Reference docker bind mount nginx gotcha

When an nginx config is bind-mounted into a Docker container by file path (not by directory), deploying a new version via tar extraction replaces the file by inode swap. The running nginx daemon continues pointing at the orphaned old inode and does not pick up the new config on nginx -s reload or equivalent.

Fix: restart the container after deploying a new nginx config file in this setup (docker restart <container>).

Why: Docker bind-mounts track the inode, not the filename. Tar replaces the file at the filesystem level but does not preserve the inode, so the mount pointer becomes stale.

How to apply: Any time a deploy script ships an nginx config to a container that uses a file-level bind mount (e.g., justinandkrystal-landing on media-server CT 102), follow up with a container restart, not just a config reload.

[Claude Code]

[auto-memory session e0097de0-7e40-40e5-8991-c9d652aef72a, confidence 0.75, mode staged]