Coming from Harvester and wanting to use proxmox again for this, I jumped for joy when I discovered this project. However, I feel there is a pretty massive gap that is frustrating for my use case.
In other cluster node drivers, like Harvester, there is the ability to supplement the cloud-init user-data bits. It is not possible with this provider, and instead we are required to manage templates that are fully built and customized. I get that it makes it easier on the development of this project, however prebuilt cloud-images already exist, and that eliminates the need to create those ourselves. Adding user-data on top of that allows us to really customize the end result of the booted VM.
Here is a screenshot from the cluster pool creation advanced settings showing the user-data bits as an example:
With that said, qemu-guest-agent is a requirement of this node-driver to work, so:
1.) Why would it not be included in the generated cloud-init user-data that is deployed with the new VMs,
or:
2.) Can it be developed in a way to allow us to include supplemental user-data config so that we don't have to build our own templates from scratch and can instead supplement the cloud-images that already exist?
Option 2 is obviously ideal because I need more than just the qemu-guest-agent package. (nfs-common to use nfs in the generated clusters for shared data access).
My anticipated workflow with this would've been to use the proxmox and rancher terraform providers to build the cluster(s), by creating a template with the desired cloud-image from something like debian, which I currently use today with the Harvester clusters I built. The template would never need to be booted; just being created with the imported cloud image would suffice, and can then be turned around into the rancher node driver to build out the cluster as I do today with Harvester. The missing piece is the user-data supplementation that I cannot add. Otherwise it would be a fairly elegant solution.
Is something like this in the roadmap? Thanks.
Coming from Harvester and wanting to use proxmox again for this, I jumped for joy when I discovered this project. However, I feel there is a pretty massive gap that is frustrating for my use case.
In other cluster node drivers, like Harvester, there is the ability to supplement the cloud-init user-data bits. It is not possible with this provider, and instead we are required to manage templates that are fully built and customized. I get that it makes it easier on the development of this project, however prebuilt cloud-images already exist, and that eliminates the need to create those ourselves. Adding user-data on top of that allows us to really customize the end result of the booted VM.
Here is a screenshot from the cluster pool creation advanced settings showing the user-data bits as an example:
With that said, qemu-guest-agent is a requirement of this node-driver to work, so:
1.) Why would it not be included in the generated cloud-init user-data that is deployed with the new VMs,
or:
2.) Can it be developed in a way to allow us to include supplemental user-data config so that we don't have to build our own templates from scratch and can instead supplement the cloud-images that already exist?
Option 2 is obviously ideal because I need more than just the qemu-guest-agent package. (nfs-common to use nfs in the generated clusters for shared data access).
My anticipated workflow with this would've been to use the proxmox and rancher terraform providers to build the cluster(s), by creating a template with the desired cloud-image from something like debian, which I currently use today with the Harvester clusters I built. The template would never need to be booted; just being created with the imported cloud image would suffice, and can then be turned around into the rancher node driver to build out the cluster as I do today with Harvester. The missing piece is the user-data supplementation that I cannot add. Otherwise it would be a fairly elegant solution.
Is something like this in the roadmap? Thanks.