Repro readme for cargo-lambda/cargo-lambda#467
Created by following the getting started guide: https://www.cargo-lambda.info/guide/getting-started.html
cargo lambda new lambda_test
Answers: 1. Not HTTP, 2. Event type is s3::S3Event
. (Answers aren't important.)
cargo lambda build --release --arm64
Arm64 shouldn't matter either, but it's what I used.
cargo lambda deploy
This works! (Good!) If I log onto AWS web console, I can run test data, and it runs successfully.
Let's clean up, and then see where it breaks:
-
rm -rf target/lambda/
to cleanup. -
export CARGO_TARGET_DIR=~/.cache/cargo/target
to set a shared target dir.
The exact directory doesn't matter, so long as it's not the default ./target
-
cargo lambda build --release --arm64
runs again - great! -
cargo lambda deploy
fails with:
Error:
× binary file for lambda_test not found, use `cargo lambda build` to create it
First bug: lambda deploy
doesn't know to use CARGO_TARGET_DIR.
Let's try to work around that:
cargo lambda deploy --binary-path ~/.cache/cargo/target/lambda/lambda_test/bootstrap
Ok, this command worked! Albeit with the following issue:
Usability issue: it's not obvious that you're forced to supply the name of the function. It'd be nice if it were
required if
--binary-path is set. To deploy to the same function name, I have to do: cargo lambda deploy --binary-path ~/. cache/cargo/target/lambda/lambda_test/bootstrap lambda_test
Anyway, after running the modified command, it deployed to the correct function name. Let's run the function. Log into AWS console. Run test data. It fails with the error in the original ticket:
START RequestId: a482048b-2a77-46df-bb56-ab26cb1a6ecd Version: $LATEST
RequestId: a482048b-2a77-46df-bb56-ab26cb1a6ecd Error: Couldn't find valid bootstrap(s): [/var/task/bootstrap /opt/bootstrap]
Runtime.InvalidEntrypoint
END RequestId: a482048b-2a77-46df-bb56-ab26cb1a6ecd
REPORT RequestId: a482048b-2a77-46df-bb56-ab26cb1a6ecd Duration: 14.25 ms Billed Duration: 15 ms Memory Size: 128 MB Max Memory Used: 2 MB
After doing some digging, I believe it's because AWS assumes it's a zip file, and not a statically linked binary. If there's a setting I'm missing that tells AWS to treat it as a binary, please let me know. The two options appear to be zip or container.
Anyway, it's configured as a zip, so let's try zipping before we pass it to deploy.
-
cd ~/.cache/cargo/target/lambda/lambda_test/
-
zip -r lambda_test.zip bootstrap
-
cd -
to go back to the project. Let's try deploying again: -
cargo lambda deploy --binary-path ~/.cache/cargo/target/lambda/lambda_test/lambda_test.zip lambda_test
Oops. This fails with:
Error:
× Could not read file magic
Unfortunately, lambda deploy
does not accept a architecture flag, which would allow it to then skip the "file
magic" check.
There's no workaround for this, so I'm stuck. The --binary-path
arg for lambda deploy
appears to be broken.
lambda deploy
should use the CARGO_TARGET_DIR directory when looking for binaries.--binary-path
should zip the binary before uploading.- If
--binary-path
is used, theNAME
should be required. The value currently defaults to the binary name (bootstrap
), and it seems unlikely someone would ever want that value to be the default.