Hi all.
First of all, sorry for my english.
We have deployed an Openfire XMPP server, initially using a standalone solution on Amazon.
In production enviroment we have installed and have running the version Openfire 3.10.0, with this plugins:
- Clustering Plugin 1.3.0 Jive Software
- DB Access 1.1.0 Daniel Henninger
- Email Listener 1.1.0 Jive Software
- Monitoring Service 1.4.2 Jive Software
- Presence Service 1.6.0 Jive Software
- Search 1.6.0 Ryan Graham
- User Import Export 2.4.0 Ryan Graham
- User Service 2.0.2 Roman Soldatow, Justin Hunt
The infraestructure is:
- One server m3.xlarge, where openfire is running
- One database:
- SQL Server SE 10.50.2789.0.v1
- Instance Class db.m1.large
- Storage TypeStorage Magnetic
- IOPSFixed disabled
- Storage 300 GB
We currently have two major problems:
- Suddenly Openfire just degraded and lost client connections (from 6,6k to 2k approx.)
- When it restarts, suddenly it does not accept all connections from existing customers (6,6k)
To neutralize this we are preemptively restarting the server every 15 days, but every time we encounter the problem of concurrency.
We believe that the current instance has a theoretical limit of 15k connections, does it?
In addition, for the expected growth we have tried to configure and deploy into production with the version 3.11.0 using Hazelcast for clustering with 2 nodes (2 * m3.xlarge). The problem is that this solution is more unstable than the current standalone. This kept operating eight hours the last time we tried it and then fell.
We would like to know if somebody has experienced similar problems using Amazon platform or this volumetric of connections (currently near to 7k).
Thanks in advance.
Best regards.