VMware Cloud Director as a cloud management solution is built on top of the underlying compute and networking platforms that virtualize the physical infrastructure. For the compute and storage part VMware vSphere was always used. However, the networking platform is more interesting. It all started with vShield Edge which was later rebranded to vCloud Networking and Security, Cisco Nexus 1000V was briefly an option, but currently NSX for vSphere (NSX-V) and NSX-T Data Center are supported.
VMware has announced the sunsetting of NSX-V (current end of general support is planned for (January 2022) and is fully committed going forward to the NSX-T Data Center flavor. The two NSX releases while similar are completely different products and there is no direct upgrade path from one to the other. So it is natural that all existing NSX-V users are asking how to migrate their environments to the NSX-T?
NSX-T Data Center Migration Coordinator has been available for some time but the way it works is quite destructive for Cloud Director and cannot be used in such environments.
Therefore with VMware Cloud Director 10.1 VMware is releasing compatible migration tool called VMware NSX Migration for VMware Cloud Director.
The philosophy of the tool is following:
- Enable granular migration of tenant workloads and networking at Org VDC granularity with minimum downtime from NSX-V backed Provider VDC (PVDC) to NSX-T backed PVDC.
- Check and allow migration of only supported networking features
- Evolve with new releases of NSX-T and Cloud Director
In other words, it is not in-place migration. The provider will need to stand up new NSX-T backed cluster(s) next to the NSX-V backed ones in the same vCenter Server. Also the current NSX-T feature set in Cloud Director is not equivalent to the NSX-V. Therefore there are networking features that cannot in principle be migrated. To see comparison of the NSX-V and NSX-T Cloud Director feature set see the table at the end of this blog post.
The service provider will thus need to evaluate what Org VDCs can be migrated today based on existing limitations and functionality. Start with the simple Org VDCs and as new releases are provided migrate the rest.
How does the tool work?
- It is Python based CLI tool that is installed and run by the system administrator. It uses public APIs to communicate with Cloud Director, NSX-T and vCenter Server to perform the migrations.
- The environment must be prepared is such way that there exists NSX-T backed PVDC in the same vCenter Server as the source NSX-V PVDC and that their external networks are at the infrastructure level equivalent as existing external IP addresses are seamlessly migrated as well.
- The service provider defines which source Org VDC (NSX-V backed) is going to be migrated and what is the target Provider VDC (NSX-T backed)
- The service provider must prepare dedicated NSX-T Edge Cluster whose sole purpose is to perform Layer-2 bridging of source and destination Org VDC networks. This Edge Cluster needs one node for each migrated network and must be deployed in the NSX-V prepared cluster as it will perform VXLAN port group to NSX-T Overlay (Geneve) Logical Segment bridging.
- When the tool is started, it will first discover the source Org VDC feature set and assess if there are any incompatible (unsupported) features. If so, the migration will be halted.
- Then it will create the target Org VDC and start cloning the network topology, establish bridging, disconnect target networks and run basic network checks to see if the bridges work properly. If not then roll-back is performed (bridges and target Org VDC are destroyed).
- In the next step the north / south networking flow will be reconfigured to flow through the target Org VDC. This is done by disconnecting the source networks from the gateway and reconnecting the target ones. During this step brief N/S network disruption is expected. Also notice that the source Org VDC Edge GW needs to be connected temporarily to a dummy network as NSX-V requires at least one connected interface on the Edge at all times.
- Each vApp is then vMotioned from the source Org VDC to the target one. As this is live vMotion no significant network/compute disruption is expected.
- Once the provider verifies the correct functionality of the target VDC she can manually trigger the cleanup step that migrates source catalogs, destroys bridges and the source Org VDC and renames the target Org VDC.
- Rinse and repeat for the other Org VDCs.
Please make sure you read the release notes and user guide for the list of supported solutions and features. The tool will rapidly evolve – short roadmap already includes pre-validation and roll-back features. You are also encouraged to provide early feedback to help VMware decide how the tool should evolve.