Reactor2TcpClient does not remove TcpClient instances after connection closed [SPR-14231]

[spring-projects-issues]

Rossen Stoyanchev opened SPR-14231 and commented

This was first reported in Boot 5810 where there is a detailed description.

---

Affects: 4.2.5

Issue Links:

• Reactor2TcpClient leaks threads on shutdown [SPR-14229] #18803 Reactor2TcpClient leaks threads on shutdown

Referenced from: commits 220711d, d48eeb2

[spring-projects-issues]

Rossen Stoyanchev commented

Resolving for now since I verified the fix locally but I'm expecting a confirmation under the Boot ticket and will reopen if necessary.

[spring-projects-issues]

dave burns commented

Wasn't sure if I should comment on the original boot issue or open a new ticket but the issue (or similar issue) still occurs for a broker/server receiving simultaneous connections. I've modified the sample project and added a multi-threaded Junit method as a client. This launches 2000 concurrent threads, about 1100 successfully connect to the server (on my setup). I don't know if its the load or the failed connections (rather than clean disconnects) but the org.springframework.messaging.tcp.reactor.Reactor2TcpClient can be seen to be leaking following by the same instructions as in the original boot issue.

Weird thing is the original leak showed both the Reactor2TcpClient and StompBrokerRelayMessagHandler leaking, this just shows the reactor 2 client.

Noticed this in our production application, i.e. system accepting concurrent connections. The multi-threaded harness mimics this behavior.

[spring-projects-issues]

Rossen Stoyanchev commented

hi it's fine to open a ticket here but could you create a new one since this one is already closed and released.

[spring-projects-issues]

dave burns commented

Thanks, have created new ticket