Use multi-stage build dockerfile in datafusion-cli and reduce image size from 2.16GB to 89.9MB

[jimexist]

Which issue does this PR close?

We generally require a GitHub issue to be filed for all bug fixes and enhancements and this helps us generate change logs for our releases. You can link an issue to this PR using the GitHub syntax. For example Closes apache/arrow-datafusion#123 indicates that this PR will close issue #123.

Closes #267.

Old image size 2.16GB, new image size is 89.9MB

Rationale for this change

Why are you proposing this change? If this is already explained clearly in the issue then this section is not needed.
Explaining clearly why changes are proposed helps reviewers understand your changes and offer better suggestions for fixes.

We should optimize for the docker image size of datafusion-cli. Using a multi-stage build would contribute to a cleaner build with minimal intermediate build artifacts.

What changes are included in this PR?

There is no need to duplicate the description in the issue here but it is sometimes worth providing a summary of the individual changes in this PR.

Are there any user-facing changes?

If there are user-facing changes then we may require documentation to be updated before approving the PR.

If there are any breaking changes to public APIs, please add the breaking change label.

[jimexist]

related apache/datafusion-ballista#455

[codecov-commenter]

Codecov Report

Merging #266 (2a82cf2) into master (3be087a) will not change coverage.
The diff coverage is n/a.

❗ Current head 2a82cf2 differs from pull request most recent head d36d768. Consider uploading reports for the commit d36d768 to get more accurate results

@@           Coverage Diff           @@
##           master     apache/arrow-datafusion#266   +/-   ##
=======================================
  Coverage   76.21%   76.21%           
=======================================
  Files         140      140           
  Lines       23553    23553           
=======================================
  Hits        17950    17950           
  Misses       5603     5603           

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 3be087a...d36d768. Read the comment docs.

[returnString]

I'm not sure whether there are any expectations around CLI performance in Docker, but it's probably worth seeing whether @andygrove's 2020 article on Rust/Alpine perf issues still holds true: https://andygrove.io/2020/05/why-musl-extremely-slow

[jimexist]

I'm not sure whether there are any expectations around CLI performance in Docker, but it's probably worth seeing whether @andygrove's 2020 article on Rust/Alpine perf issues still holds true: https://andygrove.io/2020/05/why-musl-extremely-slow

i guess dropping alpine part of the change is safer.

[andygrove]

💯

[returnString]

Looks great, excellent savings especially if we're looking at publishing this to Docker Hub soonish :)

[alamb]

 docker build . -f  ./datafusion-cli/Dockerfile -t datafusion-cli

And then tried to run the image:

docker run -it datafusion-cli

I got an error:

(arrow_dev) alamb@ip-10-0-0-124:~/Software/arrow-datafusion$ docker run -it datafusion-cli
thread 'main' panicked at 'called `Result::unwrap()` on an `Err` value: Os { code: 2, kind: NotFound, message: "No such file or directory" }', src/main.rs:56:34
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace

Perhaps we need to update the entry point to be /usr/local/bin/datafusion-cli ?

[jimexist]

@alamb I guess you would need to volume map the /data dir. the error message wasn't specific enough

[alamb]

Indeed @jimexist you are correct:

docker run -v /tmp:/data  -it datafusion-cli

works just fine

[houqp]

minor suggestion not related to this change. i recommend using github container registry instead of dockerhub to host public docker images. dockerhub's rate limit makes it impossible to use the image in any serious production or ci/cd pipeline.

[jorgecarleitao]

I agree with @houqp 👍

[jorgecarleitao]

Looks great.

Note that this makes it more difficult to use caching in docker builds on CI. I think that we can do that on a follow-up PR