Skip to content

Conversation

@ryanzhang-oss
Copy link
Contributor

@ryanzhang-oss ryanzhang-oss commented Oct 5, 2022

Description of your changes

Add a load test to measure the client-side latency for placement operations.

I have:

  • Run make reviewable to ensure this PR is ready for review.

How has this code been tested

Special notes for your reviewer

I have run the test against fleet but hit a guard issue that prevent it from running more than 20 in parallel

Copy link
Contributor

@Fei-Guo Fei-Guo left a comment

Choose a reason for hiding this comment

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

One suggestion, can we specify the test length (e.g, the total number of created CRP crs) and report the metrics result in the end. Hence the test can be used as a benchmark.

@ryanzhang-oss
Copy link
Contributor Author

One suggestion, can we specify the test length (e.g, the total number of created CRP crs) and report the metrics result in the end. Hence the test can be used as a benchmark.

Good idea to have a test length. It turned out that getting metrics with Prometheus client go lib is not easy (especially with the histogram). It's a lot easier to curl the metrics endpoint.

@Fei-Guo
Copy link
Contributor

Fei-Guo commented Oct 13, 2022

I would suggest adding a few atomic_int variables to count the total number of CRP crs, the number of timeout crs, etc, and print out the result when the program exits, we don't need to dump the histogram. This should good enough for finding some critical issues.

Fei-Guo
Fei-Guo previously approved these changes Oct 14, 2022
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