I’ve run into this a few times before - if a prior deploy failed, Ansible will be unable to remove it (the release folder) during a subsequent deploy.
Looking at the path referenced, I’d guess your deploy on 12/09/2017 @ 22:27:49 didn’t complete, but the folder is still there.
SSH into the machine, and run rm -rf /srv/www/mysite.com/releases/20171209222749/ - you should also run rm -rf /srv/www/mysite.com/releases/20171212123959/ since the deploy failed on that too, so eventually you’ll run into the same issue when Trellis is trying to remove this release.