Context
\nFor context, I'm building a couple of multi-arch images with a db.tar.gz artifact that I'm adding into each of the images. The db.tar.gz artifact is around 60MiB compressed, but over 700MiB once it's been decompressed, so I'm trying to share the decompressed files everywhere I can. There's nothing platform specific about these files either, so it should be fine to only add one cache layer to build cache. The setup looks something like this.
FROM alpine\n\nADD ./db.tar.gz /home/gitlab/.stateUnfortunately, when I build the image I end up with two separate copies of the extracted tar archive totaling at 1.5GB per image for this one ADD instruction.
ID\t\t\t\t\t\tRECLAIMABLE\tSIZE\t\tLAST ACCESSED\nspn6p45bt5dqc7nli19o0ypm4 \ttrue \t\t730.1MB* \t5 seconds ago\n1ilj1w2u763t3ud4d7xis4lji \ttrue \t\t730.1MB* \t5 seconds ago\n8wudq9jlp03gr9pu0fnhfr7zx* \ttrue \t\t65.48MB \t5 seconds ago\nct3ew7tlb0f0jb36tzw30mw89 \ttrue \t\t12.15MB* \t5 seconds ago\nbt0rz9syzlxhvgeox3rswuvvc \ttrue \t\t3.993MB* \t5 seconds ago\ngpbmiwerdds0jdfihfyp35e5m* \ttrue \t\t8.192kB \t5 seconds ago\nxjiw9k64x9afdf4m9qb6u983z* \ttrue \t\t4.096kB \t5 seconds ago\nShared:\t\t1.476GB\nPrivate:\t65.49MB\nReclaimable:\t1.542GB\nTotal:\t\t1.542GB\nInterestingly enough, when I switch to FROM scratch, this behavior no longer exists.
FROM scratch\n\nADD ./db.tar.gz /home/user/.local/db/stateOutput of docker buildx du.
ID\t\t\t\t\t\tRECLAIMABLE\tSIZE\t\tLAST ACCESSED\npie3a3z1pan546kfhjc7escrr \ttrue \t\t730.1MB* \t12 seconds ago\np1wq4dxhco6g0kqio8ot9j9bh* \ttrue \t\t65.48MB \t12 seconds ago\nh8jdp9vp4xuz3hbty44th0fnw* \ttrue \t\t8.192kB \t12 seconds ago\nye9ozzskeim4c39qevy44vddu* \ttrue \t\t4.096kB \t12 seconds ago\nShared:\t\t730.1MB\nPrivate:\t65.49MB\nReclaimable:\t795.5MB\nTotal:\t\t795.5MB\nI'd really like to share this layer across all the images if possible, even if it's only shared in the build cache. I had to upgrade the runner size so that this job could succeed, so if I can achieve this somehow, I'll be able to save on some resources.
","upvoteCount":1,"answerCount":2,"acceptedAnswer":{"@type":"Answer","text":"I think this would work if you use the --link option (see ADD --link and COPY --link).
With that option, the ADD is copied in am implicit FROM scratch layer, which allows the layer to be shared, even if the \"parent\" layers differ. This option should only be used if the location where the content is added does not depend on any parent state (i.e., empty directory, and not expected for that directory to have content that's based on the parent image).
This is a quick try with this, but probably best to try for yourself if I'm correct.
