Extend the Podman machine to test and demo this project.
dnf install podman-machine
flatpak install io.podman_desktop.PodmanDesktoppodman machine init --now
podman machine os apply --restart 'quay.io/spotlesstofu/podman-cri:5.1'Alternatively, build and push the Containerfile yourself and use that.
A vast selection of hacks allows to get things going, until we have a Podman desktop extension to sort things out.
Start the machine:
podman machine start
Copy the binary into the machine:
cat target/debug/podman-cri | podman machine ssh --username root "cat > podman-cri"
Enter the machine:
podman machine ssh --username root
Replace the Podman socket:
mv /run/user/1000/podman/podman.sock /run/user/1000/podman/podman2.sock
Start podman-cri, make it listen on the Podman socket:
PODMAN_ENDPOINT=/run/podman/podman.sock \
PODMAN_CRI_ENDPOINT=/run/user/1000/podman/podman.sock \
./podman-cri
chown core /run/user/1000/podman/podman.sock
Back to your host, ping podman-cri on the Podman socket. You should get a 200 OK:
curl -I --unix-socket /run/user/1000/podman/podman-machine-default-api.sock http://example.com/cri/_ping
Now you can start Podman desktop, its requests will go through podman-cri.
To force the AI extension to use this setup, change the permissions for the Podman desktop flatpak. Replace xdg-run/podman:create with xdg-run/podman/podman-machine-default-api.sock:create. You may use the Flatseal app to do this change.
Run the cloud API adaptor (CAA) inside the machine:
podman run -ti --rm \
--entrypoint /usr/local/bin/cloud-api-adaptor \
--env-file caa.env \
-v /run/peerpods:/run/peerpods \
quay.io/confidential-containers/cloud-api-adaptor:v0.8.2-amd64 azure \
-disable-cvm \
-subscriptionid "${AZURE_SUBSCRIPTION_ID}" \
-region "${AZURE_REGION}" \
-instance-size "${AZURE_INSTANCE_SIZE}" \
-resourcegroup "${AZURE_RESOURCE_GROUP}" \
-vxlan-port 8472 \
-subnetid "${AZURE_SUBNET_ID}" \
-securitygroupid "${AZURE_NSG_ID}" \
-imageid "${AZURE_IMAGE_ID}" \
-disable-cvm