RFE: Make 'skopeo inspect' provide the size of the OCI image

[debarshiray]

It would be nice if 'skopeo inspect' provided the size of the OCI image. It seems like useful information that one might want to have before pulling an image to avoid blowing the data cap or interfering with other network activity.

eg., it could be useful to implement containers/toolbox#134, but it's a different question whether one wants to add a dependency on another 20+ MB binary just for a bit of UI niceness. :)

[TomSweeneyRedHat]

Thanks @debarshiray for the RFE.

[github-actions]

A friendly reminder that this issue had no activity for 30 days.

[infiniteregrets]

Any progress on this issue?
I think this would be a really nice feature to have! I tried skopeo using the following command:
skopeo inspect --override-os linux docker://registry.fedoraproject.org/fedora:latest

{
    "Name": "registry.fedoraproject.org/fedora",
    "Digest": "sha256:8b01cffca564ca914d5d3c8dc8c6eca12a755ee4d1d898e22e83ad7128fae256",
    "RepoTags": [
        "34-aarch64",
        "34-ppc64le",
        "34-x86_64",
        "34",
        "34-s390x",
        "34-armhfp",
        "33-armhfp",
        "32-armhfp",
        "35-aarch64",
        "35-ppc64le",
        "35-s390x",
        "35-x86_64",
        "35",
        "35-armhfp",
        "30-aarch64",
        "30-ppc64le",
        "30-s390x",
        "30-x86_64",
        "30",
        "latest",
        "rawhide",
        "30-armhfp",
        "31-aarch64",
        "31-x86_64",
        "31",
        "31-armhfp",
        "31-s390x",
        "31-ppc64le",
        "32-aarch64",
        "32-ppc64le",
        "32-s390x",
        "32-x86_64",
        "32",
        "33-aarch64",
        "33-ppc64le",
        "33-s390x",
        "33-x86_64",
        "33"
    ],
    "Created": "2021-06-11T05:48:31Z",
    "DockerVersion": "1.10.1",
    "Labels": {
        "license": "MIT",
        "name": "fedora",
        "vendor": "Fedora Project",
        "version": "34"
    },
    "Architecture": "amd64",
    "Os": "linux",
    "Layers": [
        "sha256:ca594071dfff2c0e563feabb63b1df444a7b76eac5be65103e6c047cec6eee3f"
    ],
    "Env": [
        "DISTTAG=f34container",
        "FGC=f34",
        "container=oci"
    ]
}

This does not render the size...
I tried looking at the source code but kinda got confused
But I think this is definitely possible. I tried to do this using the docker hub API using a script I wrote:

import asyncio
import aiohttp
import json

async def fetch(name=None):
    async with aiohttp.ClientSession() as session:
        async with session.get(f"https://hub.docker.com/v2/repositories/library/{name}/tags") as response:
            if response.status != 200:
                print('`Falied to fetch data | Try again later`')
                exit(1)
            data = await response.text()
            json_data = json.loads(str(data))
            for result in json_data["results"][:10]:
                print(name+':'+result["name"], round(result["full_size"]/(1048576),2))

asyncio.run(fetch(name="fedora"))

Results:

fedora:latest 61.92
fedora:rawhide 62.04
fedora:35 62.04
fedora:34 61.92
fedora:33 59.0
fedora:32 67.84
fedora:31 58.58
fedora:30 67.54
fedora:29 85.81
fedora:28 85.49

I am not completely sure how skopeo is doing it sorry, but having this would be very useful!

[mtrmac]

        async with session.get(f"https://hub.docker.com/v2/repositories/library/{name}/tags") as response:

That’s not a documented Registry API per https://github.com/distribution/distribution/blob/main/docs/spec/api.md . (But with new enough images we can get sizes of individual blobs, and sum that up.)

---

Overall, sure, a reasonable RFE.

Note: Compare also #1309 requesting sizes of individual layers, a rather more invasive output format change.

[infiniteregrets]

@mtrmac are you working on this? If not, could you please point me in the right direction where I will have to make changes?

[mtrmac]

I’m not working on this, PRs would be welcome.

There are two parts to this:

• In containers/image, extend types.Image.Inspect to also provide sizes if known (without breaking API, otherwise should be trivial).
• Design how to output this in skopeo inspect output, without breaking existing users and without repetition in the default format. This is the harder part. One alternative would be to make this data only available to the --format option, though that’s pretty ugly and hard to find. An extra option to output both the existing layer IDs and something with sizes? An extra option to report layer ID+sizes instead of the existing field?

[lsm5]

@infiniteregrets are you working on this?

[infiniteregrets]

@lsm5 I tried working on this, but got stuck so I just didnt try again. I would love to give it another try, but busy with school ): Sorry

[lsm5]

@infiniteregrets thanks and no worries. I'll assign a volunteers wanted but feel free to pick this up whenever you can.

[WarheadsSE]

A simple method of fetching this information (when the data is present) is with skopeo inspect --raw IMAGE | jq '[ .layers[].size ] | add'

Be aware that this will not work on a "mediaType": "application/vnd.docker.distribution.manifest.list.v2+json", such as docker://registry.docker.com/library/alpine:latest. For that, you will need to pull out the image manifest appropriate for you before performing the above query for that image.

$ skopeo inspect --raw docker://registry.docker.com/library/alpine:latest \
  | jq -r '.manifests[] | select(.platform.architecture == "amd64" and .platform.os == "linux") | .digest'
sha256:e7d88de73db3d3fd9b2d63aa7f447a10fd0220b7cbf39803c803f2af9ba256b3
$ skopeo inspect --raw docker://registry.docker.com/library/alpine@sha256:e7d88de73db3d3fd9b2d63aa7f447a10fd0220b7cbf39803c803f2af9ba256b3 \
  | jq -r '[ .layers[].size ] | add'
2818413

[github-actions]

A friendly reminder that this issue had no activity for 30 days.

[rhatdan]

Is the solution by @WarheadsSE Enough to close this?

[mtrmac]

No, why would it be?

containers/image#1608 is a possible first step.

[rhatdan]

I was asking the reporter. If you think it is important to implement then fine.

[mtrmac]

#1738 (see #1738 (comment) for an example) now provides individual layer sizes, if known.

[debarshiray]

Is the solution by @WarheadsSE Enough to close this?

Yeah, the solution by @WarheadsSE might have been enough to improve Toolbx's UI. However, I am delighted that @ningmingxiao made it easier by removing one step because most of the images Toolbx deals with are application/vnd.docker.distribution.manifest.list.v2+json.

[debarshiray]

#1738 (see #1738 (comment) for an example) now provides individual layer sizes, if known.

That's wonderful! I suppose this issue can now be closed, yes.

[mtrmac]

Thanks! Note that the individual layer size can be -1 if unknown (schema1 images)