Comment 4 for bug 1891076

Per comment #3, an issue which appears to be the same occurred yesterday with juju 2.6.8. Part of upgrading software version running in the model, included adding new subordinate units to long running machines. A few of the new subordinates got stuck in agent allocating.

After looking at additional logs, the problem seen yesterday happens before the download stage.

2020-08-25 18:35:30 ERROR juju.worker.dependency engine.go:663 "unit-agent-deployer" manifold worker returned unexpected error: dbus link request failed for application "jujud-unit-docker-9": Failed to activate service 'org.freedesktop.systemd1': timed out (service_start_timeout=25000ms)
2020-08-25 18:36:01 ERROR juju.service.systemd service.go:148 dbus link request failed for application "jujud-unit-docker-9": Failed to activate service 'org.freedesktop.systemd1': timed out (service_start_timeout=25000ms)
2020-08-25 18:36:01 ERROR juju.service.systemd service.go:460 failed to install service "jujud-unit-docker-9": dbus link request failed for application "jujud-unit-docker-9": Failed to activate service 'org.freedesktop.systemd1': timed out (service_start_timeout=25000ms)

We're trying to link the agent files but it keeps failing on this call:
https://github.com/juju/juju/blob/71ee73f42e5574f14b5c93273adaffb0b29bcc91/service/systemd/service.go#L593

The unit-agent-deployer worker is stuck in a error loop, continually restarting.

We also know that dbus got wedged on the machine. Which may indicate why restarting the jujud services did not resolve the issue but rebooting did.

Interestingly a script was able to unwedge one of the machines this was seen on. Then the unit agent did install.