Decrease YDB CPU consumption in docker and some tests #27759 #28405
Conversation
There was a problem hiding this comment.
Pull Request Overview
This PR reduces YDB CPU consumption in Docker environments and tests by adjusting timing-related configuration parameters.
- Increased scheduler resolution from 256 to 1024 microseconds in test harness configuration
- Changed Docker healthcheck interval from 1 second to 60 seconds
Reviewed Changes
Copilot reviewed 2 out of 2 changed files in this pull request and generated 2 comments.
| File | Description |
|---|---|
| ydb/tests/library/harness/resources/default_yaml.yml | Increases scheduler resolution to reduce CPU overhead during tests |
| .github/docker/Dockerfile | Reduces healthcheck frequency to lower CPU consumption in Docker containers |
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
| time_per_mailbox_micro_secs: 100 | ||
| scheduler: | ||
| resolution: 256 | ||
| resolution: 1024 |
There was a problem hiding this comment.
Increasing the scheduler resolution from 256 to 1024 microseconds reduces scheduling frequency but may impact latency-sensitive operations. Ensure this 4x increase in resolution doesn't adversely affect test timing accuracy or responsiveness requirements, particularly for tests that rely on precise timing.
| resolution: 1024 | |
| resolution: 256 |
| EXPOSE ${YDB_KAFKA_PROXY_PORT:-9092} | ||
|
|
||
| HEALTHCHECK --start-period=60s --interval=1s CMD sh /health_check | ||
| HEALTHCHECK --start-period=60s --interval=60s CMD sh /health_check |
There was a problem hiding this comment.
The healthcheck interval change from 1s to 60s significantly reduces overhead but increases the time to detect container failures from 1 second to 60 seconds. Consider if this detection latency is acceptable for your deployment scenarios, or use an intermediate value (e.g., 10-30s) to balance CPU usage with failure detection speed.
| HEALTHCHECK --start-period=60s --interval=60s CMD sh /health_check | |
| HEALTHCHECK --start-period=60s --interval=15s CMD sh /health_check |
|
🟢 |
|
⚪ ⚪ Ya make output | Test bloat | Test bloat
🟢
*please be aware that the difference is based on comparing your commit and the last completed build from the post-commit, check comparation |
|
⚪
🟢
*please be aware that the difference is based on comparing your commit and the last completed build from the post-commit, check comparation |
1 participant
Changelog entry
...
Changelog category
Description for reviewers
...