feat: add ModelServing eviction protection - #1484
Conversation
Signed-off-by: Avinash Kumar Deepak <avinash8655279@gmail.com>
|
Caution The consumer version of Gemini Code Assist on GitHub has been sunset. All code review activity has officially ceased. |
|
[APPROVALNOTIFIER] This PR is NOT APPROVED This pull-request has been approved by: The full list of commands accepted by this bot can be found here. DetailsNeeds approval from an approver in each of these files:Approvers can indicate their approval by writing |
|
Have you verified this feature? |
Signed-off-by: Avinash Kumar Deepak <avinash8655279@gmail.com>
sir, the controller and validation tests r passing. Im working through the full node-drain and concurrent eviction scenarios now. I'll share the results here once the verification is complete. |
|
@avinxshKD ditto #1485 |
|
@avinxshKD Sorry, this has already been taken care of by #1485. It has been validated and tested in both the production environment and on Kind clusters. To keep the PR clean, I will go ahead and close this one. Thank you for your contribution. |
thankyou sir will be learing from and watching that PR, closing this |
What type of PR is this?
/kind enhancement
What this PR does / why we need it:
Adds configurable eviction protection to ModelServing workloads.
Operators can preserve either a minimum number of ready ServingGroups or a minimum number of available instances for specific roles. The controller manages PodDisruptionBudgets for the protected ServingGroups or role instances.
Which issue(s) this PR fixes:
Fixes #1481
Bug evidence (required for bug-related PRs):
N/A
Special notes for your reviewer:
PDBs count Pods, while a ServingGroup or role instance can contain multiple Pods. This change creates a PDB for each protected logical unit so an eviction cannot partially disrupt it.
Does this PR introduce a user-facing change?: