fix: shutdown queue on publish error if not done #412
Merged
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Outdated suggestions cannot be applied.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
This is a patch on top of 0.13.2. While debugging network disconnect issues in Boost during retrieval we discovered a leak in go-routines for graphsync. The issue is that a hard network disconnect may still result in the MessageQueue being restarted if there are messages the server is still attempting to send.
This change does not fully fix the issue but after multiple runs of force disconnecting 500 retrievals, we saw a 10x reduction in open goroutines. More robust handling for hard network disconnects may be warranted in the 0.14.x line, but that is also likely a non trivial effort. This gets us most of the way.
Goroutine dumps from boost
The below were goroutine diffs between startup and after executing 500 forceful disconnects.
Before
After