It turns out someone already made something like this! Check out https://github.com/povilasv/systemd_exporter.
Proof of concept systemd monitor daemon. It is modeled as a cross between https://github.com/prometheus/node_exporter and https://github.com/google/cadvisor. In kubernetes, I usually just want to monitor Kubelet, docker, the Node Problem Detector, journald, and maybe a few others. cAdvisor is hard to configure, since everything is in terms of cgroups.
make docker
SHA=<sha>
docker run -p 8080:8080 --volume=/run/systemd:/run/systemd:ro --volume=/sys/fs/cgroup:/sys/fs/cgroup:ro $SHA
curl localhost:8080/metrics
make docker
# Tag and push your image
# Edit deploy/kubernetes/daemonset.yaml to specify your pushed image
kubectl apply -f deploy/kubernetes/daemonset.yaml
kubectl get po
# pick a systemd-exporter pod to query
POD=systemd-exporter-12345
kubectl get --raw /api/v1/namespaces/default/pods/$POD/proxy/metrics
"authentication protocol error" Systemd 239 has a regression: systemd/systemd#9553. Use a different versoin of systemd.