Backup and restore docker named volume

Question

I have question regarding volumes and ownership.

As an example I'm using this image: privatebin, but this is the same for any case.

First I'm creating volume:

$ docker volume create privatebin-data

From docker inspect, I can see where data is located:

$ docker inspect privatebin-data 
[
    {
        "CreatedAt": "2018-12-04T21:42:46+01:00",
        "Driver": "local",
        "Labels": {},
        "Mountpoint": "/var/lib/docker/volumes/privatebin-data/_data",
        "Name": "privatebin-data",
        "Options": {},
        "Scope": "local"
    }
]

Following instructions from docker hub, I'm starting image:

$ docker run -d --restart="always" --read-only -p 8080:80 -v privatebin-data:/srv/data privatebin/nginx-fpm-alpine:1.1.1

Then I visit http://localhost:8080 and everything is working as expected.

Content of volume now:

$ ls -l /var/lib/docker/volumes/privatebin-data/_data
total 16
drwx------ 3 82 82 4096 Dec  4 21:49 73
-rw-r----- 1 82 82   46 Dec  4 21:49 purge_limiter.php
-rw-r----- 1 82 82  529 Dec  4 21:49 salt.php
-rw-r----- 1 82 82  131 Dec  4 21:49 traffic_limiter.php

I want to backup directory by archiving it:

tar -C /var/lib/docker/volumes/privatebin-data -czf privatebin-data-backup.tar.gz _data

My question is: Can I safely assume that if I restart image, for example on other server, user and group owner will still be 82? Is this proper way of backuping and restoring docker volumes?

score 14 · Accepted Answer · answered Dec 04 '18 at 22:49

14

The UID/GID come from inside your image, privatebin/nginx-fpm-alpine. So as long as you create users in the same way/order in there, and nothing changes in your base image, then those ID's will be the same regardless of where you run the image.

My preferred way to backup and restore volumes is to use a utility container, just in case the backend of docker changes, or you decide to move your named volume to another location or external data store. The commands to do that look like:

docker run --rm \
  -v privatebin-data:/source:ro \
  busybox tar -czC /source . >privatebin-data-backup.tar.gz

and

docker run --rm -i \
  -v privatebin-data:/target \
  busybox tar -xzC /target <privatebin-data-backup.tar.gz

answered Dec 04 '18 at 22:49

BMitch

231,797
42
475
450

For some reason I thought that 82:82 was somehow random. But after your reply I dug up that this is not random at all. So that clarifies it. As for second part, quick followup question: is there any adventage of taking data from inside container over directly from host? What's the difference? – Robert Maguda Dec 05 '18 at 06:42
@RobertMaguda Using a utility container allows the named volume to change its source without changing the backup command. You can mount more than just volumes from /var/lib/docker using the local driver, and there are other drivers that let you use 3rd party storage systems. – BMitch Dec 05 '18 at 13:32
@BMitch Could you add some explanation to the backup and restore `docker run` commands please? They seem to be the commands I am looking for, but I couldn't understand completely as I lack busybox / tar related knowledge. – AD8 Jun 10 '20 at 14:48
1

@AD8 busybox is just a minimal distribution that includes commands like tar. The man page on tar covers most of what you're looking for. Other than that, it's using stdin/stdout from the container to transfer the tar file to/from the container with the volume mounted from/to the host. https://www.man7.org/linux/man-pages/man1/tar.1.html – BMitch Jun 10 '20 at 15:00
@BMitch Perfect, I had no idea what busybox and tar are. I think I have enough information to look further into this, thanks to you. Is there an industry standard for backing up docker volumes by any chance? Official docker docs seem to be suggesting an approach similar to your answer above: https://docs.docker.com/storage/volumes/#backup-restore-or-migrate-data-volumes – AD8 Jun 10 '20 at 15:07
@BMitch I believe a great benefit of using Busybox would be it's smaller footprint perhaps – AD8 Jun 10 '20 at 15:08
1

@AD8 compared to ubuntu at almost 74MB, busybox is a little over 1MB, so I use busybox when I don't need the functionality of a full distribution (e.g. not needing to install additional packages with a package manager). This is as close to a standard as I'm aware of. Other solutions that backup the underlying filesystem directly will fail when you configure named volumes to store data in a different location. – BMitch Jun 10 '20 at 15:15
@BMitch makes perfect sense. Thank you for sharing your knowledge. Stay safe :) – AD8 Jun 10 '20 at 15:17

score 1 · Answer 2 · answered Oct 29 '22 at 17:48

1

You can use docker-volume-snapshot to backup docker named volume in.

After installing the utility, just enter this command

docker-volume-snapshot create privatebin-data privatebin-data-backup.tar

answered Oct 29 '22 at 17:48

Juned Khatri

451
5
8

Backup and restore docker named volume

2 Answers2