Skip to content

[WAGED Tag Isolation 8/16] Attribute a capacity deficit to the cliques that caused it - #288

Draft
LZD-PratyushBhatt wants to merge 1 commit into
lzd/waged-iso-07-hooksfrom
lzd/waged-iso-08-deficit
Draft

LZD-PratyushBhatt wants to merge 1 commit into
lzd/waged-iso-07-hooksfrom
lzd/waged-iso-08-deficit

Conversation

@LZD-PratyushBhatt

@LZD-PratyushBhatt LZD-PratyushBhatt commented Sep 20, 2026

Copy link
Copy Markdown
Collaborator

Description

The cluster wide capacity check that runs before any placement is a tag blind sum, so one wildly oversubscribed clique can drag a dimension negative while every other clique still fits comfortably on its own nodes. Throwing there would freeze the whole cluster, which is what this mode exists to prevent.

Walk the attribution blocks instead, set aside the ones whose own replicas cannot fit on their own nodes, and re-evaluate the deficit on what is left. A deficit that survives that, or a cluster that is one single block, still throws exactly as before.

This completes the production code for the feature.

Stacked PR 8 of 16 of the WAGED instance tag isolation stack.

  • Base branch: lzd/waged-iso-07-hooks
  • Only the top commit belongs to this PR
  • The full stack merges into waged-tag-based-isolation, which merges into dev at the end

Tests

Existing WAGED suites stay green. The attribution tests arrive in lzd/waged-iso-16-tests-capacity.

Tests run: 65, Failures: 0, Errors: 0, Skipped: 0

Changes that Break Backward Compatibility

None. The new cluster config flag defaults to false, and with it off every code path added here behaves exactly as it does today.

Documentation

The design doc lands in lzd/waged-iso-13-doc as docs/design/002-waged-instance-tag-isolation.md.

Code Quality

Formatted with helix-style.xml.

The cluster wide capacity check that runs before any placement is a tag
blind sum, so one wildly oversubscribed clique can drag a dimension
negative while every other clique still fits comfortably on its own
nodes. Throwing there would freeze the whole cluster, which is what this
mode exists to prevent.

Walk the attribution blocks instead, set aside the ones whose own
replicas cannot fit on their own nodes, and re-evaluate the deficit on
what is left. The blocks set aside are carried over like any other
failed group. A deficit that survives that, or a cluster that is one
single block, still throws exactly as before.

Only reachable on a path where the default mode has already decided to
throw, so nothing changes for a cluster that fits.
@LZD-PratyushBhatt
LZD-PratyushBhatt added this pull request to stack #297 September 20, 2026 05:46
@LZD-PratyushBhatt LZD-PratyushBhatt changed the title Attribute a capacity deficit to the cliques that caused it [WAGED Tag Isolation 8/16] Attribute a capacity deficit to the cliques that caused it Sep 20, 2026
@LZD-PratyushBhatt LZD-PratyushBhatt added the waged-tag-level-isolation Opt-in WAGED failure isolation per instance group tag (clique) label Sep 20, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

waged-tag-level-isolation Opt-in WAGED failure isolation per instance group tag (clique)

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant