Red Hat Bugzilla – Bug 990494
Netty worker threads consume all cpu time after deployment and undeployment of service with udp binding
Last modified: 2015-08-02 19:44:10 EDT
Created attachment 781034 [details]
Description of problem:
After deployment and undeployment of archive with switchyard service, netty worker threads consume all cpu power.
Example project is attached.
I can confirm this behavior locally and I also see netty threads consuming the CPU. We need to dig into this deeper, but I would vote against this being a blocker as deployment still works and the server is still able to shutdown cleanly if the user gets into a situation where undeploy pegs the CPU. Undeploying the app and bouncing the server results in a clean undeployment. Obviously not ideal, but at least the server is left in a consistent state.
BTW, I do not see this behavior with Netty TCP bindings, so the impact here is limited to UDP bindings only.
Keith, I've seen this behaviour with TCP bindings too, but just once and cannot reproduce it now. So there may be issues with Netty regardless of TCP/UDP choice.
Keith Babo <firstname.lastname@example.org> made a comment on jira SWITCHYARD-1633
tried several times, seems to be fixed, marking as verified in ER7-2
Keith Babo <email@example.com> updated the status of jira SWITCHYARD-1633 to Closed