Comment 10 for bug 1580728

This started spiking in failures in the last 24 hours once oslo.log 3.26.0 was used in upper-constraints on master, and it's failing in the journal handler:

https://github.com/openstack/oslo.log/blob/3.26.0/oslo_log/handlers.py#L120

The switch in devstack to use systemd by default happened yesterday too:

https://github.com/openstack-dev/devstack/commit/7f8df450dbdfe754eb3f7311de56327878a3d1c3

So everything is running with that handler now and we're seeing the spike in failures:

http://logstash.openstack.org/#dashboard/file/logstash.json?query=message%3A%5C%22Setting%20instance%20vm_state%20to%20ERROR%5C%22%20AND%20message%3A%5C%22_unplug_vifs%5C%22%20AND%20message%3A%5C%22UnicodeDecodeError%5C%22%20AND%20tags%3A%5C%22screen-n-cpu.txt%5C%22&from=7d

It's failing specifically hard in the cells v1 job, I'm not really sure why.