Right now the python runner retrieval works asynchronously. This could potentially lead to using a single runner multiple times. To get rid of this problem two approaches are possible.
Use vertx cluster wide locks and retrieve the AsyncMap only while having an active lock.
Mark created Kubernetes pods as being in use before delivering payload.
Upon start implementation decide for option one or two and the reasoning behind it.
Right now the python runner retrieval works asynchronously. This could potentially lead to using a single runner multiple times. To get rid of this problem two approaches are possible.
Upon start implementation decide for option one or two and the reasoning behind it.
Refs: This issue is a follow up to