Auto Scalability Feature

[Ravi Kumar]

Would like to find out if Auto Scalability feature for the Virtual Machines is there in the latest version of Proxmox, if a Cluster is formed using Proxmox.

If we want to start something similar to Digital Ocean using Proxmox, is it possible or not

Request you to give the answer or pointers in this direction

 

[Alwin]

There is no auto scaling, this can be implemented through the use of our api.
https://pve.proxmox.com/wiki/Proxmox_VE_API

 

[Dan Nicolae]

There is a plugin for WHMCS for that. I have no ideea if and how it works. If there is someone on this forum that is using it, maybe could share some info.

 

[shantanu]

@Alwin

Having some sort of "autoscaling group" functionality would be a great addition to Proxmox.

Of course, to keep the semantics similar to a "cloud platform", what would be useful is to be able to create a VM/LXC from a remote image url.

I have home-grown scripts which download the image from an http server and then do a "qmrestore" locally.

Are there any plans to integrate this functionality of being able to deploy from an "http(s)" source planned?
(where the download and qmrestore equivalent would be done my PVE itself)

Regards,
Shantanu

 

[fab_space]

With different approaches and hopefully with the help of contributors.. some initial Proof of Concepts

Enjoy and contribute:
- https://github.com/fabriziosalmi/proxmox-lxc-autoscale
- https://github.com/fabriziosalmi/proxmox-vm-autoscale
- https://github.com/fabriziosalmi/proxmox-lxc-autoscale-ml
- https://github.com/fabriziosalmi/proxmox-cluster-balancer

 

[starttoaster]

@fab_space kudos on the interesting tools! I think what would be interesting for a VM autoscaler is actually more similar to AWS ASGs, which support horizontal scaling, provisioning of new virtual machines, and balancing them amongst a PVE datacenter's nodes. Your vertical autoscaler and cluster balancer seem to get a lot of the way towards that end, but not quite.

 

[fab_space]

Thanks for the kudos and for the precise framing, you are right about the gap.

To clarify what exists today: LXC AutoScale ships an experimental horizontal scaling feature (https://fabriziosalmi.github.io/proxmox-lxc-autoscale/guide/horizontal-scaling.html). It works on group-level average CPU/memory: scale-out clones a base container from a snapshot (pct snapshot + pct clone), assigns networking (DHCP or a static IP range) and starts it; scale-in stops the last clone. Grace periods throttle both directions.

Where it falls short of an AWS ASG, exactly as you point out:

• LXC only. The VM autoscaler is vertical only for now, since cloning VMs safely (cloud-init, MAC/IP, guest agent readiness) is a bigger surface.
• Single-node. Clones are created via pct on one host, there is no target-node selection and no integration with the cluster balancer for placement. A proper implementation should go through the PVE API with explicit target selection and shared-storage awareness.
• No full lifecycle. Scale-in stops instead of destroying, and there is no health-check-and-replace loop or load balancer registration.

Merging the placement logic of proxmox-cluster-balancer into the scale-out path is the natural next step, and moving from pct calls to the API would unlock both cross-node placement and VM support.

If you (or anyone reading) want to help shape that, issues and PRs are very welcome, the ASG semantics you describe are a good spec to design against, of course I will try to drop some on my own too.