RFR: 8330940: Impossible to create a socket backlog greater than 200 on Windows 8+ [v2]

jwilliams990 duke at openjdk.org
Tue Aug 12 15:29:22 UTC 2025


On Tue, 12 Aug 2025 04:41:51 GMT, Jaikiran Pai <jpai at openjdk.org> wrote:

>>> Hi @jaikiran, Thanks for your response. Any chance this could be included in Java 21 LTS in the future?
>> 
>> Backports to OpenJDK are managed by a separate "jdk-updates" project https://openjdk.org/projects/jdk-updates/. Members of that project decide which changes should be backported (and when).
>> 
>> For this specific change, Java 26 release isn't scheduled for this year. Once it is released and after that version sees usage in real world applications, only then enhancements like these are usually considered for backports. So it's too early to say if/when this change will be backported.
>> 
>> In the meantime, what would be good is if you suspect that the backlog limit is indeed causing this issue on your setup, you could try to use the 26 early access build which is available here https://jdk.java.net/26/ and see if that helps you get past this issue. Are you creating a ServerSocket (or some other construct?) with some explicit value for backlog that is greater than 200?
>
>> Hi @jaikiran, I am running Apache Tomcat on Java. Tomcat has an acceptCount set to a large number (greater than 200). I believe Tomcat creates a ServerSocket with the configured acceptCount, and I suspect that it is not effective on Windows for the reasons mentioned above in this trail.
> 
> Thank you for that detail, that's useful to know. Based on what you note and what's documented of that Connector attribute in Tomcat docs - https://tomcat.apache.org/tomcat-9.0-doc/config/http.html#Attributes_Standard%20Implementation_acceptCount, I think what you are suggesting is correct - the backlog might be playing a role here.
> 
> If you get a chance to try this with Java 26 early access build, that would be a good experiment to verify that this change does indeed help this use case (I'll be curious to hear how that test goes).

Hi @jaikiran - I have tested with the Java 26 early access build and can report that the issue i faced on Windows OS with TCP/HTTP connection refused errors even at relatively low concurrency has been resolved in this build. Thanks.

-------------

PR Comment: https://git.openjdk.org/jdk/pull/25819#issuecomment-3179832921


More information about the net-dev mailing list