Skip to content

SLM: Job Pod replacement during node maintenance #6262

Description

@rthallisey

Enhancement Description

When a Node becomes unreachable during maintenance, its Pods may never reach a terminal phase that the Job controller can observe. A Job using podReplacementPolicy: Failed can then wait indefinitely instead of creating a replacement Pod, leaving the Job and higher-level queueing systems stuck.

The Job controller will use the MaintenanceInProgress condition as Node context when deciding whether a Pod can be replaced without waiting for its terminal phase. The KEP will define the replacement and accounting semantics so Jobs can make progress without unintentionally running duplicate Pods.

Please keep this description up to date. This will help the Enhancement Team to track the evolution of the enhancement efficiently.

/sig apps
/sig node
/wg node-lifecycle

Metadata

Metadata

Labels

sig/appsCategorizes an issue or PR as relevant to SIG Apps.sig/nodeCategorizes an issue or PR as relevant to SIG Node.wg/node-lifecycleCategorizes an issue or PR as relevant to WG Node Lifecycle.

Type

No type

Projects

Status
Needs Triage
Status
No status
Status
No status

Milestone

No milestone

Relationships

None yet

Development

No branches or pull requests

Issue actions