Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

[kuberay] More aggressive autoscaling defaults #26985

Merged

Conversation

DmitriGekhtman
Copy link
Contributor

@DmitriGekhtman DmitriGekhtman commented Jul 25, 2022

Signed-off-by: Dmitri Gekhtman [email protected]

Why are these changes needed?

Together with KubeRay PR ray-project/kuberay#414, addresses KubeRay issue ray-project/kuberay#359: gives the autoscaler more aggressive defaults when using KubeRay
(no upscale rate limit, faster idle timeout).

Related issue number

Checks

  • I've run scripts/format.sh to lint the changes in this PR.
  • I've included any doc changes needed for https://docs.ray.io/en/master/.
  • I've made sure the tests are passing. Note that there might be a few flaky tests, see the recent failures at https://flakey-tests.ray.io/
  • Testing Strategy
    • Unit tests
    • Release tests
    • This PR is not tested :(

Signed-off-by: Dmitri Gekhtman <[email protected]>
idle_timeout_minutes = 5.0
if autoscaler_options.get(UPSCALING_KEY) == UPSCALING_VALUE_AGGRESSIVE:
upscaling_speed = 1000 # i.e. big
idle_timeout_minutes = 1.0
Copy link
Contributor

Choose a reason for hiding this comment

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

The result from line 111 can end up being 0 due to rounding down.

Copy link
Contributor Author

Choose a reason for hiding this comment

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

I don't think that's true.
From the CRD definition, the field being divided is an integer. Doing / by a float in Python3 does float division.

However, I don't think we currently validate that the integer is positive -- we can do that later in KubeRay.

upscaling_speed = 1 # Rate-limit upscaling if "Conservative" is set by user.
# This elif is redudant but included for clarity.
elif autoscaler_options.get(UPSCALING_KEY) == UPSCALING_VALUE_DEFAULT:
upscaling_speed = 1000 # i.e. big, no rate-limiting by default
Copy link
Contributor

Choose a reason for hiding this comment

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

I assume people know about this change and its implications?

Copy link
Contributor Author

Choose a reason for hiding this comment

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

People can refer to two groups

Users:
Autoscaling support is new in KubeRay --- it suffices to cover in the docs, which I will do.

Ray developers:
Are aware of the change and its implications.

@DmitriGekhtman DmitriGekhtman merged commit 0860dd2 into ray-project:master Jul 26, 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.

6 participants