Hello, i'm having some trouble with my ejabberd 2.1.7 and muc rooms.
I'm trying to get 600 anonymous users to join a MUC room with Tsung. But Ejabberd is unable to process the request as fast as they are arriving, (and to process everything).
So far I have a 60~ users bottleneck in a single room.
The first users join super-fast until 20-30, then 1/0.5sec then 1/sec then 1/2sec... etc till the servers crashes around 200users in the same room with the error:
Slogan: eheap_alloc: Cannot allocate 912262800 bytes of memory (of type "heap").
- messages are routed with a few seconds delay (5-10 seconds and more, as users approach 60+)
- new web clients are unable to connect, when hitting these limits
- these anomalies are witnessed even if we connect to this muc room through a typical jabber client to port 5222, like Adium.
- the beam process is taking 99% of the cpu usage
- when experiencing this slowdown in one room, if we create or join another room with very few users via web client or Adium, there is _no_ latency on messages or joins -- everything behaves as normal
Min presence interval is set to 0 and my max_fsm_queue to 1000.
I have upped the ulimit, the ERL_MAX_PORTS & ERL_PROCESSES, kernel polling is set to true.
The server running ejabberd is an Intel(R) Xeon(R) CPU E5507 @ 2.27GHz with 2GB RAM.
Using the same server I can successfully make 10.000 users to connect in 60sec without doing anything else after and it holds. But the CPU is at 100% and the CPU load ~0.25
Any ideas ?
Almost Solved: Note to self
Almost Solved: Note to self LogLevel 5 is a BAD idea.
I am also facing the similar
I am also facing the similar problem.
Are you using default ejabberd configuration or customized one? It would be nice to see it and as well as to hear how configuration (ejabberd, os and etc..) was tuned to handle such load.
Thank you!
I'm using the base
I'm using the base configuration in ejabberd.conf , i just activated the mod muc and upped everything I could in ejabberdctl.conf. I've tried with 600users speaking at the same time in a room with a 8Core 7gb ram. And it is still bottlenecking. The servers holds but the amount of messages is making the queue grow and messages are delayed for 5-20min before they are send when users are joining and speaking at the same time.