fix(rest): resume paused JSON-shred parser when its channel drains empty (back-pressure deadlock)#1005
Merged
Conversation
…nnel empty The streaming JSON shredder's consumer resumed the back-pressured parser only after draining `resumeThreshold` (50k) events since the last resume. If the consumer drained the bounded channel empty while still paused and below that threshold, it never resumed the parser, so no further events (and never the end-of-stream) arrived: the consumer blocked forever on an empty channel while the producer stayed paused — a permanent CPU-idle deadlock for any input whose post-pause tail is smaller than the threshold (reachable directly with a small channelCapacity). Replace the `for (event in channel)` iterator with an explicit tryReceive()/receiveCatching() loop that, before suspending on an empty channel, resumes the producer when it is paused (low-water fallback). The 50k batched resume is kept purely as a throughput optimisation. Add a deterministic regression test (channelCapacity=16 → deadlocks before the fix, passes after) and a production-scale large-payload async test. Bumps version to 1.0.0-alpha15.
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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
Fixes a back-pressure deadlock in the REST API's
KotlinJsonStreamingShredder.The channel-based shredder runs a producer (Vert.x event loop → bounded channel) and a consumer (worker thread → blocking DB inserts), using
parser.pause()/resume()for back-pressure. The consumer resumed the paused parser only after drainingresumeThreshold(50,000) events since the last resume. If the consumer drained the channel empty while the parser was still paused and fewer than 50k events had been processed since that resume, it never resumed the parser — so no further events (and never the end-of-stream) arrived. The consumer then blocked forever on the empty channel while the producer stayed paused: a permanent CPU-idle deadlock.This is reachable whenever an input's post-pause tail is smaller than the resume threshold — directly and deterministically with a small
channelCapacity.Fix
Replace the
for (event in channel)iterator (which silently suspends on an empty channel) with an explicittryReceive()→receiveCatching()loop. Before suspending on an empty channel, if the producer is paused it is resumed (low-water fallback). The existing 50k batched resume is retained purely as a throughput optimisation — it lets the producer refill while the consumer drains the remainder.Tests
testBackpressureTailDoesNotDeadlock— deterministic regression:channelCapacity=16forces a pause whose drain never reaches the 50k threshold; deadlocks (TimeoutException) before the fix, passes after.testProductionScaleChunkedAsync— production-scale (~5000 nested objects / ~150k events) at the default 100k capacity, exercising real back-pressure end-to-end.All existing
JsonStreamingShredderTestsync + async tests continue to pass.