How can I delete images without labels (label `<none>`) using podman?

Question

When I run podman images, I see a bunch of images like this,

$ podman images
REPOSITORY                 TAG        IMAGE ID      CREATED       SIZE
localhost/tcms             latest     e6d31cbc5385  10 hours ago  86.4 MB
<none>                     <none>     8f773eba96cd  10 hours ago  86.4 MB
<none>                     <none>     3e335f6f49fe  13 hours ago  93.9 MB
<none>                     <none>     6e41c1796a0f  13 hours ago  93.9 MB
<none>                     <none>     bdde160250db  13 hours ago  93.3 MB
<none>                     <none>     d323538cae8f  13 hours ago  93.3 MB

How can I remove those images where the label is <none> en masse? I don't see command that does this under podman images --help

Answer 1

However insane the design decision

• podman images
• podman image

are different commands, you'll want to use

podman image prune

Which will remove unused images with a label of <NONE>. If you add the -a flag it will remove all images, even those with a label, if they're unused by a container.

Answer 2

The prune command didn't work for me.

What worked was to delete the 'ghost' images using the Id.

No need to specify the full id, just the first 2 chars are enough

$ podman image ls -a | grep '<none>'

<none> 75108329e92d 10 minutes ago 782 MB

Then the first image can be deleted with:

podman rmi 75

It will be faster to start deleting the first image from top to bottom. Then continue with the remaining images.

Answer 3

You can use the following command

sudo podman image ls -a | grep '<none>' | awk '{print $3}' | xargs sudo podman rmi -f

to find your images with label \<none>, extract the ID (the awk part) and pass each ID to podman rmi (-f means force).

Answer 4

I am late to the party, but here's another way to delete images without tags en masse:

podman rmi $(podman images -f "dangling=true" -q)

You'll get a warning if the images are in use. You can use podman rmi --force, but it's better check them first.