Subnostr

Tested - but not fully deployed yet. I have to get Jenkins working first to start dogfeeding the cluster new images it can utilize.

That said, just launch into an alpine-golang image (im sure this exists), mount this whole thing in /app/policy, point strfry to run /app/policy/policy_impl and you're good to go!

I can probably implement a sidecar container down the line that watches for changes and recompiles the utility on the fly - it's not a whole lot. :)

Now then... Off to fix Jenkins! I _need_ it.

Reply to this note

Please Login to reply.

Discussion

AU9913 1y ago

What's the reason you're self hosting kubernetes? Isn't that overkill?

Ingwie Phoenix (aka. birb) 1y ago

What mleku said - and, I have _way_ too many nodes under my control now. A RockPro64, NanoPi R6s, VisionFive2 and a 4 vCPU VPS with Hetzner (Ampere ARM - Not to be confused with the NVIDIA one). So... Building a cluster feels like the next logical step o.o Also... if I can build good and reusable deployments, I can publish them as Helm charts which may or may not end up coming in handy for future Nostr hosting. Lord knows how Primal and the others host their stuff; chances are, when Nostr grows, it'll need to scale - and this is something Kubernetes is quite good in. :) So while solving my own little conundrum, I might as well put together something that others can re-use down the line.

AU9913 1y ago

I manage 5 k8s clusters at work, but they're all in aws so it's way easier.

NY home lab has proxmox on it, so Im just running an lxc per service. So far I have the automation for the initial setup if a gitea lxc and ansible and I have it spinning up one more lxc.

Ingwie Phoenix (aka. birb) 1y ago

LXC is so neat! I am using that to rebuild my bitcoin/cln node, right now. Far less ephemeral than a container, but way less overhead-y than a VM. Its super neat! Though I use Incus; iirc, it's an LXC fork so... not too different.

I really wanna try Proxmox in the future, it seems like an amazing HCI solution!

AU9913 1y ago

It's been fun to play around with. I've only been doing the home lab stuff for a few months, but my intention was just to use start9, but within the first week I realized it didn't give me the control I wanted, so I've been working on this.

Once I get a few service spun up and migrated off my start9 vm into these individual containers, I want to start experimenting with ways abstract the state of the LXCs the that I could see what happens when have one instance of each DB flavor running on proxmox and then connect each app to a segregated DB on those so DB management is centralized which would enable a simplified DR solution.