As we were booting up npm agent on windows, we saw ~4K endpoints on a node with 16 pods, when there should be a 1:1 endpoint to pod ratio.
Issue Fixed:
Changed logic: When booting up the data plane, instead of getting all remote endpoints, we only get local endpoints which will be the endpoints for the pods in the node.
Testing:
Took a node with 11 non-system pods on it.
Ran windows NPM with new image and saw total of 12 endpoints getting reset vs ~4K that were getting reset originally.
Reason for Change:
As we were booting up npm agent on windows, we saw ~4K endpoints on a node with 16 pods, when there should be a 1:1 endpoint to pod ratio.
Issue Fixed:
Changed logic: When booting up the data plane, instead of getting all remote endpoints, we only get local endpoints which will be the endpoints for the pods in the node. Testing: Took a node with 11 non-system pods on it. Ran windows NPM with new image and saw total of 12 endpoints getting reset vs ~4K that were getting reset originally.
Requirements:
Notes: