3 Node Deployment (Unified Dashboard)

[plusone-nick]

@will They are freakin amazing! lol cant wait to make a mini cloudron cluster!

[malvim]

I'd be very interested in a multi-node cloudron setup. I have some stuff I'd be able to host at home, and other stuff I'd like to run on a VPS or physical server somewhere.
Also, there's software that needs tons of storage (like nextcloud), and others that need a bunch of memory but very little storage, so I'm currently renting a server with lots of storage and memory, which could easily be switched to a cheaper dual-node setup, I think.
So yeah, eagerly awaiting that 6.0 release.

[girish]

@malvim said in 3 Node Deployment (Unified Dashboard):

have some stuff I'd be able to host at home, and other stuff I'd like to run on a VPS or physical server somewhere.

Indeed, this is exactly the use case we are looking to support. I have a similar situation where my primary cloudron is hosted at home and my mail server is on a VPS and I have another server for test deployments. I would like all this to be unified in some way.

I bring this up because I have also been looking into solutions like k8s/swarm/nomad etc but they seem more suited for horizontal scaling whereas what I think we want is just a way to manage multiple cloudron installations.

[marcusquinn]

@girish TBH k8s/swarm/nomad all have latency issues to contend with that need specific network setups unless the Cloud provider already offers K8S with a dedicated network for the nodes. Been down this road and there's usually significant performance compromises as soon as anything needs to use a database.

What I'd be more interested would be hot-swappable Master-Slave(s) setups.

So you'd have your main hardware serving everything, and use Unison for file replication and whatever database Cluster versions for the Master-Slave replication.

Some decent guides on this for Wordpress:

• https://www.wpintense.com/category/performance/stack-guides/

Then you have a hot-swap timeout (say 5 minutes or whatever is enough for normal reboots with checks & fixes).

[malvim]

@girish said in 3 Node Deployment (Unified Dashboard):

nomad

Yeah, I think in this case a way to manage different cloudrons in a centralized way would be the way to go.

For me, the main objective would be to install an app from my admin interface and have it go up in a specific machine, with domains and certificates working accordingly and such.

Haven't been able to go on with the raspberry pi setup, as my provider only has NAS disks, and overlay doesn't support it, and the other storage drivers were such a pain to setup I couldn't get it to work.

Looking for a provider with a better setup (and closer to what one would have at home) to try.

Edit: I'm sorry for the raspberry pi talk, I got the threads mixed up. But yeah, my "dream" multi-node setup is basically being able to install apps on different machines from the same web/cli interface.

[robi]

This appears doable by creating a new App that orchestrates 2 or more cloudron APIs in one interface.

It could be a mobile/web admin app for any number of cloudrons with features like adding apps to each, user management across all, monitoring, etc.

[plusone-nick]

@robi I think they are open to design/feature requests: https://forum.cloudron.io/topic/3205/what-s-coming-in-6-0-take-2/2

[marcusquinn]

@plusone-nick Just looking at this: https://www.scalecomputing.com Do you have any experience with them? Any recommendations for it one wanted to make a multi-node, multi-site setup with failover redundancy?

[plusone-nick]

@marcusquinn have seen them but no hands on exp unfortunately - kept more of an eye on https://www.mirantis.com/ basically open-stack/kubernetes

but then gain that's basically just https://www.proxmox.com/en/ & Kubernetes
or https://opennebula.io/ & Kubernetes
lol

[marcusquinn]

@plusone-nick Yeah, that's application & storage layer but the reason for looking at Scale Computing is OS layer, so it covers server & client OS images. Kinda like your own private hosting provider in a box (or boxes).