I am hoping someone could shed some light on an issue I appear to have regarding the installation of 11gR2 Clusterware. The main issue is that I have a host-vip.subdomain that fails to startup on installation. This is my 3rd cluster in my environment; the other 2 installed/configured ..fine. This install fails because the host-vip.subdomain does not resolve with its own GNS service.
Further investigation lead me down a path involving my other clusters. I found in my DNS server /var/log/messages file that the host-vip.subdomain was trying to resolve to host-vip.subdmain.subdomain_clust1 on IP xx.xxx.130.20. However, this cluster's GNS service is listening on xx.xxx.130.22
More detail on the environment:
I currently have in production, two 2-node clusters. Not RAC..just 2-node linux clusters; On the production clusters, I do have a single-instance database running in a 'warm failover' configuration. I do use SCAN to access each of the database on their respective clusters.
I am using GNS and DHCP (obviously) to generate the VIPs for each cluster. Cluster1 GNS IP - xx.xxx.130.20 ; Cluster2 GNS IP xx.xxx.130.21. Both are configured in DNS and resolve the SCAN address to each of the VIPs on that server.
When testing SCAN access to the database, I noticed that a tnsping DB1 - which uses SCAN-name1.subdomain1 connects fine. Also, tnsping to DB2 using SCAN-name2.subdomain2 connects fine as well. However, the weird part is that if I use each others subdomain..they still connect ...ie....tnsping DB1 using SCAN-name1.subdomain2 ...still connects to DB1.. albeit , it takes about 10x longer to get a response.
This has lead me to the idea that since GNS is basically a mDNS service, broadcasting on the subnet .130...could it be possible that ANY GNS service on that subnet could resolve a name lookup request for any other GNS on the same subnet , and during installation of a new server, cause a new VIP to go to the wrong GNS service ?
So, my question is this : Is there any requirement I may have missed that state multiple Clusters using GNS / SCAN MUST BE on different subnets, as to not interfere with each other's lookup requests ?