http: Detect remote disconnect - 2nd attempt #27245
Closed
Conversation
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
Co-authored-by: Fabian Jahr <fjahr@protonmail.com>
|
The following sections might be updated with supplementary metadata relevant to reviewers and maintainers. ReviewsSee the guideline for information on the review process. ConflictsReviewers, this pull request conflicts with the following ones:
If you consider this pull request important, please also help to review the conflicting pull requests. Ideally, start with the one that should be merged first. |
|
1 task
|
Closed in favor of #27909 |
fanquake
added a commit
that referenced
this pull request
Oct 4, 2023
…lient disconnection 68f23f5 http: bugfix: track closed connection (stickies-v) 084d037 http: log connection instead of request count (stickies-v) 41f9027 http: refactor: use encapsulated HTTPRequestTracker (stickies-v) Pull request description: #26742 significantly increased the http server shutdown speed, but also introduced a bug (#27722 - see #27722 (comment) for steps to reproduce on master) that causes http server shutdown to halt in case of a remote client disconnection. This happens because `evhttp_request_set_on_complete_cb` is never called and thus the request is never removed from `g_requests`. This PR fixes that bug, and improves robustness of the code by encapsulating the request tracking logic. Earlier approaches (#27909, #27245, #19434) attempted to resolve this but [imo are fundamentally unsafe](#27909 (comment)) because of differences in lifetime between an `evhttp_request` and `evhttp_connection`. We don't need to keep track of open requests or connections, we just [need to ensure](#19420 (comment)) that there are no active requests on server shutdown. Because a connection can have multiple requests, and a request can be completed in various ways (the request actually being handled, or the client performing a remote disconnect), keeping a counter per connection seems like the approach with the least overhead to me. Fixes #27722 ACKs for top commit: vasild: ACK 68f23f5 theStack: ACK 68f23f5 Tree-SHA512: dfa711ff55ec75ba44d73e9e6fac16b0be25cf3c20868c2145a844a7878ad9fc6998d9ff62d72f3a210bfa411ef03d3757b73d68a7c22926e874c421e51444d6
Frank-GER
pushed a commit
to syscoin/syscoin
that referenced
this pull request
Oct 13, 2023
…emote client disconnection 68f23f5 http: bugfix: track closed connection (stickies-v) 084d037 http: log connection instead of request count (stickies-v) 41f9027 http: refactor: use encapsulated HTTPRequestTracker (stickies-v) Pull request description: bitcoin#26742 significantly increased the http server shutdown speed, but also introduced a bug (bitcoin#27722 - see bitcoin#27722 (comment) for steps to reproduce on master) that causes http server shutdown to halt in case of a remote client disconnection. This happens because `evhttp_request_set_on_complete_cb` is never called and thus the request is never removed from `g_requests`. This PR fixes that bug, and improves robustness of the code by encapsulating the request tracking logic. Earlier approaches (bitcoin#27909, bitcoin#27245, bitcoin#19434) attempted to resolve this but [imo are fundamentally unsafe](bitcoin#27909 (comment)) because of differences in lifetime between an `evhttp_request` and `evhttp_connection`. We don't need to keep track of open requests or connections, we just [need to ensure](bitcoin#19420 (comment)) that there are no active requests on server shutdown. Because a connection can have multiple requests, and a request can be completed in various ways (the request actually being handled, or the client performing a remote disconnect), keeping a counter per connection seems like the approach with the least overhead to me. Fixes bitcoin#27722 ACKs for top commit: vasild: ACK 68f23f5 theStack: ACK 68f23f5 Tree-SHA512: dfa711ff55ec75ba44d73e9e6fac16b0be25cf3c20868c2145a844a7878ad9fc6998d9ff62d72f3a210bfa411ef03d3757b73d68a7c22926e874c421e51444d6
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
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.
A revival of #19434 but with a few substantial changes.
The code is valuable but from my understanding, it is offset by the workaround code from #11593. So I have wrapped the code so it's only used for versions where the workaround does not apply. The original PR removed the workaround code but also led to failures in
test/functional/mining_getblocktemplate_longpoll.pyso I think this is not an option.Since
libevent2.2 is still not out and it's hard to say when that will happen I am putting this into draft mode.