Improve 1.23 documentation

[zyberon]

Hello I have just downloaded the new version, but I'm not sure if the documentation is updated to the new feature, how this works with the ASG to cordon and drain the nodes before shutting down, does the new nodes wait for the old ones to being deleted? and it is getting the events through the metadata?

Also is this tool making the nodes drain gracefully and one by one? what i mean is that when the asg send the termination event is the tool prepared to manage that events as a queue and proceed to drain a node, add the new not ready node to the cluster, and then proceed with draining the other node.

Thank you so much in advance.

[Lu-David]

Hi @zyberon,

Will update the docs to be clearer but to answer your question:

how this works with the ASG to cordon and drain the nodes before shutting down, does the new nodes wait for the old ones to being deleted? and it is getting the events through the metadata?

1. Yes ASG Lifecycle state support is getting events through instance metadata.

Also is this tool making the nodes drain gracefully and one by one? what i mean is that when the asg send the termination event is the tool prepared to manage that events as a queue and proceed to drain a node, add the new not ready node to the cluster, and then proceed with draining the other node.

2. It depends which mode of NTH you are using.

There is no queue if you are using IMDS mode. Each node listens to this endpoint : /latest/meta-data/autoscaling/target-lifecycle-state according to the docs. Whenever the node will soon be entering the Terminated State, NTH will then cordon and drain. So each node is draining asynchronously.

However, if you want to use ASG Lifecycle with Queue Processor, then yes the nodes would get cordoned and drained one by one

Hope this helps! Thanks