#BEAM and Kubernetes

1 messages · Page 1 of 1 (latest)

hidden pelican
#

I recently found gleam and I really want to dive into it, but I was curious if someone more experienced with the BEAM has any thoughts on best practices for deployment. More specifically about how the runtime works with Kubernetes. I'm curious if the recommended deployment pattern would be container based or something like an autoscaling ec2 group on AWS. I've done a lot of SRE work and have had personal troubles with the idea of putting a JVM inside a container running on a virtual machine, it's too much abstraction imo and it very quickly becomes a resource constraint, especially when the code is not super memory efficient.

Just curious what the thoughts are around containerization and orchestration.

tender yoke
#

I myself try to avoid excess abstraction and prefer to orchestrate using ansible on bare metal … but I never run thousands of nodes so then perhaps k8s is worth the effort

hidden pelican
#

oh cool, tbh I have no need for scale and could probably just deploy on bare metal (single server), but my use cases will mostly be for my homelab, and I have limited compute resouces, and when I want to also deploy OSS applications like grafana or miniflux, the orchestration and abstraction that K8s provides is really valuable for making a platform for many different use cases.

ocean wedge
#

I’ve never had any issues with putting the BEAM inside containers or k8s specifically. Can treat it like you would nodejs as resource wise it’s more like that VM than the JVM

#

I often don’t bother to set up clustering

#

I prefer deploying my software directly on the host using systemd or similar, but that’s a personal preference more than anything else

#

I think the best deployment strategy is whatever your ops folks are more comfortable with. The BEAM is pretty forgiving operationally