Comment 3 for bug 1993319

With my change above, and using the loki-k8s charm, this appears to be resolved. Confirmation would be helpful.

I deployed the loki-k8s charm:

$ juju deploy loki-k8s --resource loki-image=2

Revision 2 was used for the Loki-image because it had a different SHA from the most recent resource revision.

$ microk8s kubectl get pod loki-k8s-0 -n two -o json | grep "loki-image"
                "image": "registry.jujucharms.com/charm/ayt4gaga2x20gye32qthgygevco13qgq6rfff/loki-image@sha256:4e6dca46091ac8e08f8e06e734d657ee6c0000f1727050fd70ca59af7086841d",

$ juju refresh loki-k8s
charm "loki-k8s": already up-to-date
resources to be upgraded
Leaving endpoints in "alpha": alertmanager, grafana-dashboard, grafana-source, ingress, logging, metrics-endpoint

Use `juju show-status-log loki-k8s-0` to see the upgrade-charm hook after `juju refresh` run.

$ microk8s kubectl get pod loki-k8s-0 -n two -o json | grep "loki-image"
                "image": "registry.jujucharms.com/charm/ayt4gaga2x20gye32qthgygevco13qgq6rfff/loki-image@sha256:b61a65dbf591ccebebf3874c8d5e37d4ba1a8d2017625da97a83a4ced2ff252e",

The image has a new SHA.