fix(router): don't consume circuit breaker probe slot during routing#172
Merged
fix(router): don't consume circuit breaker probe slot during routing#172
Conversation
…agent traffic Claude Code can fire multiple concurrent upstream requests via subagents. A single HTTP/2 connection was sufficient for basic use but could bottleneck during parallel agent workflows. Bumping to 3 connections per provider per session gives headroom for typical subagent concurrency.
…resolution Replace canProceed() with getState() === "open" in the distribution routing path. canProceed() has a side-effect of transitioning open→half-open and consuming the probe slot, which prevented the actual request in forwardWithFallback() from probing the provider. This caused providers to stay blocked after circuit cooldown in distribution mode — the routing layer consumed the probe but no actual request was sent, so recovery depended on the slow health-probe timeout path (5s). Add regression tests proving getState() is read-only and canProceed() correctly gates probe access. Closes #156
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Summary
canProceed()withgetState()in distribution routing to prevent probe slot consumptiongetState()is read-only andcanProceed()correctly gates probesProblem
In
router.ts, the distribution routing code calledcanProceed()to check if a provider's circuit breaker was open. This had a side-effect: it transitioned the breaker fromopen→half-openand consumed the probe slot. When the actual request reachedforwardWithFallback()and calledcanProceed()again, the probe was already consumed (halfOpenInProgress=true), so the provider was skipped.Result: providers stayed blocked after circuit cooldown in distribution mode. Recovery only happened via the slow health-probe timeout path (5s delay per cycle).
Fix
Use
getState() === "open"for the read-only routing check. The actualcanProceed()call inforwardWithFallback()is now the only place that can consume the probe slot.Test plan
circuit-breaker.test.tsCloses #156