chadoe / docker-cleanup-volumes Goto Github PK

Getting error when running against Docker 1.8.3, due to Cannot connect to the Docker daemon when calling docker info:

docker-cleanup-volumes nritholtz$ docker run -it -v /var/run/docker.sock:/var/run/docker.sock -v /var/lib/docker:/var/lib/docker --entrypoint=bash --rm martin/docker-cleanup-volumes
bash-4.3# docker info
FATA[0000] Cannot connect to the Docker daemon. Is 'docker -d' running on this host? 
bash-4.3# ls /var/run/
docker.sock
bash-4.3# ls /var/lib/docker/
aufs               graph              linkgraph.db       tmp                vfs
containers         init               repositories-aufs  trust              volumes
bash-4.3# exit
exit

docker-cleanup-volumes nritholtz$ docker info
Containers: 0
Images: 280
Storage Driver: aufs
 Root Dir: /mnt/sda1/var/lib/docker/aufs
 Backing Filesystem: extfs
 Dirs: 280
 Dirperm1 Supported: true
Execution Driver: native-0.2
Logging Driver: json-file
Kernel Version: 4.1.10-boot2docker
Operating System: Boot2Docker 1.8.3 (TCL 6.4); master : af8b089 - Mon Oct 12 18:56:54 UTC 2015
CPUs: 4
Total Memory: 1.955 GiB
Name: default
ID: IXJO:JTHJ:ZABX:BB3M:ZC2L:6TQR:H7QO:MLW2:YKDX:5IQX:SYAW:KQPW
Debug mode (server): true
File Descriptors: 12
Goroutines: 18
System Time: 2015-10-13T18:58:42.28216136Z
EventsListeners: 0
Init SHA1: 
Init Path: /usr/local/bin/docker
Docker Root Dir: /mnt/sda1/var/lib/docker
Username: nritholtz
Registry: https://index.docker.io/v1/
Labels:
 provider=virtualbox

First off, thanks for this handy script!

I ran into an issue because my build server had ~85k orphaned volumes (image building server from late 2014 :) ) where the ls -d ${targetdir}/* command was failing.

I was able to replace it with find ${targetdir} -maxdepth 1 -type d and continue on my merry way, however I don't have the time or criteria to test and ensure that works in all cases, so I figured an issue was better than a PR.

Thanks again and hope this helps!

regarding:
https://github.com/chadoe/docker-cleanup-volumes/blob/master/docker-cleanup-volumes.sh#L62

Maybe this should just be an echo warning in case of access-denied errors.

Ran the docker and this just went through and deleted all volumes period even on running containers, now I get to rebuild them. Joy is me.

docker@boot2docker:$ sh docker-cleanup-volumes.sh
docker-cleanup-volumes.sh: line 10: syntax error: unexpected "("
docker@boot2docker:$

When you say, in the precautions:

Don't use with boot2docker, docker machine or any tools that use a virtual machine to run docker, unless you run the script on the virtual machine itself (meaning don't run this script or docker command from your MAC).

I'm using Docker 1.9 on OSX with docker-machine, is it safe for me to use the command provided docker volume rm $(docker volume ls -qf dangling=true) from my host? I did it from within the VM as you suggested, but I'm wondering if the warning is just for running your script on versions < 1.9?

just one volume directory:

root@boot2docker:/mnt/sda1/var/lib/docker/volumes# ls -la
total 12
drwx------    3 root     root          4096 Jun 24 21:21 .
drwxr-xr-x    9 root     root          4096 Jun 24 21:24 ..
drwxr-xr-x    3 root     root          4096 Jun 24 21:21 bd1e586a629637ffb1153847260e63f810c43d7f9f28f382eecacf9f4a5eb749

and as you can see it IS used by a running container:

✗ docker inspect --format '{{.Volumes}}' database                                                                                                                                                                                                                                                               ruby-2.2.0 (master) deedf83
map[/var/lib/postgresql/data:/mnt/sda1/var/lib/docker/volumes/bd1e586a629637ffb1153847260e63f810c43d7f9f28f382eecacf9f4a5eb749/_data]

now running the script:

✗ docker run -v /var/run/docker.sock:/var/run/docker.sock -v /var/lib/docker:/var/lib/docker --rm martin/docker-cleanup-volumes --dry-run                                                                                                                                                                       ruby-2.2.0 (master) deedf83

Delete unused volume directories from /var/lib/docker/volumes
Would have deleted bd1e586a629637ffb1153847260e63f810c43d7f9f28f382eecacf9f4a5eb749

Directory /var/lib/docker/vfs/dir does not exist, skipping.

If you configured an alternative dockerdir (default: /var/lib/docker) the script doesn't work. It would be nice to get this variable configurable.

From a code inspection it seems to me that this will delete bind-mount volumes from /var/lib/docker/volumes, even if they are in use.

For bind-mount volumes (that is, when you do docker run -v /hostpath:/containerpath), docker inspect prints something like this:

"Volumes": {
    "/containerpath": "/hostpath"
},

...so that volume isn't added into allvolumes in your script, but you do delete it from /var/lib/docker/volumes.

I was using https://github.com/spotify/docker-gc, but it deletes volumes associated this containers that were never run such as with Data Only Containers.

Will this script remove them as well or is it safe?

Hello!

I've noticed this script removing running container volumes on centos + docker 1.4.1 (in boot2docker)... using the docker container way to run the script.

What more information can I provide?

