built on top of it CLVM) is a 
> "supported " setup ?
Do not use mdadm for mirroring. If You SAN support remote mirroring use 
it or configure DRBD(it supports only two servers) [1].
This post [2] could help You also.

[1] - http://www.drbd.org/
[2] - http://archives.free.net.ph/message/20090311.200230.23f4f917.pl.html

Best Regards
Maciej Bogucki