Missing logs and APM traces when a lambda reaches maximum memory

[glenthomas]

Similar to #89.

When a lambda reaches maximum memory it is killed by the runtime and the following is written to the CloudWatch log stream:

RequestId: 45b36944-5896-4d3b-a4eb-b2af494eb662 Error: Runtime exited with error: signal: killed
Runtime.ExitError

Logs and traces are not present in Datadog.

The following layer versions are present:

[hghotra]

Hi @glenthomas, can you try the latest version of the extension and report if this is still an issue? You should be able to see the logs with the latest extension.

[smoya]

@hghotra does it work for the containerized version of the extension? I have it running but I see no traces nor logs when oom-killed.

[DylanLovesCoffee]

Hey @glenthomas, I'm curious is there is a subsequent invocation after the Lambda gets OOM killed? Typically we will buffer logs during an OOM invocation and then send them to Datadog in the next invocation. However, if invocations are sparse and the Lambda environment is shut down then those logs will be lost.

@smoya just to clarify, you are running your Lambda as a container image? And also the same as the above, are there subsequent invocations or are they sparse?