A good friend of mine despatched me a hyperlink to a prolonged convoluted doc describing the 17-step process (with the final step having 10 micro-steps) to comply with if you wish to run NSX supervisor on prime of N-VDS, or as they name it: Deploy a Absolutely Collapsed vSphere Cluster NSX-T on Hosts Operating N-VDS Switches1.
You won’t be conversant in vSphere networking and the best way NSX-T makes use of that (during which case I can extremely suggest vSphere and NSX webinars), so right here’s a CliffsNotes model of it: you wish to put the administration part of NSX-T on prime of the digital swap it’s managing, and make it accessible solely by that digital swap. What might presumably go fallacious?
Nicely, even VMware technical advertising didn’t danger NOT describing the most important caveat:
In a single cluster configuration, administration parts are hosted on an N-VDS swap as VMs. The N-VDS port to which the administration part connects to by default is initialized as a blocked port because of safety concerns. If there’s a energy failure requiring all of the 4 hosts to reboot, the administration VM port shall be initialized in a blocked state2.
To “resolve” that, it’s important to comply with the ten substeps of step 17. These steps contain writing JSON paperwork in a textual content editor and executing curl. Yeah, that’s positively one of the best they technique to configure any product. The final time I’ve seen one thing like that was the Wellfleet Technician Interface the place you had the privilege of configuring the field by writing values into SNMP OIDs.
For these of you who nonetheless don’t recognize how ridiculous the entire concept is: think about migrating from EIGRP to OSPF, however with the next minor limitations:
- There isn’t a out-of-band interface (fairly regular to date).
- You can’t run the routing protocols in parallel (WTx?)
- It’s important to sort in particular person instructions (OK).
- You can’t do cut-and-paste, and also you positively can not cheat by copying instructions right into a file and executing them regionally (OMG).
- Each command is executed instantly (as typical) and instantly saved into the everlasting configuration file (WTx???).
- Reloading or power-cycling the field can not convey the field to a earlier state (I’m out of right here)
- Whereas there’s some rollback functionality, you may have to have the ability to attain the field to execute it (enjoyable occasions).
Now don’t get me fallacious. If somebody has a determined urge to take part in Pink Bull Flying Contest, I’m all for it, however describing the ensuing Hero’s Journey as a part of official product documentation is a bit excessive. I can solely hope there was a ginormous buy order behind this requirement; bravado for bravado’s sake by no means made a lot sense to me.
Cautious readers would possibly level out that Nutanix makes use of the same trick: a VM exposing an iSCSI or NFS goal is operating on prime of a hypervisor utilizing that very same goal3. There’s only a slight distinction between the 2: Nutanix comes as a prepackaged answer, not as a unfastened assortment of hard-to-fit elements and IKEA-like directions.
Lastly, let’s assume VMware does care sufficient about prospects who wish to deploy NSX-T on a 4-node cluster. The one sane technique to meet that requirement could be to create a prepackaged one-click answer (aka “automate it”, however with correct rollbacks when an error is encountered), not a Rube Goldberg machine. Come on, VMware, we all know you CAN do higher than that.
Launch Historical past
- Footnote: Nutanix doesn’t create round dependencies (based mostly on the remark by Erik Auerswald).