Hi everyone,
New here. We have a problem with the syncronization of user’s presence using OpenFire clustered.
Our test case is as follows:
- We start both openfire nodes in clustered mode and subscribe 20 users, next, we take a look to the sessions in the openfire admin console
in node 2, the picture is as follows:
- We stop the openfire windows service in the node 1 in order to simulate a failover and force the traffic is redirected to the node 2.
We take a look to the sessions in the openfire admin console in node 2, the picture is as follows:
- We start the openfire windows service in the node 1 in order to simulate that the node 1 has come back to life and we wait for hazelcast plugin share
the memory, then, we take a look to the sessions in the openfire admin console in node 1, the picture is as follows:
As you can see, in the node 1 the presence of all users is marked as “offline”, but in the node 2, the presence of all users is correct.
We have the feeling that the hazelcast plugin is not working as expected, we don’t know is due to a bug or because of misconfiguration or maybe this
is the right behaviour … but we think that after sharing the memory, in the node 1 we have to see the same picture which we see in the node 2 …
Anyway this behaviour is random … sometimes the presence is correctly syncronized and other times not …
The configuration of our environment is as follows:
- Windows Server 2012 R2 8 GB RAM:
- OpenFire 4.0.1
- Hazelcast Clustering Plugin 2.2.0
- Presence Service 1.7.0
The configuration of both openfire nodes (we have two) is as follows:
-
openfire.xml
-
openfire system properties
-
Hazelcast configuration file
Is attached.
Kind Regards
hazelcast-cache-config.xml.zip (3509 Bytes)