The issue manifests in our StopContextTimeoutTest running under Windows. In this tests we stop a container while it is receiving tens of thousands of parallel requests. So the process of stopping/removing takes a very long time, and sometimes the test framework's start.stop.timeout kicks in and deletes the contexts while the server is stopping.
Tomcat HostConfig.checkResources sleeps for 500ms in a synchronized method if a resource is not found. This sleep delays the stop of the container and can cause the errors seen in our StopContextTimeoutTest and reported previously in MODCLUSTER-782.
After some discussion with rhn-engineering-jclere he opened PR-641 that is still not merged either in the tomcat-10 or in the tomcat-9 branches.
- blocks
-
JWS-2881 Release Notes for 5.8
-
- Closed
-
- clones
-
JWS-3092 Race condition in stop under stress: Context checks block Container stop
-
- Closed
-
- is cloned by
-
-
- Resolved
-
- is incorporated by
-
JWS-3241 Upgrade mod_cluster to 1.4.5.Final
-
- Closed
-
