Skip to content

docs: clarify DaemonSet overhead is used for sizing, not reserved - #9442

Open
vincentcreusot wants to merge 2 commits into
aws:mainfrom
vincentcreusot:docs/9269-daemonset-overhead-not-reserved
Open

docs: clarify DaemonSet overhead is used for sizing, not reserved#9442
vincentcreusot wants to merge 2 commits into
aws:mainfrom
vincentcreusot:docs/9269-daemonset-overhead-not-reserved

Conversation

@vincentcreusot

@vincentcreusot vincentcreusot commented Jul 25, 2026

Copy link
Copy Markdown

Signed-off-by: Vincent Creusot vincent.creusot@gmail.com

Fixes #9269

Description
Added a small paragraph explaining that Karpenter only uses the Daemonset overhead to size the nodes and do not reserve the space

How was this change tested?
Checked on preview

Does this change impact docs?

  • Yes, PR includes docs updates
  • Yes, issue opened: #
  • No

By submitting this pull request, I confirm that my contribution is made under the terms of the Apache 2.0 license.

@vincentcreusot
vincentcreusot marked this pull request as ready for review July 25, 2026 17:20
@vincentcreusot
vincentcreusot requested a review from a team as a code owner July 25, 2026 17:20
@vincentcreusot
vincentcreusot force-pushed the docs/9269-daemonset-overhead-not-reserved branch from a99e6f6 to 1598bb6 Compare July 31, 2026 14:00

@ryan-mist ryan-mist left a comment

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

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

Thanks for opening this one! Just had a few minor comments

pods itself.

{{% alert title="Note" color="primary" %}}
To guarantee DaemonSets get their capacity, give them a

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

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

I'm not sure if using PriorityClass is an absolute guarantee. IIUC it enables reclaiming capacity for the DS pods, but doesn't guarantee it because of PDBs, equal priority classes, etc.

Copy link
Copy Markdown
Author

Choose a reason for hiding this comment

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

That seems fair, it's not an absolute guarantee. Modified the wording.

Comment thread website/content/en/docs/concepts/scheduling.md Outdated
that capacity on the node after it joins the cluster, and it does not bind
pods itself.

{{% alert title="Note" color="primary" %}}

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

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

We should also add startup taints as an option

Copy link
Copy Markdown
Author

Choose a reason for hiding this comment

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

That makes sense I am adding as an option.

@vincentcreusot
vincentcreusot force-pushed the docs/9269-daemonset-overhead-not-reserved branch from 1598bb6 to 4b3e59c Compare August 23, 2026 23:59
@vincentcreusot

Copy link
Copy Markdown
Author

Thank you for your review @ryan-mist. Took your comment into account and provided more details. Please review changes when you get time.

@vincentcreusot
vincentcreusot force-pushed the docs/9269-daemonset-overhead-not-reserved branch 3 times, most recently from 25e7d82 to e88861a Compare August 27, 2026 03:57
vincentcreusot and others added 2 commits August 27, 2026 21:54
Signed-off-by: Vincent Creusot <vincent.creusot@gmail.com>
Soften the PriorityClass guidance, which is best-effort rather than a
guarantee, and add startup taints as the other option for keeping
workload pods off a node until DaemonSets are ready.

Also reflow the section to the one-sentence-per-line style used
throughout scheduling.md.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Signed-off-by: Vincent Creusot <vincent.creusot@gmail.com>
@vincentcreusot
vincentcreusot force-pushed the docs/9269-daemonset-overhead-not-reserved branch from e88861a to f66ca30 Compare August 28, 2026 03:54
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.

DaemonSet pods can remain Pending when workload pods consume Karpenter-calculated DaemonSet overhead on new nodes

2 participants