Jason Aliyetti

My knowledge of the codebase is limited but I'm wondering if making the [LinkList call](https://github.com/cilium/cilium/blob/3068add5d64d3e122b0d9d44a27edfed0f3f3653/pkg/ipam/crd_eni.go#L139) retry would mitigate this.

Doing some digging, I noticed about 5 months ago the netlink library Cilium uses had a change that may be responsible for this suddenly behaving differently: https://github.com/cilium/cilium/blob/7cf5782920423b07e7a4e6e67fafd81e1d4a1d33/vendor/github.com/vishvananda/netlink/nl/nl_linux.go#L563 —> https://github.com/vishvananda/netlink/commit/aa4f20db57d498a8db66aaa202ddae15c1fa81a5 I'm...

> I do believe #32099 was a good change, however I'm still seeing this error (and a lot of slowness) a lot in my tests even with the patch Not...

I can confirm that I still see issues in 1.15.6.

Actually, I don't see "timed out waiting for ENIs to be attached" in the logs, just the message indicating the retry.

Isn't the "failed to obtain eni list" message just indicative that it had to retry? I think there's only a real issue if you get "timed out waiting for ENIs...

@vivekgkwd see above. That log message indicates an error that should get retried. Does the issue not resolve itself and allow the pod to come up? There was another issue...