It depends what you're after; there seem to be a few different ways to try and monitor containers. For example, DataDog's agent will monitor the docker socket in order to get realtime stats. cAdvisor is probably the most powerful one around and is what Kubernetes is using.
Docker stats is not a solution since you can't configure it, it's a good effort but is not sufficient. Docker stats over hundredths of containers generates too much load, you can't configure anything. An external project is welcomed to fulfill this lack.
Step 1: Spend time implementing a monitoring solution.
Step 2: Realise what you actually need is security and availability monitoring, plus well tested automated resolution strategies for various classes of events.
Step 3: Go back to step 1, and include broader requirements. Simultaneously realise that your strategy for resolution (or indeed containerization) is not universally applicable, so you actually need to build an abstract system that enables the appropriate use of multiple architectures if you don't want the problem again in future. Scope creep!
8 comments
[ 3.7 ms ] story [ 29.5 ms ] threadStep 2: Realise what you actually need is security and availability monitoring, plus well tested automated resolution strategies for various classes of events.
Step 3: Go back to step 1, and include broader requirements. Simultaneously realise that your strategy for resolution (or indeed containerization) is not universally applicable, so you actually need to build an abstract system that enables the appropriate use of multiple architectures if you don't want the problem again in future. Scope creep!