feat: Support CRaC and priming of powertools large messages

[David-DAM]

Issue number: Closes #2000

Description of changes:

• Added CRaC support in powertools-large-messages, following the guide and blogs mentioned on the ticket.
• Added automatic priming for powertools-large-messages by preloading all the required classes, the list of classes is fetched from the JVM at compile time.
• Modified Protobuf versions on the powertools-examples-kafka

Checklist

• Meet tenets criteria
• Update tests
• Update docs
• PR title follows conventional commit semantics

Breaking change checklist

RFC issue #:

• Migration process documented
• Implement warnings (if it can live side by side)

By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of your choice.

[David-DAM]

closes #2000

[David-DAM]

Hello @phipag, can you review my PR when you have time? Thanks!

[phipag]

Hey @David-DAM, thank you this looks good on first sight. I will have time to test this very soon and will get back to you here.

In the meantime, did you test it in your own AWS account? E.g. add debug logs to the CRaC hooks and check if they are executed?

[David-DAM]

Thanks! Also I don't know how to test it on my own AWS account, is there a guide?

[phipag]

Hey @David-DAM, apologies for the delay. I am testing now and get back to you in a bit here. The approach is quite manual at the moment:

• Checkout the PR
• Deploy a handler in Lambda using the changed code
• Enable snapstart for the handler
• Check cloudwatch logs that the priming code gets executed

[phipag]

Hey @David-DAM, thanks for this. I tested it end to end with SnapStart in my own account and hit one blocker.

The CRaC registration runs in a static initializer, which only fires when the class first loads. In a normal handler the first reference to LargeMessages is inside handleRequest, which runs at invoke time, after the SnapStart checkpoint. So the class never loads during INIT, never registers before the snapshot, and beforeCheckpoint never runs. I confirmed this in CloudWatch: no priming logs across several runs until I referenced the class during INIT.

Please follow the same pattern as our Tracing module: expose a placeholder LargeMessages.init() and document that users call it during INIT, for example in the handler constructor.

public MyHandler() {
    LargeMessages.init();
}

For the docs, add a "SnapStart priming" section to docs/utilities/large_messages.md, mirroring the one in docs/core/tracing.md that renders at Lambda SnapStart priming.

[phipag]

By the way: while testing this I found that class preloading was loading 0 classes because ClassPreLoader did not handle the leading whitespace in classesloaded.txt. I made it more robust in #2560.

[sonarqubecloud]

Quality Gate passed

Issues
2 New issues
0 Accepted issues

Measures
0 Security Hotspots
0.0% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud

[David-DAM]

@phipag Thanks for the clarification, i have added the changes related to that