Skip to content

fix(pd): wait for KV transport readiness and track transfer progress - #1600

Open
WANDY666 wants to merge 1 commit into
mainfrom
fix/pd-kv-transfer-progress
Open

WANDY666 wants to merge 1 commit into
mainfrom
fix/pd-kv-transfer-progress

Conversation

@WANDY666

Copy link
Copy Markdown
Contributor

Wait for each KV transport module to finish initialization before consuming its results, and detect startup failures instead of waiting indefinitely. Pending decode pages use request progress for idle timeouts; active transfers retain their per-page deadline.

This prevents slow startup or long prefill from being mistaken for a failed transfer. Generation failures before the first streamed output return HTTP 500 instead of starting a successful response.

Tests: readiness, worker exit, bounded startup, progress timeout and HTTP error regressions (21 passed).

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants