feat(profiler): switch to zstd compression by default#3970
Conversation
This should decrease the amount of egress traffic from the profiler by 20-30%. If neccessary, the old compression strategy (a mix of gzip-1, gzip-6 and none) can be restored by setting the following environment variable: DD_PROFILING_DEBUG_COMPRESSION_SETTINGS=legacy
BenchmarksBenchmark execution time: 2025-09-16 18:00:23 Comparing candidate commit f825e9e in PR branch Found 0 performance improvements and 0 performance regressions! Performance is the same for 24 metrics, 0 unstable metrics. |
|
Note encountered a potential flake. Re-running.
|
Edit: Took a shot at this, see: #3971 |
|
Got the same flake I got earlier again.
Edit: Turns out this is a hard fail, not a flake. But don't have time to fix right now :( |
…ble-zstd-by-default
|
/merge |
|
View all feedbacks in Devflow UI.
This merge request is not mergeable yet, because of pending checks/missing approvals. It will be added to the queue as soon as checks pass and/or get approvals.
devflow unqueued this merge request: It did not become mergeable within the expected time |
|
More flakes again 😢 |
|
/merge |
|
View all feedbacks in Devflow UI.
The expected merge time in
|



What does this PR do?
This should decrease the amount of egress traffic from the profiler by 20-30%.
If necessary, the old compression strategy (a mix of gzip-1, gzip-6 and none) can be restored by setting the following environment variable:
Motivation
Reduce network egress for customers and ingress for Datadog. See JIRA.
Reviewer Notes
This is the minimal viable patch to achieve the goal. Some of the existing tests made strong assumptions about the legacy compression strategy, so I fixed them up via
DD_PROFILING_DEBUG_COMPRESSION_SETTINGS=legacyfor now. We could invest in refactoring them to work regardless of the default compression. I think it's unnecessary, but could be convinced otherwise.Reviewer's Checklist
./scripts/lint.shlocally.Unsure? Have a question? Request a review!