Skip to content

enable instrumentation telemetry#809

Open
emmettbutler wants to merge 2 commits intomainfrom
emmett.butler/enable-instrumentation-telemetry
Open

enable instrumentation telemetry#809
emmettbutler wants to merge 2 commits intomainfrom
emmett.butler/enable-instrumentation-telemetry

Conversation

@emmettbutler
Copy link
Copy Markdown
Contributor

This change removes the configuration that disables instrumentation telemetry and replaces it with a single telemetry upload during cold start, per the Instrumentation Telemetry For Serverless RFC.

DataDog/dd-trace-py#17661 makes this change work as described in the Instrumentation Telemetry For Serverless RFC.

@emmettbutler emmettbutler requested review from a team as code owners April 24, 2026 19:29
Comment thread datadog_lambda/tracing.py

from ddtrace import patch
from ddtrace import __version__ as ddtrace_version
from ddtrace.internal.telemetry import telemetry_writer
Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I recall that importing the telemetry writer added a bunch of cold start time. Did you happen to check to see what impact it's had this time around?

Copy link
Copy Markdown
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I haven't, though the impact is expected to be reduced by the changes from the RFC. Is there a benchmark somewhere in CI that you'd look at to validate this?

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

It doesn't look like we have any on this repo, at least none that I can easily find. Is this the only change required to enable instrumentation telemetry? I can get a manual test going.

Copy link
Copy Markdown
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yes I believe it is

@purple4reina
Copy link
Copy Markdown
Contributor

purple4reina commented Apr 27, 2026

This PR does increase cold start time by approx 4.79% (see first column in chart below).

I deployed a lambda layer from this pr branch and another layer from current on main, then used https://github.com/purple4reina/dd-python-cold-start-testing for testing. It deploys two lambda functions then invokes them with frequent cold starts. Results are sent to dashboard https://ddserverless.datadoghq.com/s/61ba5v0dnxcpd8dm/3gu-sxi-fq2.

@emmettbutler you said you're expecting to see some improvements later. Can you point me to that RFC you referenced and the performance gains we're hoping to see?

Screenshot 2026-04-27 at 13 14 06

@emmettbutler
Copy link
Copy Markdown
Contributor Author

@purple4reina here's the ddtrace branch that this PR assumes / depends on for the performance gains: DataDog/dd-trace-py#17661

@purple4reina
Copy link
Copy Markdown
Contributor

@purple4reina here's the ddtrace branch that this PR assumes / depends on for the performance gains: DataDog/dd-trace-py#17661

This PR doesn't look like it's going to improve cold start times at all, which is what the benchmarks are showing we need to address.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants