global vcs

[tanislavm]

hi,

if i have node a and b in site 1 and node c and d in site 2,then if i wish to configure them into global vcs in this way-nod a and node c serve a failover group and node b and d serve a parallel group.

i saw if i wish to setup gco first in site 1 i setup gco and then in site 2 i setup  the vcs cluster  and it will be joining next to the gco.

so in site 1 could i setup gco with node a and node c in server list of the failover group,even at this time the vcs in site 2 is not setup yet?

then could i setup node b and node d to server parallel group?it means that the disk group shared will be formed up by volumes mirrored accross the sites.or only i could have parallel service on 2 nodes within site 1?

tnx a lot.

[Sunil_Yadav]

Hi,

Please correct us if we are incorrectly interpreting problem statement. We assume your deployment looks like this:

       Cluster X               |          Cluster Y

                                    |

                                    |

Node A == Node B        |         Node C == Node D

                                     |

You have already done GCO setup between Cluster X and Cluster Y. For configuring global service group, you have to configure group in both cluster(with identical name). Then populate their SystemList with local systems(viz. Node A and Node B in Cluster X’s service group. Similarly in Cluster Y). thereafter, populate their ClusterList(Cluster X and Cluster Y on both clusters).

To answer your queries

• Nodes do not serve as failover/parallel. Service groups acts like failover(only 1 instance running in cluster) or parallel(multiple instances concurrently running in cluster).

• so in site 1 could i setup gco with node a and node c in server list of the failover group, even at this time the vcs in site 2 is not setup yet?

For a service group(even if it global), SystemList can contain systems from local cluster only. Thus, for failover group, you cannot add Node A and Node C in SystemList. You have to add local systems then populate ClusterList appropriately.

• then could i setup node b and node d to server parallel group?

Again, service group cannot have servers spreading across clusters. First populate SystemList(local nodes only) then ClusterList.

Feel free to let us know for any further query/assistance.

Thanks & Regards,
Sunil Y

[mikebounds · Accepted solution]

You cannot add a system in another globally connected cluster to the SystemList - you have to use ClusterList and ClusterList  does NOTcontain systems, it contains clusters.

The ClusterList is what makes a service group a "global service group" and if you create a global service group in the GUI using the wizard, then both clusters must be up and running, but you can manually edit ClusterList if the second cluster is not available (or if the wizard fails).  If you configure a global service group, then VCS prevents that group running on both nodes at the same time.  So if you want a parallel group across sites, then simply don't configure it as a global group.

So for example you can create cluster 1 as follows:

group failoversg (
  SystemList = { nodeA = 0 }
  ClusterList = { clus1 = 0, clus2 = 1 }

group parallelsg (
  SystemList = { nodeB = 0 }

This means locally each service cannot failover, but failoversg can fail across sites (nothing in cluster 1 that defines which nodes it fails to, just that it fails to clus2) and parallelsg cannot fail across sites.

So then cluster 2 would look like:

group failoversg (
  SystemList = { nodeC = 0 }
  ClusterList = { clus1 = 0, clus2 = 1 }

group parallelsg (
  SystemList = { nodeD = 0 }

So here it is defined that global group failoversg can failover to nodeC only (nodeD is NOT in SystemList) and as parallelsg is NOT a global service group (no ClusterList attribute), then it will online on both clusters.

You can define local failover (and parallel) attributes for local clusters so cluster 1 could be defined like:

group failoversg (
  SystemList = { nodeA = 0, nodeB = 1 }
  ClusterList = { clus1 = 0, clus2 = 1 }

group parallelsg (
  SystemList = { nodeA = 0, nodeB = 1 }
  Parallel = 1

 and likewise you can do the same for cluster 2

Mike