[frs-20] Why does the system sometimes lock up, or behave nondeterministically after killing a thread?

When killing a request or thread, FusionReactor first tries to "soft-kill" it (which gives it the opportunity to end safely) but sometimes requests will not respond and we have to forcibly kill it using the most effective, non-blocking method available to us within the framework of the Java Virtual Machine. However, this method can theoretically leave the system in an inconsistent state. The chances of this happening are very low, but the system can be recovered by restarting ColdFusion MX. In practice we have never seen this case arise during development and testing.

Issue Details

Type: Technote
Issue Number: FRS-20
Components: Request Managment, Thread Management
Environment:
Resolution: Fixed
Last Updated: 14/Jun/07 3:31 PM
Affects Version:
Fixed Version:
Server:
Platform:
Related Issues:

Comments are closed.