Based on one recent interview experience I realized I feel really good, when some one challenges my technical ability. For me it not important that I win the challenge but it tells where I’m currently and which are areas for improvement. But then this externally driven, as I’m firm believer of IDIOTS(I’ll Do It On my Own TermS), I decide to take this task inside my area of control. Following posts are solid example of it and also shows I’ve really started thinking.
I’m currently reading Advance Technical Design Guide, I’m reading this second time but only now I realize something unique.
Maximum number of Paths per volume are 32 and maximum size of cluster node is also 32 in vSphere.
Now in almost all production scenario there at least 2HBA or 2 iSCSI hardware initiator. So volume will have 2 paths coming from two HBA’s per node.
So if there are 32 nodes in a cluster, each with 2 HBA. Then each volume will receive 32x 2=64 paths. Maximum path is already crossed. So either reduced the node size or reduce the HBA. As former is much desired in all cases, you can easily come to the conclusion that you can have no more than 16 node cluster with 2 HBA’s per node.
So even if HA node size is increased from 16 to 32 in vSphere to bring it in synch with DRS, practical limitation still exists and needs to be addressed.
Maximum number of HBA’s which are supported in vSphere are 8 and maximum of HBA ports are 16. With maximum number of HBA you can configure only 4 node production cluster and with maximum HBA ports you can make only 2 node production cluster, as you are limited here by Paths per Volume which is 32.
so to conclude
With 2 HBA you can create up 16 Node Cluster
With 4 HBA you can create up 8 Node Cluster
With 8 HBA you can create up 4 Node Cluster
With 16 HBA you can create up 2 Node Cluster
So as your HBA increases your node size decrease.
Disclaimer: These are just theoretical limits and maximum supported configuration. It is very likely you can prove maximum configuration guide wrong.