Hi guys,
Sorry for adding this discussion here as i know it does not really belong in this forum, just i added it to the clustering discussion a few days ago which does not seem to be too active, If some one could help me out on the following problem or point me in the right direction that would be great.
I was following instructions in the documents section on how to enabled clustering but im not sure if its working correctly?
PC1 has openfire 3.4.5 enterprise installed with Server Name say pc1,
PC2 has openfire 3.4.5 enterprise installed with Server Name say pc2.
Both servers use the same mySql database, PC1 uses localhost and pc2 the IP address of PC1 and both connect correctly.
I enable clustering on both nodes and they see each other on the
clustering overview page, also creating users and logging them in
allows the users sessions to be viewed from PC1 and PC2 openfire
install. I specify to the users that their Jabber address is pc1. Now
while PC1 is running the users will log in ok and i see thier session
from pc1 and pc2, if i stop pc1, the users will log out and i can no
longer see pc1 in the pc2 cluster overview. Also the users can not
login again untill pc1 is restarted, should they not be able to log in
using the pc2 node?
Does it matter what jabber client i use, should all clients be able to support clustering?
Best
Mark
Hey Mark,
Have you checked the document How to enable clustering in Openfire Enterprise??
Regards,
-- Gato
It seems to me the clients are only trying to connect to pc1,
If you put pc1 in thier configuration then that's where they will try to connect. What I would rccomend is setting up a round robin DNS entry for pc1 and pc2, that's the one that will go into the clients, this will both distribute the load and allow the automatic failover you are looking for.
Hi Gato and Mluich,
Have you checked the document How to enable clustering in Openfire Enterprise??
Yes i have read this article and have also tried making the server names the same, example both pc1, this resulted in the same behavior as i mentioned before.
The end of the article mentions Load Balancing but did not really elaborate much on it, Is this what controls the failover? How would this be implemented?
DNS round robin is a good solution but as far as I know it will not detect when a server is down so it will still redirect traffic to "dead" IP addresses. However, for basic testing this solution should be fine.
Thanks for this solution but failover is what im after, is there anyting else i could try?
Also i was looking at the article explaining how to load test clustering, when i installed the userCreation.jar and used the default settings to create users, just adding a user alias of usrLoadtest i was given an error of...
"Number of users per roster should be greater than total number of users. Number
of users per roster plus one should also be a multiple of total number of
users"
Other varaitions in number attempts resulted in same error, could someone shed some light on this as it does not make much sense to me? Prob something simple anyway.
Regards
Mark