I’m looking for some actually working examples of using Flux to deploy at least Immich and Jellyfin with Traefik ingress using Flux CD preferably that use NFS shares for storing data and/or configs, and prometheus monitoring. I’ve found a few just from searching, but usually it seems like they aren’t actually working, just demonstrating something.

Also, what UI do you use for Flux?

Some background: I use k0s with MetalLB and Longhorn and all db storage on a separate postgres server. I’ve gotten a good repeatable installation of all of that as well as the FluxCD Operator via kOsctl and helm charts which allows me to simply launch k0sctl and provide it a single config that is well documented that deploys all of that across all of the servers. But Flux has been a challenge, especially since I don’t really care to learn Kustomize. I already have lots of other things to learn, LOL.

I want to avoid having to use lots of CLI commands that I then have to make sure to document and takes a while to remember all of the issues that can come up and all of that. I’m basically looking for easy disaster recovery. All files and configs are stored on a NAS that is then a single point of all offsite backups. It has worked well with docker, but I want better dynamic distribution of services to take better advantage of the few small servers I have. And docker Swarm ended up being difficult to implement with a lot of applications, and I don’t trust Docker not to further enshitify as well as the number of GUIs that support it well is limited, mostly just to Portainer which has been also enshitifying lately. So I’ve been looking to k8s.

  • saddlebag@lemmy.world
    link
    fedilink
    English
    arrow-up
    4
    ·
    18 hours ago

    Have you seen the Home Ops repos? This is what I followed to get full git-ops from 0 to production using just a few commands.

    This repo is quite complicated and I don’t think you need everything. At the heart of it is the App-template helm chart template. This lets you create similar deployments for most applications that don’t have their own helm chart.

    The flow is then to use flux to converge against a a git repository.

    I’m happy to answer any questions here or you can join the home-ops discord! It’s tons of fun to learn and the community is very supportive.

  • its_me_gb@feddit.uk
    link
    fedilink
    English
    arrow-up
    1
    ·
    18 hours ago

    I have pretty much this exact setup.

    • k3s cluster running on proxmox
    • FluxCD for Deployment (not operator)
    • NFS storage on truenas
    • Longhorn with SSD storage for PVCs
    • Traefik for ingress (recently switched from nginx ingress controller)
    • cloud native Postgres for DB
    • Prometheus/Loki for monitoring
    • sops for secret encryption

    That being said, all of the configs are stored in a private git repo that isn’t exposed to the internet, so I’m afraid I can share it all with you.

    Like others, I don’t use a UI for Flux, just a combination of VSCode, Vim, kubectl, k9s and the flux cli.

    Kubernetes is not for the faint hearted and is a massive step up/learning curve from Docker. I definitely wouldn’t recommend it for everyone, I’m ‘lucky’ that I also work with it most of the time in my day job.

    I can share my immich helm release later, but I’d suggest looking into bjw-s-app template and having a look at https://kubesearch.dev/