HttpClient SSL Race Condition

Daniel Fuchs daniel.fuchs at oracle.com
Sun Feb 3 11:10:35 UTC 2019


Hi,

It looks like this is JDK-8214418 - which has been fixed
in 12.0.1 b03 and 13-ea b04. The issue was with the
half closed semantics of the SSL engine in TLS 1.3.

This is only loosely related to
https://bugs.openjdk.java.net/browse/JDK-8217094
(we found that while testing the fix to JDK-8214418)
but JDK-8217094 is also fixed in 13-ea b04.

If you're testing on 13-ea, could you make sure to test on
a recent build >= b04?

best regards,

-- daniel


On 03/02/2019 03:38, August Nagro wrote:
> Hello,
> 
> When using multiple threads to send HttpRequests from a static final 
> HttpClient instance, I noticed that CPU use spikes after a minute or so.
> 
> I created a reproducible example that I've tested on Java 11.0.2 and 
> 13-ea: https://gist.github.com/AugustNagro/fec5f7895bd8bb5db66f2ed9cfafc35c
> 
> This bug never occurs when using a single thread. Also, switching TLS to 
> 1.2 or 1.1 does not fix the issue. Looking at VisualVM it seems that 
> SSLFlowDelegate is entering some sort of race condition.
> 
> Is this a bug in java.net.HttpClient, or should I not be accessing the 
> HttpClient from multiple threads?
> 
> Regards,
> 
> August Nagro
> 
> PS: Could this be related to (fixed) 
> https://bugs.java.com/bugdatabase/view_bug.do?bug_id=8217094?



More information about the net-dev mailing list