Incus is so nice. You can really tell that it's being built by people who give a shit about UX and doing things right.
Case in point, I created and booted a VM, then started adding host bind mounts to it.
First of all the UX is the same for containers and VMs, you just tell it you want a disk with a particular source/destination, and Incus figures out internally whether to materialize that as bind mounts, virtio-fs mounts, 9p mounts, whatever.
But also as I was creating the new disks, I assumed that I'd have to powercycle the VM to have them show up. But... nope, as I created each disk it immediately appeared inside the VM. Incus went and twiddled the configuration of the running qemu instance to add the new devices, and then the Incus agent inside the VM noticed the new devices, saw that they came with a target mount path, and just went ahead and mounted them.
The VM base images are similarly well wired. For example, the Incus agent correctly orders itself to run before local-fs.target is reached, so that it can make sure all the VM's configured storage is properly mounted prior to letting most "normal" services start up, so you don't have any boot-time races on mounts being available.
It's a bunch of little "obvious" things, but they just keep adding up. I've clearly been conditioned to expect stuff to be janky and full of sharp edges, and therefore Incus keeps surprising me by being completely reasonable and useful. ❤️
Case in point, I created and booted a VM, then started adding host bind mounts to it.
First of all the UX is the same for containers and VMs, you just tell it you want a disk with a particular source/destination, and Incus figures out internally whether to materialize that as bind mounts, virtio-fs mounts, 9p mounts, whatever.
But also as I was creating the new disks, I assumed that I'd have to powercycle the VM to have them show up. But... nope, as I created each disk it immediately appeared inside the VM. Incus went and twiddled the configuration of the running qemu instance to add the new devices, and then the Incus agent inside the VM noticed the new devices, saw that they came with a target mount path, and just went ahead and mounted them.
The VM base images are similarly well wired. For example, the Incus agent correctly orders itself to run before local-fs.target is reached, so that it can make sure all the VM's configured storage is properly mounted prior to letting most "normal" services start up, so you don't have any boot-time races on mounts being available.
It's a bunch of little "obvious" things, but they just keep adding up. I've clearly been conditioned to expect stuff to be janky and full of sharp edges, and therefore Incus keeps surprising me by being completely reasonable and useful. ❤️
1
