Broker shutdown doesn't terminate active connections before closing MessageStore
--------------------------------------------------------------------------------
Key: QPID-871
URL: https://issues.apache.org/jira/browse/QPID-871
Project: Qpid
Issue Type: Bug
Components: Java Broker
Affects Versions: M2.1
Reporter: Martin Ritchie
Summary:
This may only be applicable to InVM situations. However, we need a high load
test to validate this.
When the broker is shutting down its internal processes are not stopped before
the MessageStore or indeed any other storage is closed.
The example test case is to run the TimeToLive.testPassiveTTL test with a very
low 50ms receive timeout. The timeout will fail, failing the test and so
killing the InVM broker. This shutdown process is similar to shutting down a
real broker. In the TimeToLive test this resulted in a NPE as the MessageStore
is closed. On a real broker it can manifest itself as a 'Failed to cleanup
message' error.
The cause is that on shutdown the ApplicationRegistry just closes the registry
and so the MessageStore, rather than causing the active connections to close
and stopping processing before closing the MessageStore.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
