Clustering with redundant SAN storage

[Lee Chee Wei]

Hi,

I am planning to implement clustering solution (MSCS) over two sites (a few hundred meters apart), with redundant storage on the two sites as well.

May I know if the SAN controller has any functionalities to mirror data to another physical SAN storage at another site?

Is it possible to automatically failover the storage to allow the cluster to allow information from the mirrored disk?

I'm pretty new to SAN systems so need any advise here.

[Ganesh chichakar]

Hi Lee
Well I would like to know what do you mean by redundant storage on two sites, do you mean that each site will have two storage boxes?

Yes there are storage controllers like EVA and XP which will allow you to replicate the data across the sites which MS GDC does not.

and what do you want to say in your third question? i guess you want to know whether automatic failover is possible in between two sites m i rite? if so then well i guess yes.

Welcome to SAN :)

[Lee Chee Wei]

Hi, thanks for your reply! I am using Microsoft Cluster to create a cluster, which spans over two sites a few hundred meters apart. However, the common disadvantage of clustering is that it requires a shared disk. Hence, the disk can only be placed in one of the sites. If the site goes down, the cluster will fail and hence storage becomes a single point of failure.

I'm wondering if I can have another SAN on the other site, so that data is mirrored to the that storage system. Wonder if microsoft clusters support this configuration, and will automatically write and read from the mirrored disk?

[Lee Chee Wei]

Hi, there's another doubt I have. Does SAN store information into individual disks? Let's say i have allocated a mirrored disk pair, can I then remove 1 of the disks, and insert a blank disk and have the blank disk synchronise with the esisting disk?

Sorry if I might have confused the way SAN works with other storage systems.

[JR-Pounds]

I've done exactly what you've described for a number of IT organizations using the EVA with Continuous Access (CA). CA is a replication solution that allows EVA arrays to "mirror" each other. The MSCS cluster has two methods for achieving quorum, shared disk and majority Node set (MNS).

CA mirrors write I/O operations at the array controller (block level replication) and offers a myriad of configuration scenarios.

While MSCS will failover automatically from server to server, synchronizing the storage failover with the server failover requires an arbitrator. HP offers a product called CLX for Microsoft MNS clusters (cluster extention). For shared disk clusters, failover scripting is required to make all of the pieces shift from site to site. This can be fully automated, but most shops will opt to manually trigger an automated failover because of the complexity of the decision making process and responsibility associated with moving production operations to another site.

[Patrick Terlisten]

Hello Lee,

for a campus cluster you need a storage which can mirror (EVA or XP) and the Cluster Extension for EVA or XP. Otherwise the cluster won't failover successfull if one site dies. CLX isn't supported with a quorum-based cluster, so you need at least 3 nodes for a Majority Node Set Cluster.

Hope this helps.

Best regards,
Patrick