You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Multiple CNIs can run in parallel, creating multiple data planes in a Kubernetes cluster. Goldfinger should allow me to ping over an arbitrary data plane.
It would seem that POD_IP is the place to set this, but when I set POD_IP for each goldpinger pod to an IP on an alternative dataplane, this setting is lost after discovery; goldpinger finds the peer pods but uses the primary data plane IP addresses.
Not all data planes are discoverable from the Pod API. E.g., Multus data plane specs are in Pod annotations. Meshnet-CNI specs data planes in a CRD that refers to a Pod resource by name.
The text was updated successfully, but these errors were encountered:
While a little niche, I think that's a good idea. I wonder how we'd discover the alternative IPs, if they can't be provided by the Pod API (getting them from annotations would be straightforward).
I'd be happy to review a PR, let's just make sure we still default to the current behaviour.
Multiple CNIs can run in parallel, creating multiple data planes in a Kubernetes cluster. Goldfinger should allow me to ping over an arbitrary data plane.
It would seem that POD_IP is the place to set this, but when I set POD_IP for each goldpinger pod to an IP on an alternative dataplane, this setting is lost after discovery; goldpinger finds the peer pods but uses the primary data plane IP addresses.
Not all data planes are discoverable from the Pod API. E.g., Multus data plane specs are in Pod annotations. Meshnet-CNI specs data planes in a CRD that refers to a Pod resource by name.
The text was updated successfully, but these errors were encountered: