Without logs it's hard to tell for sure, but you should take a look at the Tigase statistics (you can utilize Tigase Monitor) for possible bottlenecks.

Given that you are both registering and archiving messages it's possible that maybe database is a bottleneck.

Thank you for response me.

I use mysql5.6 as database server,here's my tigase log and the configuration of mysql .You said that monitor tigase ,if use http to monite or not?

I use tsung send 500000 users to tigase cluster,but at the end ,only half user regeist on mysql ,I adjust 50user per second to be send to tigase,this must be a slow speed ,can you help to analysis it ,thank you.

You are using excessive debuggin options (i.e. @--debug=server,xmpp.impl,db,cluster@) which will definitely impact performance. Please remove this line.

You can monitor the server using the monitor which I linked in previous message - https://projects.tigase.org/projects/tigase-monitor

Thank you for your help ,I finally find the reason of my problem,it is my env of mysql is virtual machine,the speed of IO is not good .But I found another problem ,it is I use tsung send 500000 and 800000 user to tigase cluster ,only half user created in mysql, if I have something wrong of the configuration?

hurry wang wrote:

But I found another problem ,it is I use tsung send 500000 and 800000 user to tigase cluster ,only half user created in mysql, if I have something wrong of the configuration?

The main question is whether you are load-testing Tigase XMPP server or your MySQL installation. If you want to load test Tigase server it's better to create user accounts in the database prior to load testing Tigase XMPP Server.

This thread is considered closed now. If you have additional questions please user our forums: https://projects.tigase.org/projects/tigase-server/boards