Cluster Service group IP using IPMultinicB.

[unix_san]

Hello,

We have 2 node cluster using SFHA 6.0 in a Solaris 10 enviornment.  We are using Solaris IPMP on a O.S level and MultinicB/IPMultinicB on the VCS level for IP failover for any network failure.    

Can I configure the ClusterService service group IP to use the IPMultinicB instead of a single nic (as it is configured during the inital VCS configuration process)

[Gaurav_S]

Hi, Is it a simple cluster without any replication or global clustering configured? If it is then I don't see much downside as only time the IP will be used is to access Web console However if global clustering is in use and you have WAC resource configured then I would recommend to keep it separate G

[unix_san]

Thanks for the reply. 

Yes, we are using VVR and GCO too in the cluster. 

We are keeping the ClusterService service group sperate, but instead of the NIC resource can we use the multinicB resource in it.  Please see the config that we have, hope this will work fine.

cluster sample-cluster (
        UserNames = { admin = gJKcJEjGKfKKiSKeJH }
        Administrators = { admin }
        )

system node1 (
        )

system node2 (
        )

group ClusterService (
        SystemList = { node1 = 0, node2 = 1 }
        AutoStartList = { node1, node2 }
        )

        IPMultiNICB ClusterService_IP (
                Critical = 0
                BaseResName = MultiNICB_Res
                Address = "x.x.x.x"
                NetMask = "255.255.255.0"
                )

        NotifierMngr notifier (
                Critical = 0
                SmtpServer = mailserver
                SmtpRecipients = { some@someone.com = Warning }
                )

        Proxy ClusterService_Proxy (
                Critical = 0
                TargetResName = MultiNICB_Res
                )

        ClusterService_IP requires ClusterService_Proxy
        notifier requires ClusterService_Proxy

        // resource dependency tree
        //
        //      group ClusterService
        //      {
        //      IPMultiNICB ClusterService_IP
        //          {
        //          Proxy ClusterService_Proxy
        //          }
        //      NotifierMngr notifier
        //          {
        //          Proxy ClusterService_Proxy
        //          }
        //      }

group MultiNIC_Grp (
        SystemList = { node1 = 0, node2 = 1 }
        Parallel = 1
        AutoStartList = { node1, node2 }
        )

        MultiNICB MultiNICB_Res (
                UseMpathd = 1
                MpathdCommand = "/usr/lib/inet/in.mpathd -a"
                ConfigCheck = 0
                Device = { vnet2 = 0, vnet3 = 1 }
                DefaultRouter = "x.x.x.x"
                )

        Phantom Phantom_res (
                )

        // resource dependency tree
        //
        //      group MultiNIC_Grp
        //      {
        //      MultiNICB MultiNICB_Res
        //      Phantom Phantom_res
        //      }

[mikebounds · Accepted solution]

I have done this before for several customers and it works fine.  VCS performs operations on the ClusterService group, not the individual resources and GCO just needs the IP up and using IPMP makes it more available, so your proposal is a good idea.  Changing the ClusterService is the norm as far as I am concerned as for all my customers, even if they are using NIC, then I would add a Proxy in the ClusterService group to the NIC.

Mike

[Gaurav_S · Accepted solution]

This configuration looks very much OK, this will keep your IP is redundant & is available. May be I wasn't clear before, my recommendation was to have a separate IP, which configuration above has .. so all good

G