Ensure https proxy connections don't clog up the ConnectionPool #580
Conversation
This file contains 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
Fixes #562. Despite lots of back and forth on this issue, it turns out there's a simpler diagnosis & fix that worrying too much about when sockets will throw when `close`ing (which we should certainly get all cleaned up one day). The _real_ issue here is in `ConnectionPool.sslupgrade` which was effectively swapping our `Connection` that we formally "acquired" from the connection pool with a new one that had a https-ssl-upgraded `SSLContext` to do the request tunneling. The issue with just swapping the `Connection` is that the original connection was never properly "released" to the connection pool, so connections were acquired, but never released, and eventually we hit the `connection_limit`. The proposed solution here (which works locally to avoid getting hung up on the connection pool clogging), is to introduce a new `kill!(c::Connection)` method that "releases" a connection without returning it to the connection pool. So we signal to the pool that a connection has finished, but it's not put back in the queue to be re-used, which is what we want. I should note that this does mean that proxied https requests then don't really respect the `connection_limit`, but also note that it never has due to this "swapping".
Codecov Report
@@ Coverage Diff @@
## master #580 +/- ##
==========================================
+ Coverage 75.94% 77.60% +1.66%
==========================================
Files 37 37
Lines 2257 2434 +177
==========================================
+ Hits 1714 1889 +175
- Misses 543 545 +2
Continue to review full report at Codecov.
|
|
Thanks a lot for the fix! |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.
Fixes #562. Despite lots of back and forth on this issue, it turns out
there's a simpler diagnosis & fix that worrying too much about when
sockets will throw when
closeing (which we should certainly get allcleaned up one day).
The real issue here is in
ConnectionPool.sslupgradewhich waseffectively swapping our
Connectionthat we formally "acquired" fromthe connection pool with a new one that had a https-ssl-upgraded
SSLContextto do the request tunneling. The issue with just swappingthe
Connectionis that the original connection was never properly"released" to the connection pool, so connections were acquired, but
never released, and eventually we hit the
connection_limit. Theproposed solution here (which works locally to avoid getting hung up on
the connection pool clogging), is to introduce a new
kill!(c::Connection)method that "releases" a connection withoutreturning it to the connection pool. So we signal to the pool that a
connection has finished, but it's not put back in the queue to be
re-used, which is what we want. I should note that this does mean that
proxied https requests then don't really respect the
connection_limit,but also note that it never has due to this "swapping".