HTTP Connection pool lock leak in multithread environment

[barintom]

Describe the bug?

We are experiencing a problem where temporary SSL/TLS handshake failures trigger a lock leak in Apache HttpClient’s connection pool (it is bug in HTTPCLIENT-2399 -> https://issues.apache.org/jira/browse/HTTPCLIENT-2399). When an SSL error occurs during connection creation/closing, the pool lock may not be released. After that, other threads entering the pool simply wait for a connection that never becomes available, even though the pool should have free capacity.

This makes the connection pool unusable until the application is restarted. A single SSL handshake failure is enough to cause the issue.

What is expected to happen?

With SSL issues

What is the actual behavior?

Thread cannot get connection from pool even when it's full.

Reproduction Steps?

Use connection with SSL setup not trusted by called

Additional Information?

This is known bug and requires to bump HTTPClient to 5.5.1

Java Version

JDK-21

SDK Version

Any

OS version

Linux