Network created AMQ-6978:
----------------------------
Summary: Double dequeuing in a case of network partial
Key: AMQ-6978
URL: https://issues.apache.org/jira/browse/AMQ-6978
Project: ActiveMQ
Issue Type: Bug
Components: activemq-leveldb-store
Affects Versions: 5.15.3
Reporter: Network
We set up a cluster of 3 ActiveMQ brokers (Replicated LevelDB Store). other 3
machines were used to run Zookeeper cluster. We configured ActiveMQ as the
following:
* Replicas: 3
* sync: quorum_disk
We noticed that multiple clients will dequeue the same message in a case of a
network partition. This happens as the following:
# one client create a queue and sends a message.
# Network partition such that the master broker will not be able to
communicate with all other nodes (including Zookeeper nodes)
# one client consumes the message from the master successfully (before the
master detects that it does not have a majority of nodes and steps down).
# after electing new master by the other two nodes, another client consumes
the same message successfully
We think that is a bug and should not happen. Please advice if there is
something that we are missing in configuration and setting up the cluster.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