\n👇 worth noting that I'm using docker with the containerd image-store enabled so that I can build multi-platform images with the default builder (not requiring a custom builder);
\n- \n
- https://docs.docker.com/desktop/features/containerd/ \n
- https://docs.docker.com/engine/storage/containerd/ \n
Pulling an image to export as tar (as \"content\");
\ndocker pull docker:latest\nlatest: Pulling from library/docker\nDigest: sha256:f49e1c71b5d9f8ebe53715f78996ce42b8be4b1ec03875d187dfe3c03de1dc00\nStatus: Image is up to date for docker:latest\ndocker.io/library/docker:latestSaving it to a file;
\ndocker image save -o ./content.tar docker:latest\nls -lah content.tar\n-rw------- 1 root root 126.2M Apr 24 23:30 content.tarBuild a multi-arch image with 3 architectures, all adding the same file, using the --link option;
docker build --platform linux/amd64,linux/arm64,linux/s390x -f- .<<'EOF'\n# syntax=docker/dockerfile:1\n\nFROM alpine\nADD --link ./content.tar /home/user/.local/db/state\nEOF[+] Building 7.6s (19/19) FINISHED docker:default\n => [internal] load build definition from Dockerfile 0.0s\n => => transferring dockerfile: 131B 0.0s\n => resolve image config for docker-image://docker.io/docker/dockerfile:1 0.4s\n => CACHED docker-image://docker.io/docker/dockerfile:1@sha256:4c68376a702446fc3c79af22de146a148bc3367e73c25a5803d453b6b3f722fb 0.0s\n => => resolve docker.io/docker/dockerfile:1@sha256:4c68376a702446fc3c79af22de146a148bc3367e73c25a5803d453b6b3f722fb 0.0s\n => [linux/arm64 internal] load metadata for docker.io/library/alpine:latest 2.4s\n => [linux/amd64 internal] load metadata for docker.io/library/alpine:latest 2.5s\n => [linux/s390x internal] load metadata for docker.io/library/alpine:latest 2.5s\n => [internal] load .dockerignore 0.0s\n => => transferring context: 2B 0.0s\n => [internal] load build context 0.3s\n => => transferring context: 132.32MB 0.3s\n => [linux/s390x 1/2] FROM docker.io/library/alpine:latest@sha256:a8560b36e8b8210634f77d9f7f9efd7ffa463e380b75e2e74aff4511df3ef88c 0.6s\n => => resolve docker.io/library/alpine:latest@sha256:a8560b36e8b8210634f77d9f7f9efd7ffa463e380b75e2e74aff4511df3ef88c 0.0s\n => => sha256:c1a599607158512214777614f916f8193d29fd34b656d47dfc26314af01e2af4 3.47MB / 3.47MB 0.5s\n => [linux/arm64 1/2] FROM docker.io/library/alpine:latest@sha256:a8560b36e8b8210634f77d9f7f9efd7ffa463e380b75e2e74aff4511df3ef88c 0.0s\n => => resolve docker.io/library/alpine:latest@sha256:a8560b36e8b8210634f77d9f7f9efd7ffa463e380b75e2e74aff4511df3ef88c 0.0s\n => [linux/amd64 1/2] FROM docker.io/library/alpine:latest@sha256:a8560b36e8b8210634f77d9f7f9efd7ffa463e380b75e2e74aff4511df3ef88c 0.9s\n => => resolve docker.io/library/alpine:latest@sha256:a8560b36e8b8210634f77d9f7f9efd7ffa463e380b75e2e74aff4511df3ef88c 0.0s\n => => sha256:f18232174bc91741fdf3da96d85011092101a032a93a388b79e99e69c2d5c870 3.64MB / 3.64MB 0.9s\n => [linux/s390x 2/2] ADD --link ./content.tar /home/user/.local/db/state 0.1s\n => [linux/arm64 2/2] ADD --link ./content.tar /home/user/.local/db/state 0.0s\n => [linux/amd64 2/2] ADD --link ./content.tar /home/user/.local/db/state 