Add flush method to tracer

[wu-json]

What does this PR do?

Adds a .flush() method to the tracer, which allows users to manually flush collected spans to the exporter agent. This is useful for ensuring spans aren't lost when force exiting a process with process.exit.

Motivation

My team recently migrated our Node.js crons from NewRelic to Datadog. In NewRelic land, we had access to newrelic.shutdown({ collectPendingData: true }), which allowed us to send all metrics and traces to NewRelic before killing the process with process.exit (we do not exit gracefully in our crons for reasons I will not dive into here).

In Datadog land however, we found that such a method did not exist, which means we were not getting traces when crons completed. The only workaround is to set a delay before exiting the process, which is hacky and not guaranteed to work.

There are two existing threads with conversations requesting this feature as well.

• Traces not flushing on process exit #708
• Add flush() method to tracer #1227

The other PR that attempted to add this functionality was abandoned so I'm taking a stab at picking up where they left off.

Checklist

• Unit tests.
• TypeScript definitions.
• TypeScript tests.

Additional Notes

I really want this feature - happy to work with maintainers on making whatever edits we need to get this in.

[simon-id]

FYI There is no point rebasing this PR on master anymore, as Juan recreated it to run the CI

[juan-fernandez]

hey @wu-json thanks for the contribution. We've discussed this internally and while we like it, we're afraid to deviate from the public API other datadog libraries have. We're thinking of a way forward. Stay tuned 😄

[wu-json]

hey @wu-json thanks for the contribution. We've discussed this internally and while we like it, we're afraid to deviate from the public API other datadog libraries have. We're thinking of a way forward. Stay tuned 😄

@juan-fernandez Ah - understood. Let me know if I can be of use here after the direction is sorted out internally. Would be happy to help with the solution.