2022-02-16 14:54:18 UTC - Pramod S Rao: Hello everyone. I am new to openwhisk,
and i've been reading up a lot on it. I had a few conceptual questions on how
openwhisk is deployed on Kubernetes, and couldn't really find the exact
documentation on how certain things are happening. Hopefully this is the right
place to ask.
1. For the normal deployment (without k8s), i understand that the invoker
starts the container and injects the code via the /init and /run http calls.
For k8s I can see a pod running for the invoker, and when i invoke an action
(e.g. python hello world), a new pod is created and the code runs inside the
container within the pod. My questions here are:
a. How does the invoker tell the kubernetes cluster to start the pod?
b. How is the code injected into the pod?
c. If it’s possible to invoke it directly, why bother using kubernetes?
2. In the normal situation without k8s, openwhisk invoker has a container pool
where it maintains information on the warm containers, and in case of
invocations, code is run inside this warm/pre-warm container.
a. When k8s comes into the picture, how is the container pool maintained?
Is it now a pod pool instead?
b. How does the invoker now know which pods are warm since the information
lies with k8s i assume?
c. What is the flow for a warm start if openwhisk is deployed on k8s?
wave : Rodric Rabbah
https://openwhisk-team.slack.com/archives/C3TPCAQG1/p1645023258824749?thread_ts=1645023258.824749&cid=C3TPCAQG1
----
