VMs not moving to ‘Write Back’ operational state when using vMotion TCP/IP stack.

When you configure VirtuCache for the first time, and you move VM(s) to  ‘Write-Back 1 Replica’ caching policy, if you suddenly find write latencies spiking and/or VM(s) not moving to Write-Back operational state (VirtuCache GUI > ‘Virtual Machines’ tab > ‘Operational State’) it could be because VirtuCache is using your vMotion network as the VirtuCache Replication network (this is the default configuration in VirtuCache unless the Replication Network has been changed using the VirtuCache GUI) AND your vMotion network uses VMware’s ‘vMotion’ TCP/IP stack.  VMware’s ‘vMotion’ TCP/IP stack prevents any other traffic, other than vMotion traffic, from running on that network. As a result, VirtuCache write-replication traffic cannot go over the vmotion network if the ‘vMotion’ TCP/IP stack is used. You should then create a separate network that uses VMware’s ‘Default’ TCP/IP stack and use that for VirtuCache Replication or change the TCP/IP stack on your vMotion network to be the ‘Default’ TCP/IP stack (see picture below).

The list-vms command (in esxcli > type vnxcli > type list-vms) will tell you why the VM might not be moving to write-back operational state.

TCP/IP stack in VMware’s VMKernel Adapters section