amazonica-s3

amazonica-s3

Testing whether Amazonica/s3 library can be used in a native binary image with GraalVM.

Usage

Currently testing:

[amazonica "0.3.152"]

Test with:

lein do clean, uberjar, native-config, native, run-native

To be able to run this need to provide AWS credentials:

export AWS_ACCESS_KEY_ID=xxx
export AWS_SECRET_ACCESS_KEY=yyyy
export AWS_DEFAULT_REGION=eu-west-1

Building this image requires a more complex procedure. After building the uberjar as usual we will run the application with a java-agent that will capture several parameters such as reflected classes, proxy instances, and jni config.

It is good idea to cover as much as possible of the run-time functionality during this step, the more information will be able to collect, the easier/better will be the compilation step.

java -agentlib:native-image-agent=config-output-dir=./target/config/ \
    -jar ./target/amazonica-s3-0.1.0-SNAPSHOT.jar

If multiple execution are required there are ways to merge the generated config.

We will then feed this configuration into the native image compilation by adding the following flag:

# addint the following flag to native-image
# will tell the compiler to use the information we
# extracted from the introspection step

  -H:ConfigurationFileDirectories=./target/config/

IMPORTANT: Finally, I've noticed that after the compilation, the native image was still using the AWS configuration I had at set during the compilation. So even after changing the account credentials it was still connecting to the same account I used during the compilation. To avoid this, I had to add the following flag to force the initialization of the credential provider to happen at runtime.

--initialize-at-run-time=com.amazonaws.auth.DefaultAWSCredentialsProviderChain

After rebuilding, everything worked fine.!