[webmvc-tomcat/CRaC] Graceful shutdown throws AsynchronousCloseException

[nekhtan]

Hello,

While the build runs fine, the build/jvmCrApp/output-checkpoint.txt shows:

2025-05-23T11:00:32.997+02:00  INFO 30863 --- [           main] o.s.b.w.e.tomcat.GracefulShutdown        : Commencing graceful shutdown. Waiting for active requests to complete
2025-05-23T11:00:35.010+02:00  INFO 30863 --- [tomcat-shutdown] o.s.b.w.e.tomcat.GracefulShutdown        : Graceful shutdown complete
2025-05-23T11:00:35.010+02:00 ERROR 30863 --- [auto-1-Acceptor] org.apache.tomcat.util.net.Acceptor      : Socket accept failed

java.nio.channels.AsynchronousCloseException: null
	at java.base/java.nio.channels.spi.AbstractInterruptibleChannel.end(AbstractInterruptibleChannel.java:223) ~[na:na]
	at java.base/sun.nio.ch.ServerSocketChannelImpl.end(ServerSocketChannelImpl.java:391) ~[na:na]
	at java.base/sun.nio.ch.ServerSocketChannelImpl.accept(ServerSocketChannelImpl.java:415) ~[na:na]
	at org.apache.tomcat.util.net.NioEndpoint.serverSocketAccept(NioEndpoint.java:518) ~[tomcat-embed-core-10.1.34.jar!/:na]
	at org.apache.tomcat.util.net.NioEndpoint.serverSocketAccept(NioEndpoint.java:71) ~[tomcat-embed-core-10.1.34.jar!/:na]
	at org.apache.tomcat.util.net.Acceptor.run(Acceptor.java:128) ~[tomcat-embed-core-10.1.34.jar!/:na]
	at java.base/java.lang.Thread.run(Thread.java:1583) ~[na:na]

I guess the checkpoint is not created properly then, right ?

The issue seems to appear starting from SpringBoot 3.4.x (tested in SpringBoot 3.3.3 and it looks fine there).

Note that the same issue can also be observed when upgrading SpringBoot on https://github.com/sdeleuze/spring-boot-crac-demo

Do you know how it can be fixed ?

[nekhtan]

Alright, after digging a bit more, it seems that it's not "broken": graceful shutdown is enabled by default since 3.4.0. If we revert this behavior the checkpoint is created without any exception:

server.shutdown=immediate

What's your take on this ? Should we use an immediate shutdown when creating the checkpoint and enable it back afterwards (is that even possible) ? Note that with the graceful shutdown enabled (throwing the exception), the checkpoint seems OK overall, but it's never reassuring to see such exceptions...

FYI, the stack trace leading to the AsynchronousCloseException: