fix(reconcile): re-pin scheduled models that drifted to HOT-but-unpinned - #268
Open
MichaelRipa wants to merge 1 commit into
Open
fix(reconcile): re-pin scheduled models that drifted to HOT-but-unpinned#268MichaelRipa wants to merge 1 commit into
MichaelRipa wants to merge 1 commit into
Conversation
A scheduled model whose pinned replica dies can be re-created unpinned by a hotswap request. It then passes the old "is it HOT?" reconcile check, so the pinned flag is never restored — the model serves hotswap keys but rejects every non-hotswap key, while the dashboard still shows it as pinned. Observed in prod on Llama-3.1-405B (45 entitled keys locked out). Track per-model pinned-ness (not just HOT-ness) and treat "scheduled + HOT + any unpinned replica" as drift: evict and redeploy pinned. The API gate reads replicas[0].pinned and deploy is additive, so an in-place flip is impossible — eviction is required. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
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
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
A scheduled model whose pinned replica dies can be re-created unpinned by a hotswap request. It then passes the old "is it HOT?" reconcile check, so the pinned flag is never restored — the model serves hotswap keys but rejects every non-hotswap key, while the dashboard still shows it as pinned. Observed in prod on Llama-3.1-405B (45 entitled keys locked out).
Track per-model pinned-ness (not just HOT-ness) and treat "scheduled + HOT + any unpinned replica" as drift: evict and redeploy pinned. The API gate reads replicas[0].pinned and deploy is additive, so an in-place flip is impossible — eviction is required.