RHEV Networking Best Practices (managment, interconnects, etc...)
I have opened a case to get feedback from Red Hat, but I was interested in others "real world" opinions and experiences.
We are deploying our first "prod" RHEV 3.1 cluster. Our host requires acces to multiple VLANs which we will accomplish with multiple physical interfaces (and not VLAN tagging).
We are wondering what the "Best Practice" is regarding the cluster interconnects. Should we have an interface dedicated to the RHEV-M traffic and then additional interfaces for the cluster to communicate with each of the nodes? Or - should we simply have multiple RHEV-M interfaces, bonded together?
Please let me know if my question requires further clarification.
Thanks!
Responses
The best configuration is to have a bonded interface for rhevm network and another bonded interface for cluster interconnect, if it's possible to spare a total of four NICs for this configuration.
If this is not possible (only two nics can be used), then there are two options.
- One bonded interface that shares rhevm and cluster interconnects. This has the side effect of losing mangement network if clsuter interconnects flood the network or losing the cluster interconnect if rhevm logical network gets overloaded with live migration and other management traffic.
- Separate management and cluster communication to its own interface without bonding. This has the side effect of losing network communication if interface goes down.
Welcome! Check out the Getting Started with Red Hat page for quick tours and guides for common tasks.
