Docker container consumed large number of inodes

[dantefromhell]

Which resulted in /run (a TMPFS on CentOS 7) running out of empty inodes.

Troubleshooting showed

$ { find /run -xdev -printf '%h\n' | sort | uniq -c | sort -k 1 -n; } 2>/dev/null
[... SNIP ...]
 498772 /run/docker/libcontainerd/19e9fe6cf800b0089d54064ace641a3ce4d56d8381dec570030896c58c6e9e41

The beginning of the hash value overlaps with the container ID column of docker ps -a output.

[dantefromhell]

Restarting the container released the inodes.
Previous runtime was ~6 months.