blog.ndumas.com/content/posts/digging-through-docker-volumes/index.md

draft	title	date	series	author	cover	tags	summary	showFullContent
false	How to find that one volume you're pretty sure you didn't lose	2024-06-25		Nick Dumas		bash docker	Docker volumes can be opaque, so I wrote a small bash script to help you troubleshoot.	false

What I expect you to know

This article is only relevant if you know about and use Docker volumes and have some fluency in bash. I'll explain the code as I go, if it helps.

The Problem

Over the lifetime of a Docker host machine, it's like that orphaned volumes ( and other detritus ) will accumulate over time. You might also find yourself fumbling a configuration and orphaning a volume yourself.

However we got here, we have a bunch of volumes and we need to know if any of them are important. In a perfect world, they'll have decent names. We don't live in a perfect world.

Make a list

Luckily, we have tools at our disposal to handle this. My thought process almost always starts with "Can I turn a list of the things I care about into a newline separated list?" If I can do that, I can start automating my troubleshooting.

Let's start with docker volume ls. This is how we list volumes, but the default output isn't quite what I'm looking for:

docker volume ls
DRIVER    VOLUME NAME
local     d35fce052fbce42b94b2f9b2957be0f77090fa006b1a192030eff07db3675af2
local     grafana-storage
local     plausible_db-data
local     plausible_event-data

This is human readable, and we could even do some slicing with cut or awk, but Docker gives us a flag that will take us exactly where we need to go: --format. Generally, Docker uses Go's text/template library to back this feature, and more specifically, individual flags (usually) document the template verbs available. Here, we want Name.

docker volume ls --format "{{.Name}}"
d35fce052fbce42b94b2f9b2957be0f77090fa006b1a192030eff07db3675af2
grafana-storage
plausible_db-data
plausible_event-data

And now we have a newline separated list of volume names.

Process of elimination

The next part is fairly straightforward. We loop over this list and ask Docker to create a temporary container based on alpine, with a single volume mounted at /test/.

#! /bin/bash

# Newline separated list of volume names
volumes=$(docker volume ls --format="{{.Name}}")
for volume in $volumes; do
	# Help the user keep track of which volume they're exploring
	echo "Mounting $volume in a temporary image."
    docker run --rm -v $volume:/test/ -it alpine /bin/sh
done

Running this script should do something like this:

./cycle-volumes.sh
Mounting d35fce052fbce42b94b2f9b2957be0f77090fa006b1a192030eff07db3675af2 in a temporary image.
/ # ls /test/
clickhouse-server.err.log   clickhouse-server.log.1.gz  clickhouse-server.log.4.gz  clickhouse-server.log.7.gz
clickhouse-server.log       clickhouse-server.log.2.gz  clickhouse-server.log.5.gz  clickhouse-server.log.8.gz
clickhouse-server.log.0.gz  clickhouse-server.log.3.gz  clickhouse-server.log.6.gz
/ #
Mounting grafana-storage in a temporary image.
/ # ls /test
alerting          csv               file-collections  grafana.db        plugins           png
/ # exit
Mounting plausible_db-data in a temporary image.
/ # exit
Mounting plausible_event-data in a temporary image.
/ # exit

You can use bash to explore the volume and identify its contents, make note of which ones are which, and proceed accordingly.