Connection refused while nginx connects to puma

Question

Using Ruby 2.2.4, Rails 4.2.5, Puma 2.15.3.

App is running fine, but after several weeks, I get Rails' "Something went wrong" error. Upon checking nginx log, here's what I got:

2016/12/21 19:44:08 [error] 1387#0: *1219594 connect() to unix:///home/deployer/apps/abc/shared/tmp/sockets/abc-puma.sock failed (111: Connection refused) while connecting to upstream, client: 180.76.15.135, server: app.abc.my, request: "GET / HTTP/1.1", upstream: "http://unix:///home/deployer/apps/abc/shared/tmp/sockets/abc-puma.sock:/", host: "app.abc.my"
2016/12/21 19:44:08 [info] 1387#0: *1219594 recv() failed (104: Connection reset by peer) while sending response to client, client: 180.76.15.135, server: app.abc.my, request: "GET / HTTP/1.1", upstream: "http://unix:///home/deployer/apps/abc/shared/tmp/sockets/abc-puma.sock/", host: "app.abc.my"

The only way I had to restart this is

1. Stop Puma
2. Delete abc-puma.sock
3. Start Puma

Then it worked.

I checked my nginx and puma processes:

deployer@vps1497:~$ ps aux | grep nginx
deployer  1146  0.0  0.0   6456   836 pts/1    S+   16:14   0:00 grep nginx
root      1386  0.0  0.0  31240   388 ?        Ss   Dec15   0:00 nginx: master process /usr/sbin/nginx
www-data  1387  0.0  0.0  32416  2344 ?        S    Dec15   5:53 nginx: worker process
www-data  1389  0.0  0.0  32384  2288 ?        S    Dec15   5:54 nginx: worker process
www-data  1390  0.0  0.0  32420  2280 ?        S    Dec15   5:56 nginx: worker process
www-data  1391  0.0  0.0  32380  2284 ?        S    Dec15   6:04 nginx: worker process
deployer@vps1497:/etc/nginx/sites-available$ ps aux | grep puma
deployer   791  0.0  2.2 749108 93724 ?        Sl   15:54   0:01 puma 2.15.3 (unix:///home/deployer/apps/rra/shared/tmp/sockets/abc-puma.sock) [20160710155205]                   
deployer  1340  0.0  0.0   6456   840 pts/1    S+   16:25   0:00 grep puma

What went wrong? I have 4GB RAM in the server, and I could only restart the app by first deleting the puma socket.