30-day hazard rating
- Incidents (30d)
- 7
- Severity-weighted volume
- 8.0
- Critical / high (clustered)
- 0 / 1
- Avg time to resolve
- n/a
Volume of incidents (weighted by severity) over the last 30 days, raised when fixes are slow. 1 = calm, 10 = constant breakage.
- Incidents, 30 days
- 7
- Reports, 30 days
- 7
- Active advisories
- 7
- Regression lifespan
- n/a
- nothing resolved yet
Incidents per day, last 30 days
View as table
| Date | Incidents | Reports |
|---|---|---|
| 2026-11-07 | 2 | 2 |
| 2026-11-06 | 5 | 5 |
| 2026-11-05 | 0 | 0 |
| 2026-11-04 | 0 | 0 |
| 2026-11-03 | 0 | 0 |
| 2026-11-02 | 0 | 0 |
| 2026-11-01 | 0 | 0 |
| 2026-10-31 | 0 | 0 |
| 2026-10-30 | 0 | 0 |
| 2026-10-29 | 0 | 0 |
| 2026-10-28 | 0 | 0 |
| 2026-10-27 | 0 | 0 |
| 2026-10-26 | 0 | 0 |
| 2026-10-25 | 0 | 0 |
| 2026-10-24 | 0 | 0 |
| 2026-10-23 | 0 | 0 |
| 2026-10-22 | 0 | 0 |
| 2026-10-21 | 0 | 0 |
| 2026-10-20 | 0 | 0 |
| 2026-10-19 | 0 | 0 |
| 2026-10-18 | 0 | 0 |
| 2026-10-17 | 0 | 0 |
| 2026-10-16 | 0 | 0 |
| 2026-10-15 | 0 | 0 |
| 2026-10-14 | 0 | 0 |
| 2026-10-13 | 0 | 0 |
| 2026-10-12 | 0 | 0 |
| 2026-10-11 | 0 | 0 |
| 2026-10-10 | 0 | 0 |
| 2026-10-09 | 0 | 0 |
Active advisories
- highInvestigating
Kubernetes Default Preemption Plugin Evicts Partial Gang-Scheduling Members Causing Deadlock
When a high-priority batch workload triggers preemption, the default preemption plugin evaluates node-local priorities and evicts partial members of atomic gang-scheduled workloads. This results in cluster-wide task deadlocks for distributed jobs like multi-node PyTorch or MPI workloads.
Kubernetes7 hours ago
Velocity4.0 - mediumInvestigating
Kubernetes Device Manager Checkpoint State Desynchronizes on Admission Failure
During Pod admission, kubelet's Device Manager assigns physical device IDs and checkpoints the state to disk, but if container admission subsequently fails and the kubelet is restarted, the device checkpoint state can desynchronize.
Kubernetes7 hours ago
Velocity4.0 - mediumInvestigating
Kubernetes Concurrent Watch Decoder Goroutine Panic Handling Issues
An issue was discovered in Kubernetes where adding HandleCrash calls to the concurrent watch decoder's goroutines caused callers to wait indefinitely if a worker recovered from a panic without publishing a result. Recovery mechanisms must ensure failures are properly reported so callers can finish execution.
Kubernetes20 hours ago
Velocity2.6 - mediumInvestigating
Kubernetes ResourceClaim Updates Fail Due to Non-Canonical JSON Opaque Parameters
ResourceClaim updates fail with a 'spec: field is immutable' error when opaque parameters are not in canonical JSON format, such as when created via the Python client. The issue occurs because the apiserver re-encodes JSON fields, causing a byte mismatch during the immutable spec check.
Kubernetes22 hours ago
Velocity2.6 - mediumInvestigating
Kubernetes List Requests with Limit and Field Selector Return Latest Revision Instead of Exact ResourceVersion
List requests with a limit and resourceVersion, without resourceVersionMatch, are incorrectly served at the latest revision rather than the requested exact revision when the field selector targets a single object. This affects API queries for single objects like ConfigMaps by name, returning current data and the wrong resourceVersion.
Kubernetes20 hours ago
Velocity2.5 - mediumInvestigating
Kubernetes Server-Side Apply Status Subresource Incorrectly Acquires Root Field Ownership
Server-Side Apply on a CRD's /status subresource can acquire ownership of an unchanged, non-status root field that the status strategy restores from the old object. This causes subsequent changes by the original main-endpoint manager to conflict with the unintended status ownership.
Kubernetes21 hours ago
Velocity1.9 - mediumInvestigating
Kubernetes Kubelet: Node /etc/hosts Entries Shadow hostAliases in hostNetwork Pods
In hostNetwork: true Pods, kubelet writes the node's /etc/hosts entries before appending spec.hostAliases. Consequently, when the node already contains an entry for the same hostname, lookups resolve to the node's IP first rather than the intended hostAlias IP.
Kubernetes1 day ago
Velocity1.5
Version hazard notices
No version-specific incidents for Kubernetes in the last 90 days.
Resolved incident history
No resolved incidents on record yet.