Comment 1 for bug 1888729

I found this was occurring because the etcd/2 unit had re-rendered its etcd.conf.yml file, omitting the `cluster:` field, before etcd had completed registering.

Workaround:

juju run --unit etcd/2 -- charms.reactive clear_flag etcd.registered
juju run --unit etcd/2 -- hooks/update-status