cancel
Showing results for 
Search instead for 
Did you mean: 

Maximum number of nodes for Veritas Global Cluster?

inetcfg
Level 3
The documentation says 2, but that defeats the purpose.

Can any confirm that?
8 REPLIES 8

Gene_Henriksen
Level 6
Accredited Certified
The docs say 1 to 32 nodes whether it is on Solaris, HP UX, AIX, Linux or Windows. GCO supports 4 clusters. Replication between clusters is automatically supported between two clusters, but can be configured for more. Windows includes the ability to handle VVR with 3 clusters in GCO.


Message Edited by Gene Henriksen on 10-30-2007 10:41 AM

inetcfg
Level 3
Gene-

Would you be so kind to point me to the docs that state that?  I'm trying to get Symantec to confirm this for me, but I've been unsuccessful.

I'm reading the Installation and Configuration Guide for Linux, v 5.0.

On page 41, it states, verbatim, "note that a global cluster environment only allows a maximum of two nodes per cluster."

If that's the case, it doesn't make any technical sense to me.

If you have documentation that says otherwise, that would be of the utmost value.

Thanks,

Ken

Gene_Henriksen
Level 6
Accredited Certified
Are you looking at a specific application such as Cluster File system or are you looking at something generic like VCS and GCO? The limits I quoted are for VCS and GCO. For CVM and CFS because of the nature of the application, they may be different.

inetcfg
Level 3
Gene-

What I'm looking to implement is the Symantec Storage Foundation for Oracle RAC, which as you know consists of a few components.  These are Veritas Clustered File System, Veritas Cluster Server, Veritas Volume Replicator, and Oracle RAC extensions. 

What I'd like to do is use Global Clustering Option to make two geographically dispersed sites aware of each other.  Essentially, a global Oracle cluster.  Each site will consist of a 5 node Oracle cluster controlled by VCS, and will be synchronized using VVR.  I was planning on mirroring the number of database servers at the remote site as well.  However, if the docs state that I can only have 2 nodes per cluster, then well, this solution won't work for me.  That'd be really unfortunate, as I don't see any other vendor coming close to offering a complete solution.

Does this fall in line with the documentation that you have?

Thanks,

Ken

Gene_Henriksen
Level 6
Accredited Certified
This is from the VVR Linux version 5.0 Release Notes.

VVR supports replication in a shared disk group environment. This functionality of VVR enables you to replicate data volumes that are managed by the clustering functionality of Veritas Volume Manager. VVR also supports the clustering functionality of Veritas File System and Veritas Storage Foundation for Oracle RAC.

New features
VVR enables you to replicate data volumes in a shared disk group environment, for use with parallel applications that use Veritas Cluster Server (VCS) for high availability. You can replicate data volumes in a shared disk group to a remote site, for disaster recovery or off-host processing. VVR supports configurations in which both the Primary and Secondary disk group are shared, or either the Primary or the Secondary disk group is shared. If the Primary disk group is shared, it is not required that the Secondary disk group must be a shared disk group and vice versa. Currently, replication support is limited to 4-node cluster applications.

inetcfg
Level 3
Thanks Gene-

That certainly helps out.  It doesn't make ANY sense to why there's that limitation though.  If VCS/GCO relies on VVR to synchronize the data, and VVR is limited to 4 nodes, then well, I guess this solution isn't going to help.

Any thoughts with why there's such a low limitation?

Ken

Gene_Henriksen
Level 6
Accredited Certified
They started out with 2 nodes for VVR with CFM and increased it to 4. I am susre it has to do with coordinating all the writes to maintain write order fidelity on the local and remote.

inetcfg
Level 3
Gene-

Thanks for your help, and the information.

I'm going to contact my SE, and push him for an answer as to why the low limit of nodes, as well as a path forward to allow for more nodes.

I'll update when/if I get an answer.

Thanks again!