Users randomly disconnect

Openfire Openfire Support
1

Hello everyone,

We have been using Openfire in our department for a few months with no problems, basically just testing it out and setting it up. Recently we switched everyone in the company over to the Openfire server (Potentially about 180 office employees, but average of 80 connect at one time), and since then users randomly disconnect. Not just every once in a while, there will be a disconnect spree, where about 10-20 people disconnect then reconnect. This coincides with a high packet per minute (5000-14000 packets per minute).

Here is our setup:

Openfire 3.6.4 on SLES 10 SP2

Plugins: Broadcast, Client Control, Monitoring Service, Presence Service, Search

LDAP authentication (eDirectory server local to the Openfire server)

3 locations, one server (connected via T1)

Here is what we have tried so far:

Limit types of clients that can connect (only Pidgin, Spark and Adium)

Upgrade all clients to Pidgen 2.5.5

We removed some of the plugins (Monitoring Service) but it didn’t help.

Changed Java from IBM Java 1.5 to Sun Java 1.6 and increased Java cache.

This isn’t a lot of users, and I can’t imagine why we would need a server at each location, the old jabberd server worked just fine from one location.

If anyone has any other ideas on what to look for, please let me know. We are getting ready to either switch back to the old jabberd server or look into a different solution. I really like the management and functionality of Openfire, so I would love to solve this problem.

Thanks in advance,

Travis

2

We are still having user disconnection problems, This really seems strange.

Here is a little more information:

MySQL back end running on a separate server, db load very low.

This just seems like a packet problem, when to many packets are received (it really doesn’t seem like that many, 7000-14000 ppm) and over 70 users are online, we get random disconnects. That isn’t very many people, and the hardware is sound (Intel Xeon Processors, 4GB memory):

95% of the users are using Pidgin 2.5.5 or 2.5.6. A couple are now using Spark. The disconnects do not seem to happen when running Spark.

It seems like quite the stretch that the “fix” for the server problem is to switch everyone to the Spark client. Has anyone else experienced this? I doubt the users will want to switch from Pidgin (you know how people hate change), especially when the old jabberd2 server was working just fine (albeit from an old samba_tng domain running on an old bsd server). With the management headaches of the old jabberd2 server, I would love to get Openfire/Pidgin working as we have dedicated a consider amount of time getting it up and running on our new domain.

Does anyone have any more ideas before we axe this thing for something else?

Thank you very much for your assistance.

Travis

3

I’ve the same problem: users randomly disconnects and high packet traffic (from 5000 to 12000 ppm). This is my configuration:

Linode VPS Server with Debian 5

16GB Disk

512 Mb Ram

Openfire 3.6.4

Proprietary client

The problem started when i reached 100 users (with more than 70 online concurrently). I cannot change the client so i need to understand if there is another kind of solution. Otherwise i must switch to another xmpp server because my users are complaining. I’m sorry for this because i like very much this software but reliability is my first condition.

Thank you

4

It seems that for the author of this thread switching to Spark has fixed the problem, at least it looked like it helped. There were some issues with previous Pidgin versions, like it wasn’t sending heartbeat packets like Openfire was expecting and then server was disconnecting clients for being idle.

You can first of all try this with Spark to determine is it really related to a client. Also you can try changing xmpp.client.idle system property to some huge value (it is in miliseconds), so it won’t disconnect clients for being idle. Also it would be great to get some logs here from the server.