Topper

When running from docker - there is ability to use host docker binary if we set it via host volume. For example:

$ docker run -v $(which docker):/bin/docker -v /var/run/docker.sock:/var/run/docker.sock -v /var/lib/docker:/var/lib/docker --rm martin/docker-cleanup-volumes --dry-run

Looks different, but it allows to feel free which docker version is used on host and no need to keep builds for last n docker versions.

I haven't had any issues, but a potentially damaging script like this should default to --dry-run to avoid issues like #19. git clean defaults to following this practice:

$ git clean
fatal: clean.requireForce defaults to true and neither -i, -n nor -f given; refusing to clean
$ git clean -n
Would remove Dockerfile
Would remove docker-compose.yml

Hallo,

Learning docker, I installed docker toolbox on my Mac and follow the docker book from James Turnbull.

$ docker-machine --version
docker-machine version 0.4.0 (9d0dc7a)

Run from my Mac.

$ docker run -v /var/run/docker.sock:/var/run/docker.sock -v /var/lib/docker:/var/lib/docker --rm martin/docker-cleanup-volumes --dry-run --verbose
dockerdir -> /var/lib/docker
dockerdir_match -> /var/lib/docker
Processing volumepath /var/lib/docker for container cf4df59e1a292b37ee52bd68cbe052ba64b9a5ba3759df37fad66bbcf1c50a68
Processing volumepath /var/run/docker.sock for container cf4df59e1a292b37ee52bd68cbe052ba64b9a5ba3759df37fad66bbcf1c50a68
Processing volumepath /Users/serege/thuis/docker/james_blog for container d03b3329be410d6d03f3484eb8b5ee1f2df8749bc60b67b08b1bf0549d71d4df
Processing volumepath /mnt/sda1/var/lib/docker/volumes/967e423a2c907eb4aa0ccbc1e33ffca3f8051859c4198f8247d1a7903a738132/_data for container d03b3329be410d6d03f3484eb8b5ee1f2df8749bc60b67b08b1bf0549d71d4df

Delete unused volume directories from /var/lib/docker/volumes
Would have deleted 967e423a2c907eb4aa0ccbc1e33ffca3f8051859c4198f8247d1a7903a738132

Directory /var/lib/docker/vfs/dir does not exist, skipping.

Why would it delete the volume when is has a volume path? Or did I need to run an other command?

Regards,
Gerrit

On my system, /var/lib/docker is symlinked to another directory on the system. After running docker-cleanup-volumes.sh it removes volumes for my data-only containers that are not currently running (and possibly others and I just didn't notice it). Changing dockerdir in the script to the actual location fixes this for me.

FYI I ran your dockerized script but it kept outputting the below "grep" warning about "recursive directory loop". Probably nothing for you to do as a fix, just posting in case other users see the error.

Deleting the volume mentioned in the message fixed the problem for me -- it was an old volume for jenkins.

Below are my steps. I first ran "docker-machine ssh dev" to access my docker server. It's a virtual box / boot2docker machine on my macbook, continuously upgraded from ~1.6 to 1.9.1.

root@dev:/mnt/sda1/var/lib/docker# docker version
Client:
 Version:      1.9.1
 API version:  1.21
 Go version:   go1.4.3
 Git commit:   a34a1d5
 Built:        Fri Nov 20 17:56:04 UTC 2015
 OS/Arch:      linux/amd64

Server:
 Version:      1.9.1
 API version:  1.21
 Go version:   go1.4.3
 Git commit:   a34a1d5
 Built:        Fri Nov 20 17:56:04 UTC 2015
 OS/Arch:      linux/amd64

root@dev:/mnt/sda1/var/lib/docker# docker run -v /var/run/docker.sock:/var/run/docker.sock -v $(readlink -f /var/lib/docker):/var/lib/docker --rm martin/docker-cleanup-volumes --dry-run
...
grep: warning: /var/lib/docker/volumes/0eb33134d5df1f8c88afd051fe4725c0bc53d8f6e3068ce3f64edd8ca4f4a144/_data/sys/class/power_supply/AC/device/subsystem/devices/LNXVIDEO:00/input/input6/subsystem/event6/device/device/subsystem/devices/0000:00:0d.0/ata1/ata_port/ata1/subsystem/ata5/device/host4/subsystem/devices/host25/scsi_host/host25/subsystem/host0/device/target0:0:0/0:0:0:0/scsi_device/0:0:0:0/subsystem/1:0:0:0/device/block/sda/sda1/subsystem/loop3/bdi/subsystem/7:2/subsystem: recursive directory loop

That does seem like a weird directory:

root@dev:/mnt/sda1/var/lib/docker# ls -l /var/lib/docker/volumes/0eb33134d5df1f8c88afd051fe4725c0bc53d8f6e3068ce3f64edd8ca4f4a144/_data/sys/class/power_supply/AC/device/subsyste
m/devices/
total 80
lrwxrwxrwx    1 root     root            63 Sep  4 04:11 ACPI0003:00 -> ../../../devices/LNXSYSTM:00/LNXSYBUS:00/PNP0A03:00/ACPI0003:00
...

That particular volume appears to be for my Jenkins container (/var/lib/jenkins) from a few months ago. Deleting (rm -rf /var/lib/docker/volumes/0eb33134d5df1f8c88afd051fe4725c0bc53d8f6e3068ce3f64edd8ca4f4a144) it allowed your script to run correctly.