Metrocluster with SRDF

[Joon]

Hi all

I am considering about Metrocluster with SRDF after EMC SRDF DR solution setup.please have a look attached file(present configuration and after considerable configuration-metrocluster/SRDF)

My Questions are..

1.Is the attached file's configuration possible and any risks when change from local MC/SG to metrocluster??

2.Do I need to setup third site called is "Quorum server" or "Arbitrated server"?

3.What is different between Quorum and arbitrated server? and what are they roles?

4.What are advantage and disadvantage for metrocluster?

5.DWDM is being used for Data sync.what about node connection - heartbeat,public - should I use T1 or T3? I heard ATM not support metrocluster.

6.local MC/SG use shared disk as you know.
I am wondering metrocluster mechanism about shard disk(EMC).Is it possible for EMC disk to be separated in each site?
After MC/SRDF installation in each node,what happen to Primary and Backup site's EMC disk?
are they changed to one logical shared disk??

Thanks in advance to all of answerer.

[melvyn burnard]

Configuring MetroCluster is not an insignificant task.
I strongly recommend you read the manuals at
http://docs.hp.com/hpux/ha/#MetroCluster

and then also consider getting some consultancy assistance from a trained consultant

[Stephen Doud]

1. Best have an HA consultant do a thorough review of the rearchitected plan.

As Melvyn said - have a long look at this document: http://docs.hp.com/hpux/onlinedocs/B7660-90013/B7660-90013.html

2. Serviceguard will perform a cluster reformation when heartbeat transmission ceases between nodes in the original cluster. If a total heartbeat network outage occurs between a majority and minority of the nodes in the cluster, the majority will reform a new cluster and the minority will reboot. In the event that the all heartbeat networks break down, splitting the cluster in half, cluster arbitration is needed in order to determine which half will take over cluster operations, and which half will reboot. A quorum server can arbitrate at 50% split (clusters N and S), but in your re-architected plan, the M cluster has a minority at the DR site - it will always reboot in such a network outage. It would be necessary to have 2 nodes at a 3rd site to make the DR site a majority if a HB network outage occured between the primary site and the DR site.

3. A quorum server is not a member of the cluster - it merely provides arbitration for a 50%/50% HB network split. Arbitration servers on the other hand add to the total number of servers in the cluster. This helps define whether a HB network outage between sites results in a majority of nodes being on one side of the network outage or the other. No arbitration (cluster lock or quorum server) is needed in such a split.

4./5. Recommend the manual for this.

6. EMC has an SRDF facility, which Metrocluster supports. The Metrocluster product includes a special control script that manages changing R2 sites into R1 sites when necessary. Again - read the manual for a larger discussion.

-=StephenD.