0.0s\n => exporting to image 4.1s\n => => exporting layers 2.8s\n => => exporting manifest sha256:1e3d128e5164ba7501e70ca1173e066346240f3bdc312142a8631a689100fa9b 0.0s\n => => exporting config sha256:d68dadf9c442ddd8bb47bf6d3d61594a5541e87b2a04167f6e912266ba878d26 0.0s\n => => exporting attestation manifest sha256:4391c33e547075b2103e40e9bf8384a17cd45d0d321d0196b29ed199914a66fe 0.0s\n => => exporting manifest sha256:11df05f8103507ec96a0d1974790278635494dc92c337ac96032fb2feff00351 0.0s\n => => exporting config sha256:1f27853335f3c02d0f68ae48fe22ea61425663bb9e6a47d5da4d89902b6925e8 0.0s\n => => exporting attestation manifest sha256:b8d00bb9e19d72b3a82aa09ff7575d7bb65ba3721151bd7e9e5b0609aaeeafbe 0.0s\n => => exporting manifest sha256:44e790ee7e180edb5204b4f84b81d65930166fa7f748ea29ff47579f939a9caa 0.0s\n => => exporting config sha256:47813f37e50c569ba1abd16005a80d991f727b4f2b15d0ff45a34f6bf29a837a 0.0s\n => => exporting attestation manifest sha256:80af267def7cff2c3e5b19a0b6003b1ea82a5d787bc8dc979d1a884c59f0af6a 0.0s\n => => exporting manifest list sha256:26b0b36c895978d0cfa611d790a168d55a7e74b7f62110da7268a4510bd27a4d 0.0s\n => => naming to moby-dangling@sha256:26b0b36c895978d0cfa611d790a168d55a7e74b7f62110da7268a4510bd27a4d 0.0s\n => => unpacking to moby-dangling@sha256:26b0b36c895978d0cfa611d790a168d55a7e74b7f62110da7268a4510bd27a4d 0.4sImage huilt;
\ndocker image ls --tree\nIMAGE ID DISK USAGE CONTENT SIZE EXTRA\n<untagged> 26b0b36c8959 548MB 406MB\n├─ linux/amd64 1e3d128e5164 135MB 135MB\n├─ linux/arm64 11df05f81035 277MB 136MB\n└─ linux/s390x 44e790ee7e18 135MB 135MBdocker buildx du\nID\t\t\t\t\t\tRECLAIMABLE\tSIZE\t\tLAST ACCESSED\nilwlv3up0ko6hf6rbte9ui9wl \ttrue \t\t264.1MB* \t45 seconds ago\nvkj7p0cdulb0od9mhnna4pt6i* \ttrue \t\t132.3MB \t45 seconds ago\nynwtwzuwfepczwuoxx3i2qo28 \ttrue \t\t37.54MB \t45 seconds ago\nw8r4bfgupifuhu68ii2sxirfz \ttrue \t\t3.993MB* \t45 seconds ago\n6cqonmk41d6tl21yt91jpj5go \ttrue \t\t3.642MB* \t45 seconds ago\n2o1taudcfcgfzq7al4ccg9h6y \ttrue \t\t3.468MB* \t45 seconds ago\nxw0sd9sy7a8urox5ghw9tmhiy* \ttrue \t\t8.192kB \t45 seconds ago\nvp8a4ygnqkqzdg5cydonktbzo* \ttrue \t\t4.096kB \t45 seconds ago\nz8ixtkjvpnne6aqnt2ih4fovd \ttrue \t\t0B* \t45 seconds ago\ns4szzze2pnb8py277vweleb4z \ttrue \t\t0B* \t45 seconds ago\n6rt3w0wr1hjvjbq5x7x9ixvsv \ttrue \t\t0B* \t45 seconds ago\nShared:\t\t275.2MB\nPrivate:\t169.9MB\nReclaimable:\t445.1MB\nTotal:\t\t445.1MB
-
Docker infoContextFor context, I'm building a couple of multi-arch images with a FROM alpine
ADD ./db.tar.gz /home/gitlab/.stateUnfortunately, when I build the image I end up with two separate copies of the extracted tar archive totaling at 1.5GB per image for this one Interestingly enough, when I switch to FROM scratch
ADD ./db.tar.gz /home/user/.local/db/stateOutput of I'd really like to share this layer across all the images if possible, even if it's only shared in the build cache. I had to upgrade the runner size so that this job could succeed, so if I can achieve this somehow, I'll be able to save on some resources. |
Beta Was this translation helpful? Give feedback.
