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

Improve default bucket count in FTE #21977

Merged
merged 1 commit into from
May 15, 2024

Conversation

dekimir
Copy link
Contributor

@dekimir dekimir commented May 15, 2024

Description

We observed high skew in FTE tasks due to overly low bucket count.

Additional context and related issues

Release notes

(x) This is not user-visible or is docs only, and no release notes are required.
( ) Release notes are required. Please propose a release note for me.
( ) Release notes are required, with the following suggested text:

# Section
* Fix some things. ({issue}`issuenumber`)

@cla-bot cla-bot bot added the cla-signed label May 15, 2024
@dekimir dekimir requested review from losipiuk and sopel39 May 15, 2024 00:14
@losipiuk
Copy link
Member

CI: #18697

We observed high skew in FTE tasks due to overly low bucket count.
@losipiuk losipiuk merged commit ab7570f into trinodb:master May 15, 2024
92 of 93 checks passed
@github-actions github-actions bot added this to the 448 milestone May 15, 2024
// then each task gets one bucket. So let's return the total number of tasks here. This number, unfortunately, is hard to know precisely because of the variety of ways in
// which schedulers create tasks. We therefore have to estimate.
if (getRetryPolicy(session) != RetryPolicy.TASK) {
// Pipeline schedulers typically create as many tasks as there are nodes.
Copy link
Member

Choose a reason for hiding this comment

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

this is too low even for pipeline mode as it affects local scaling too

Copy link
Contributor Author

Choose a reason for hiding this comment

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

Can you say more? What is "local scaling"? I see that stages schedulers create a task per stage; can this exceed the node count and is it relevant?

Copy link
Member

Choose a reason for hiding this comment

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

I've created follow up PR #22000

Copy link
Member

Choose a reason for hiding this comment

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

See

PartitionFunction partitionFunction = createPartitionFunction(
for example. If bufferCount is larger than bucket count, then we are not fully utilizing writer concurrency within a task

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

Successfully merging this pull request may close these issues.

None yet

3 participants