fix: deep-copy capacity map to prevent informer cache corruption - #1065
Merged
Conversation
The capacity filter assigned hypervisor Status.Capacity map references directly into a local working map, then mutated those maps during free-resource calculations. Since the maps are shared with the controller-runtime informer cache, each scheduling pass permanently decremented cached capacity values, progressively starving scheduling until all hosts appeared full. Fix: deep-copy the capacity map per hypervisor before subtracting allocations so the informer cache is never mutated. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
cortex-ai-agents
Bot
requested review from
PhilippMatthes,
SoWieMarkus,
auhlig,
juliusclausnitzer,
mblos and
umswmayj
as code owners
July 20, 2026 07:15
Contributor
Test Coverage ReportTest Coverage 📊: 70.3% |
juliusclausnitzer
approved these changes
Jul 27, 2026
Merged
juliusclausnitzer
added a commit
that referenced
this pull request
Jul 27, 2026
## Release cortex v0.3.4 Release prep PR: #1094 ✅ ### Changes since v0.3.3 #### Bug Fixes - **Deep-copy capacity map to prevent informer cache corruption** — the capacity filter assigned hypervisor Status.Capacity map references directly into a local working map, then mutated those maps during free-resource calculations. Since the maps are shared with the controller-runtime informer cache, each scheduling pass permanently decremented cached capacity values, progressively starving scheduling until all hosts appeared full (#1065) - **Reset gauge metrics in reservation monitor to prevent stale values** — adds `Reset()` calls on `numberOfReservations` and `reservedResources` GaugeVecs before repopulating in `Collect`, preventing stale label combinations from persisting when reservations are deleted or change status (#1092) - **Stop committed resource controller reconcile storm** — `setAccepted` unconditionally wrote a new AcceptedAt timestamp and a time-varying StatusSummary on every call, causing a non-empty status patch even when the CR was already accepted. Adds an early-return guard that skips the patch when Ready=True/Accepted is already set for the current generation (#1089) #### Non-breaking Changes - Add History CRD documentation and multicluster guide link (#1091) - Update `github.com/prometheus/client_golang` to v1.24.1 (#1090) - Update `k8s.io/api`, `k8s.io/apimachinery`, `k8s.io/client-go` to v0.36.3 (#1086) - Update `github.com/sapcc/go-bits` (#1085) - Update `kube-prometheus-stack` to 87.19.2 (#1090) - Remove year from licensing section in readme (#1087) ### Version Bumps | Chart | Previous | New | |-------|----------|-----| | cortex (library) | 0.3.3 | 0.3.4 | | cortex-shim (library) | 0.1.9 | 0.1.10 | | cortex-nova | 0.0.83 | 0.0.84 | | cortex-cinder | 0.0.83 | 0.0.84 | | cortex-manila | 0.0.83 | 0.0.84 | | cortex-crds | 0.0.83 | 0.0.84 | | cortex-ironcore | 0.0.83 | 0.0.84 | | cortex-pods | 0.0.83 | 0.0.84 | | cortex-placement-shim | 0.1.9 | 0.1.10 |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
The capacity filter assigned hypervisor Status.Capacity map references directly into a local working map, then mutated those maps during free-resource calculations. Since the maps are shared with the controller-runtime informer cache, each scheduling pass permanently decremented cached capacity values, progressively starving scheduling until all hosts appeared full.
Fix: deep-copy the capacity map per hypervisor before subtracting allocations so the informer cache is never mutated.
Assisted-by: Claude Code:claude-opus-4-8 [Bash] [Read] [Edit]