Savithru Lokanath
1 min readApr 22, 2020

--

Michael LaRocca glad you found it helpful!

  1. Yes, that is correct. We set the env variable in aws-node daemonSet. If you’re use case needs multiple WARM_IP_TARGETS (per node basis), then you will need to have multiple daemonSets, each having an affinity for a specific set of nodes. You can play around with the nodeAffinity/toleration setting in the daemonSet spec to accomplish this.
  2. Yes, so when WARM_IP_TARGET=10 & you have a single pod, then there are a total of 11 IPs assigned to that node; 1 for the pod running on that node & 10 in the pool. Note that these are secondary IPs assigned to the interface of the node.
  3. Each interface has a limit on the secondary IPs it can hold. So if the limit is let’s say 30 IPs/interface, then in the scenario where there are >20 pods on a node, then the CNI plugin will automatically attach one more interface to the node in order to add more IPs.
  4. STEP3 kicks in once the 30 IPs/interface limit has been reached on the newly added interface & results in a third interface being added to the node.
  5. STEP3–4 continues till the interface limit has been reached on that EC2 node.

--

--

No responses yet