Unicorn/Nginx process missing, socket open

Question

I am trying to deploy code using Capistrano, and it fails on deploy:start or deploy:stop because the Unicorn process is already killed. However if I try to cap deploy:start, I get a stderr claiming that Address already in use - /tmp/my_app.socket. How would this happen, and how might I get out of this mess?

Answer 1

Still not sure how this happens, but the following solution seems to work:

lsof /tmp/my_app.socket - lists the pids

kill -9 pid - (replace 'pid' with one of those listed)

Then cap deploy:start from the local terminal.