This message was deleted.
# intros
s
This message was deleted.
m
Welcome, Kevin! This is the thing I've been wandering - Why does anyone pick Kubernetes for projects and teams that are so small?... So much technical debt is incurred that way. What I'm voting for at all times - if you really need a dynamically-scaled system, Docker Swarm will be good enough for you. 1-2 people on the platform can barely handle even a fully managed solution. Kubernetes is an overkill. Thoughts?
(can't delete this message)
k
Hi Mario. Actually in my experience, Kubernetes (managed by the hyperscaler) runs pretty smooth. I do not have the feeling that there is much technical dept except updating the cluster. I was responsible for the platform for a small startup and was on my own. That worked out quite well. The advantage in my opinion is the big landscape of open-source tools paired with pre-build Helm-Charts. I was able to provide a lot of helpful services which otherwise would have taken me longer to set up with something like Swarm.
d
Welcome to the community!
m
@Kevin Scholz Nice to hear! Can you maybe tell for how long have you been running Kubernetes on these projects? Any outages you needed to tackle? Thanks!
k
We used GKE (I think around v1.16) on GCP and Kubernetes itself never had any outages. There was one issue that when we updated nginx-ingress, the load balancer still routed traffic to old containers, which prevented us from zero-downtime updates of NGINX. I think it is tracked here, which seems to be a general K8s problem. Otherwise everything else was our misconfiguration (which did not happen often). One learning though was to not over-provision memory 😄
Otherwise the stack was pretty default: • prometheus • EFK • nginx-ingress • cert-manager • jaeger • keycloak
m
Thanks!
cert-manager is a fine piece of software, I really liked it 🙂
when you say "100 containers" you mean 100 different apps/services?
k
Basically, yes. Although one application of ours (e.g. a Rails app) can have multiple containers.