On the most foundation part is Kubernetes with multiple different container images.
On top of them, we have a python library which dynamically loads our code, which we write and install on those containers after deployment. We write stateful applications on k8, that's why this kind of architecture.
Let's call our code beta
and the lib which loads it alpha
alpha
, which loads beta
, which we actively developbeta
, which gets loaded after running alpha
on the k8 pod -> containerFor now the possible solution was to do the following: - add a remote debugger to the code, which listens upon start - port forward the traffic - clone the code (install), such that the lib gets run, as soon as you run the main - run the code - it waits until attach - attach to the debugger.
Voila it works.