We from KPN always keen to use latest and greatest technologies. In today modern technology world Container is no more fancy word. We would like to have the docker container built to run the UIM Robot within Container.
Current coupling between Robot and monitoring probes does not allow to have repeatable dockerized image for scale-out containerized solution.
as we are internally working on our drawing boards for the neat solution i have a query: For most part of usage, is probes fixed for a container i.e. you upfront know which applications are being contained and monitored?
Depends on what exactly you mean. Since containers are usually well defined before startup and tend to reflect a single type of service you usually know what you will want to monitor before starting up a container. But will all containers provide the same service and need identical monitoring, no I think is the general answer.
What I have experienced though is that because of the mobile nature of a container, one doesn't care about the name of the container being fixed over time. One cares if the whole cluster is providing the services it is supposed to. In our usage of containers, it is sufficient to know that one of the three database containers has failed on a given cluster. It doesn't matter anywhere as much that it happened to be running on worker node 7 at the time or was the third of the three instances to be started up.
So I'd really think about developing an equivalence between the concept of Robot and a rule governing the running of a set of related containers. I don;t care where a container is running if the rule is satisfied. And when there's a failure there's a rule violation - I care about that.
Also there's an overall concept of utilization - containers will get moved around to satisfy resource constraints so I really don't care about individual nodes - it's analogous to caring about the utilization of a single core of a CPU or not. I care if any one is at a maximum but not if core 3 specifically is at max. Same with containers - I care if there's enough resources to start another container, not really if one worker node is at 50% and another is at 25%.
Retrieving data ...