Running BalenaOS in cloud

Is there any support for running BalenaOS in either AWS or Azure?
We are seeing some use cases where a physical device is not optimal but we wish to keep the same fleet management over all deployments.

3 Likes

Hi there,

Thanks for reaching out. While we don’t provide official support for that use case, internally we have run BalenaOS on AWS and DigitalOcean successfully. The generic x86 64-bit image is what we used.

It’s an interesting use case we’re exploring. We’d love to hear how you’d like to make use of it.

1 Like

Hi, as a follow up and in case you want to experiment with this let me point you to how we generate the AMI images for AWS. balena-yocto-scripts/jenkins_generate_ami.sh at master · balena-os/balena-yocto-scripts · GitHub

I’m jumping back into an older thread but we’ve also run across this use case recently. We’re exploring using balenaOS devices to distribute content (linear and non-linear) to remote customers. We know some of the sites would prefer a “hosted” solution where they don’t have hardware on site. We’d like to run the same balenaOS device in a cloud environment as a “hosted” solution for the remote customer but have it integrate into the same monitoring and operations systems we have for remote physical devices (i.e. balenaCloud). The customer would then pull content from the device via SFTP or web streams (e.g. Icecast or HLS).

We’ve successfully done this in our on-prem datacenter using VMWare VMs but we’d like to move it to the cloud. We selected DigitalOcean as a provider because the bandwidth costs are a fraction of AWS (e.g. about 9x cheaper). From what I can tell, DO requires custom images to use cloud-init. I read through your scripts for generating the AMI but it doesn’t seem to modify the OS image to add or configure cloud-init. Maybe I’m just not seeing it.

You mentioned that you had balenaOS running on DigitalOcean at one point. Do you have any scripts you can share or any pointers for how to configure the custom image? I’m aware there’s a balena “virt” project that uses qemu to run balenaOS as a VM on a DO instance, but I’d love to be able to boot the OS directly as a VM without the extra virtualization layer.

Thanks.