Question: Do you intend to run any Windows containers, or are you keeping them as VMs and only use Linux containers?
Answer: We might. For now, the strategy is to migrate as-is to AWS which means traditional VMs. From there, we'll see how quickly we're able to transition to .net core and Linux-based containers. So at the moment, we're not actively researching Windows containers but that might change down the road.
Question: I see the version of Traefik is a little old (1.7.5). From your experience, how easy is to update traefik in production environment?
Answer: Very easy! For VM-based Traefik instances everything is managed by Ansible in Autoscaling Groups, so it's a matter of flipping the version, and perform a scale-out/scale-in to replace nodes. For Traefik running in Kubernetes its even easier (I even show this in as a demo in the webast). It's worth mentioning how important it is to tune the shutdown process in Traefik and make sure it aligns with how any load balancers outside Traefik is configured.
Question: Where do you see benefits/downsides in running Traefik as DaemonSet instead of a Deployment? Answer: For us it's just easier, since we're running relatively small clusters - it's also tied to how we route traffic thru AWS ALB to our clusters. But Kubernetes is super-flexible, and there's a ton of different ways to set up the Traefik pods.
Question: Do you load-balance in-node only or over all workers? Answer: We run Traefik on all Kubernetes workers, and AWS ALB evenly distributes load between the workers.
And, some links:
Meetup recording: https://www.youtube.com/watch?v=022gbgxYJZk
Slides: https://docs.google.com/presentation/d/1E_BpV5dkU3fftjW_HLeSWR-6Ss5eDgSa3IGWBXOhaj0/edit