Restore model probe for service routers - #4322
Merged
Merged
Conversation
Since #4270, a router replica of a service with `model` and no `probes` was probed with `GET /health` instead of the chat completions request. The latter deadlocked: the router only answers it once workers are registered, and the worker sync skipped routers that were not ready. #4313 dropped the readiness check from the sync, so this deadlock is gone. `/health` passes before the router has any workers. In a rolling deployment, the replacement router was considered ready, and the old one was scaled down, before the replacement could serve requests. Now that #4320 syncs workers with every running router, the replacement passes the chat completions probe once it has workers. Workers behind a router still get no default probe, as they may not serve HTTP at all (gRPC workers). With a Dynamo router, a router rolling deployment now waits for the replacement indefinitely instead of scaling down the old router, as workers stay attached to the old router's etcd/NATS. Such deployments were broken before as well and are to be fixed separately. Co-Authored-By: Claude Opus 5.5 (1M context) <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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Since #4270, a router replica of a service with
modeland noprobeswas probed withGET /healthinstead of the chat completions request. The latter deadlocked: the router only answers it once workers are registered, and the worker sync skipped routers that were not ready. #4313 dropped the readiness check from the sync, so this deadlock is gone./healthpasses before the router has any workers. In a rolling deployment, the replacement router was considered ready, and the old one was scaled down, before the replacement could serve requests. Now that #4320 syncs workers with every running router, the replacement passes the chat completions probe once it has workers.Workers behind a router still get no default probe, as they may not serve HTTP at all (gRPC workers).
With a Dynamo router, a router rolling deployment now waits for the replacement indefinitely instead of scaling down the old router, as workers stay attached to the old router's etcd/NATS. Such deployments were broken before as well and are to be fixed separately.