tsung, high error rate

4 posts / 0 new
Last post
tsung, high error rate


Testing via tsung is very good! I already spotted some mistakes, and I got no idea why it is like that.

Logs (loglevel = INFO) show a bunch of

2012-04-20 05:39:32 ClientConnectionManager.processSocketData() INFO: Hm, receiver is not set yet (misconfiguration error)..., ignoring: from=c2s@localhost/, to=null, DATA=CData size: 100, SIZE=213, XMLNS=jabber:client, PRIORITY=NORMAL, PERMISSION=NONE, TYPE=chat, connection:, type: accept, Socket: Socket[addr=/,port=33058,localport=443]


2012-04-20 04:33:23 SessionManager.closeConnection() WARNING: Found stale XMPPResourceConnection: user_jid=+391199309@xmpp.xgate.com.hk/tsung, packets=23, connectioId=c2s@localhost/, domain=xmpp.xgate.com.hk, authState=AUTHORIZED, isAnon=false, removing...

Which I can't really combine into any bell ringer. Maybe there is some place you can point me to given these ?

Thank you,


Actually some of errors like this are a result of combination of high debug logging with high traffic.
Normally, the way you do load tests is in a following way:

  1. You setup Tigase with test accounts and debug set to 'server'
  2. You login to a few test accounts from a normal XMPP clients to see if authentication works, contact list works, presence works, etc...
  3. You run Tsung on a very low traffic, up to 100 accounts with login rate no higher than 1/sec and look into logs if you can see any problem
  4. You switch debug log off completely and then run tsung at full speed.

Now, for the last step you need a way to monitor Tigase somehow to see how it copes with load. There are a few options for this:

  1. With Psi client you can pull server statistics from time to time.
  2. With JConsole you can pull server statistics via JMX
  3. You can write a plugin for statistics component to periodically dump statistics to a file or DB
  4. You could contact me via contact form and I could provide you with a trial version of our Tigase Monitor which can connect to running server and display various performance metrics


turning logs off seem to have done the trick. That Tigase Monitor may be of interest to us indeed. Is there any quick facts about its capabilities ?


The Tigase Monitor pull various metrics from the Tigase server or cluster and presents them as charts on the screen.
Please have a look at the article for screenshots example.