Replies: 2 comments 1 reply
-
|
I think this would work if you use the With that option, the This is a quick try with this, but probably best to try for yourself if I'm correct. 👇 worth noting that I'm using docker with the containerd image-store enabled so that I can build multi-platform images with the default builder (not requiring a custom builder);
Pulling an image to export as tar (as "content"); docker pull docker:latest
latest: Pulling from library/docker
Digest: sha256:f49e1c71b5d9f8ebe53715f78996ce42b8be4b1ec03875d187dfe3c03de1dc00
Status: Image is up to date for docker:latest
docker.io/library/docker:latestSaving it to a file; docker image save -o ./content.tar docker:latest
ls -lah content.tar
-rw------- 1 root root 126.2M Apr 24 23:30 content.tarBuild a multi-arch image with 3 architectures, all adding the same file, using the docker build --platform linux/amd64,linux/arm64,linux/s390x -f- .<<'EOF'
# syntax=docker/dockerfile:1
FROM alpine
ADD --link ./content.tar /home/user/.local/db/state
EOF[+] Building 7.6s (19/19) FINISHED docker:default
=> [internal] load build definition from Dockerfile 0.0s
=> => transferring dockerfile: 131B 0.0s
=> resolve image config for docker-image://docker.io/docker/dockerfile:1 0.4s
=> CACHED docker-image://docker.io/docker/dockerfile:1@sha256:4c68376a702446fc3c79af22de146a148bc3367e73c25a5803d453b6b3f722fb 0.0s
=> => resolve docker.io/docker/dockerfile:1@sha256:4c68376a702446fc3c79af22de146a148bc3367e73c25a5803d453b6b3f722fb 0.0s
=> [linux/arm64 internal] load metadata for docker.io/library/alpine:latest 2.4s
=> [linux/amd64 internal] load metadata for docker.io/library/alpine:latest 2.5s
=> [linux/s390x internal] load metadata for docker.io/library/alpine:latest 2.5s
=> [internal] load .dockerignore 0.0s
=> => transferring context: 2B 0.0s
=> [internal] load build context 0.3s
=> => transferring context: 132.32MB 0.3s
=> [linux/s390x 1/2] FROM docker.io/library/alpine:latest@sha256:a8560b36e8b8210634f77d9f7f9efd7ffa463e380b75e2e74aff4511df3ef88c 0.6s
=> => resolve docker.io/library/alpine:latest@sha256:a8560b36e8b8210634f77d9f7f9efd7ffa463e380b75e2e74aff4511df3ef88c 0.0s
=> => sha256:c1a599607158512214777614f916f8193d29fd34b656d47dfc26314af01e2af4 3.47MB / 3.47MB 0.5s
=> [linux/arm64 1/2] FROM docker.io/library/alpine:latest@sha256:a8560b36e8b8210634f77d9f7f9efd7ffa463e380b75e2e74aff4511df3ef88c 0.0s
=> => resolve docker.io/library/alpine:latest@sha256:a8560b36e8b8210634f77d9f7f9efd7ffa463e380b75e2e74aff4511df3ef88c 0.0s
=> [linux/amd64 1/2] FROM docker.io/library/alpine:latest@sha256:a8560b36e8b8210634f77d9f7f9efd7ffa463e380b75e2e74aff4511df3ef88c 0.9s
=> => resolve docker.io/library/alpine:latest@sha256:a8560b36e8b8210634f77d9f7f9efd7ffa463e380b75e2e74aff4511df3ef88c 0.0s
=> => sha256:f18232174bc91741fdf3da96d85011092101a032a93a388b79e99e69c2d5c870 3.64MB / 3.64MB 0.9s
=> [linux/s390x 2/2] ADD --link ./content.tar /home/user/.local/db/state 0.1s
=> [linux/arm64 2/2] ADD --link ./content.tar /home/user/.local/db/state 0.0s
=> [linux/amd64 2/2] ADD --link ./content.tar /home/user/.local/db/state 0.0s
=> exporting to image 4.1s
=> => exporting layers 2.8s
=> => exporting manifest sha256:1e3d128e5164ba7501e70ca1173e066346240f3bdc312142a8631a689100fa9b 0.0s
