0
votes

I have pubsub related problem on @ejabberd 17.04 My Cluster Configuration Is 2 Servers 16 GB + 4 Core each on AWS

My pubsub config is as follows 4 Pub Sub Nodes ( flat type )

Our concurrent pubsub load is as follows 20000+ Concurrent users connecting and subscribing on those 4 nodes.

And the problem is as follows Each pubsub blocking call takes infinite time to complete With smackexception$noresponseexception

I also tried from websocket and it happens to be having very long response time for pubsub subscribe or publish calls. The same issue arises when we try to create pubsub nodes using xmpp client.

MUC works fine. What could be wrong ? Any help ? How to trace this ?

1

1 Answers

0
votes

There are 2 things you can do to start with.

  1. Put entry and exit level logs and figure out where it is taking time.
  2. Increase value of IQDisc parameter in mod_pubsub.erl file. Parameter sets number of process to process particular time of IQ request. Since you have concurrent load for pubsub request, increase it to 20-30(from my experience) and give it try.