Silently dropped metrics when using sendDistributionMetric

[AaronMoat]

Description

I have a Lambda function, without log forwarding, that issues a high count of metrics with sendDistributionMetric.

I have noticed that, at times, some subset of a Lambda invocation's metrics are silently dropped. I tried running with { shouldRetryMetrics: true, debugLogging: true }, with no useful output from the debugger to distinguish successful runs and unsuccessful runs.

Upon looking into this repo's code, I noticed that b20c40e introduced a change where we continually recreate this.currentProcessor, meaning that when we await it during flushing at the end of the lambda invocation, it's only awaiting one of the calls to sendDistributionMetric, not all of them. This could be causing my problem.

Does this sound correct?

Specifications

• Datadog Lambda Layer version: 72
• Node version: nodejs20.x

[astuyve]

Hi @AaronMoat thanks for the note!

Would you mind opening a support ticket with us so we could take a closer look? You can use the in-app chat or even just email [email protected]. Please include this issue in the report as to help expedite the process.

Thanks!