=> => exporting config sha256:d68dadf9c442ddd8bb47bf6d3d61594a5541e87b2a04167f6e912266ba878d26 0.0s
=> => exporting attestation manifest sha256:4391c33e547075b2103e40e9bf8384a17cd45d0d321d0196b29ed199914a66fe 0.0s
=> => exporting manifest sha256:11df05f8103507ec96a0d1974790278635494dc92c337ac96032fb2feff00351 0.0s
=> => exporting config sha256:1f27853335f3c02d0f68ae48fe22ea61425663bb9e6a47d5da4d89902b6925e8 0.0s
=> => exporting attestation manifest sha256:b8d00bb9e19d72b3a82aa09ff7575d7bb65ba3721151bd7e9e5b0609aaeeafbe 0.0s
=> => exporting manifest sha256:44e790ee7e180edb5204b4f84b81d65930166fa7f748ea29ff47579f939a9caa 0.0s
=> => exporting config sha256:47813f37e50c569ba1abd16005a80d991f727b4f2b15d0ff45a34f6bf29a837a 0.0s
=> => exporting attestation manifest sha256:80af267def7cff2c3e5b19a0b6003b1ea82a5d787bc8dc979d1a884c59f0af6a 0.0s
=> => exporting manifest list sha256:26b0b36c895978d0cfa611d790a168d55a7e74b7f62110da7268a4510bd27a4d 0.0s
=> => naming to moby-dangling@sha256:26b0b36c895978d0cfa611d790a168d55a7e74b7f62110da7268a4510bd27a4d 0.0s
=> => unpacking to moby-dangling@sha256:26b0b36c895978d0cfa611d790a168d55a7e74b7f62110da7268a4510bd27a4d 0.4sImage huilt; docker image ls --tree
IMAGE ID DISK USAGE CONTENT SIZE EXTRA
<untagged> 26b0b36c8959 548MB 406MB
├─ linux/amd64 1e3d128e5164 135MB 135MB
├─ linux/arm64 11df05f81035 277MB 136MB
└─ linux/s390x 44e790ee7e18 135MB 135MBdocker buildx du
ID RECLAIMABLE SIZE LAST ACCESSED
ilwlv3up0ko6hf6rbte9ui9wl true 264.1MB* 45 seconds ago
vkj7p0cdulb0od9mhnna4pt6i* true 132.3MB 45 seconds ago
ynwtwzuwfepczwuoxx3i2qo28 true 37.54MB 45 seconds ago
w8r4bfgupifuhu68ii2sxirfz true 3.993MB* 45 seconds ago
6cqonmk41d6tl21yt91jpj5go true 3.642MB* 45 seconds ago
2o1taudcfcgfzq7al4ccg9h6y true 3.468MB* 45 seconds ago
xw0sd9sy7a8urox5ghw9tmhiy* true 8.192kB 45 seconds ago
vp8a4ygnqkqzdg5cydonktbzo* true 4.096kB 45 seconds ago
z8ixtkjvpnne6aqnt2ih4fovd true 0B* 45 seconds ago
s4szzze2pnb8py277vweleb4z true 0B* 45 seconds ago
6rt3w0wr1hjvjbq5x7x9ixvsv true 0B* 45 seconds ago
Shared: 275.2MB
Private: 169.9MB
Reclaimable: 445.1MB
Total: 445.1MB |
Beta Was this translation helpful? Give feedback.
-
|
Thank you! This does exactly what I need 😄 |
Beta Was this translation helpful? Give feedback.
-
|
An alternative could also be to push the data as a separate image, and use the new option to mount a volume from an image (note that currently that only allows the content to be mounted read-only); But that requires Docker 28.0 |
Beta Was this translation helpful? Give feedback.
I think this would work if you use the
--linkoption (seeADD --linkandCOPY --link).With that option, the
ADDis copied in am implicitFROM scratchlayer, which allows the layer to be shared, even if the "parent" layers differ. This option should only be used if the location where the content is added does not depend on any parent state (i.e., empty directory, and not expected for that directory to have content that's based on the parent image).This is a quick try with this, but probably best to try for yourself if I'm correct.
👇 worth noting that I'm using docker with the containerd image-store enabled so that I can build multi-platform images with the default builder (not requiring …