OBSERVABILITY, TELEMETRY, AND THE REALITY OF gNMI IN MULTI-VENDOR ENVIRONMENTS
While many networks are in freeze mode and changes are on hold, there is finally some space for experimentation, learning, and simply playing around with topics we normally postpone. This time at ITcare we decided to spend it diving deeper into observability using telemetry, and more specifically gRPC and gNMI. Very quickly, frustration kicked in. From one version to another and from one vendor to the next, support looks wildly different.
How are engineers supposed to learn and test streaming telemetry in a standardized way such as OpenConfig, if access and consistency are so limited? In virtual environments like Containerlab or EVE-NG, support is still very thin.
With Arista Networks, everything works almost out of the box. With Juniper Networks, I had to hunt for a very specific image version to make it usable. With Cisco, I am still stuck. Even though XRv 7.9 and newer should support gNMI, I cannot get it working properly on 7.11.1. The process does not even start, and no socket ever opens.
I am lucky to have access to a physical lab with Juniper and Arista gear, but I still have not figured out a clean way to test this on Cisco Nexus or XR platforms.
Has anyone successfully played with gNMI telemetry in a virtual environment? If yes, I would really appreciate hearing how you made it work. In a week or so I’ll share updates about where we managed to get with gNMI and multi-vendor environment.
