[ARTEMIS-2991] Leak of HttpAcceptorHandler instances when using websocket connections

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 2.9.0, 2.10.0, 2.10.1, 2.11.0, 2.12.0, 2.13.0, 2.14.0, 2.15.0, 2.16.0
• Fix Version/s: 2.17.0
• Component/s: Broker
• Assignee: Unassigned
• Reporter: Luis Miguel De Bello
• Created: 13/Nov/20 12:58
• Updated: 09/Feb/21 21:23
• Resolved: 18/Nov/20 21:08

Description

We are running Artemis 2.9.0 and we noticed the memory grows continuously after getting a heap we discovered there are a leak of multiple instances of type org.apache.activemq.artemis.core.remoting.impl.netty.HttpAcceptorHandler which are keep in a list inside org.apache.activemq.artemis.core.remoting.impl.netty.HttpKeepAliveRunnable.

 

Those instances are removed in the channelInactive method but this method is not called because when doing the upgrade to websocket the handler is removed.

 

Proposals:

1) In the upgrade clean up the reference for the HttpAcceptorHandler before removing the handler.

2) If option 1 does not work because that threads also send the keep alive for websocket, I could use the channelInactive from websocket handler to later unregister the handler

Issue Links

• links to: GitHub Pull Request #3339