We use “project.affinity” to control the scheduling of services to our worker nodes.
If we want to update a bare-metal worker node, what happens in the cluster if we remove the “project.affinity” from the node?
Will it pull the rug from the service or will it do a “graceful” move of the service to another node and switch to the new pod before terminating the first service?
Furthermore, is the behavior configurable i.e some services need graceful “moves” between nodes and others don’t.
Thanks in advance
Have you tried this in a sandbox environment and, if so, what were your observations ?
It’s ok to ask questions of course, but it’s useful for those willing to offer an answer to understand what you have found not to work from the efforts you have made thus far.
Sorry for the late response,
The question was more theoretical as a “what if” question.
No we have not tried it in a (for now missing) sandbox environment.
The reply has now been sitting unsent for a very very long time I must have been interrupted while writing it.
I appreciate the answer and feel a bit of a fool not having replied earlier.
No worries … busy times for us all