i migrated my site to the new Websocket Plugin using Openfire 3.10.2 and stanza.io on the client side. The same setup works fine using BOSH. After switching to websockets the whole http-bind resource becomes unresponsive after ~ two hours uptime accepting no more new connections (neither BOSH nor Websockets). At this time there were only around 200 active users. I’m using nginx to proxy the websocket connections to openfire using this setup NGINX as a WebSocket Proxy - NGINX
Any thoughts on this issue?
Bump! Does anyone have experiences with the new websocket plugin in a production environment?
I will be looking into this issue over the next few days, although my client and proxy configurations are slightly different than yours. Are there any errors in your server log files or other evidence of dropped connections, etc.? Also, it would be great if you would be willing to share a thread dump from your server when it gets locked up.
Thanks for your reply Tom. The only way to reproduce this for me at the moment is to enable websockets again on the production system which i would like to avoid. I will try and see if i can generate some load on the test system to get the logs. Is the plugin in general compatible with OF 3.10.2? Are you using nginx as well on your system?
Hi Tom, just as an update on this… I have now migrated my system to OF 4.0.2 with the Websocket plugin 1.1.3 and everything is running stable with a couple of thousands concurrent connections.