Details
-
Bug
-
Status: Closed
-
Critical
-
Resolution: Fixed
-
0.5
-
None
Description
Adding a new broker to a cluster while there are clients using not-acquired mode (aka "browse" mode) to retrieve messages without dequeing them can cause inconsistencies that will cause cluster members to abort. Examples of errors that may be seen as a result of this bug:
2009-12-18 10:29:04 critical cluster(20.0.100.32:23782 READY/error) local error 906715 did not occur on member 20.0.100.33:16960: resource-limit-exceeded: Policy exceeded on A_1, poli\
cy: size: max=104857600, current=952489; count: max=20000, current=20000; type=reject (../../cpp/src/qpid/broker/QueuePolicy.cpp:83)
2009-12-21 12:22:07 error Error delivering frames: local error did not occur on all cluster members : invalid-argument: anonymous.a6e88ed1-e69b-4549-b\
32d-5168f068974d: confirmed < (103+0) but only sent < (99+0) (../../cpp/src/qpid/SessionState.cpp:151) (../../cpp/src/qpid/cluster/ErrorCheck.cpp:89)
See https://bugzilla.redhat.com/show_bug.cgi?id=541927 for more details.