Lucee Websocket 3.0.0.20-Snapshot connection closed at 60s

dev
1

The websocketInfo() lists the idleTimeout at 300 (as seen in the dump). But when I was tinkering with it, the timeout happens at 60. I am not sure if this is a browser issue or websockets ignoring the websocket.json file it itself created.

I looked into websockets disconnect time for browsers and both firefox and chrome, and what I could find, says that there is no set timeout, and usually chrome will not close a websocket for several minutes after the window goes inactive or is covered.

2

I’ve dug into this and the extension is actually applying idleTimeout correctly to every session — the WebSocket endpoint is per-connection (Tomcat creates a new instance for each), so every connection gets the configured 300s timeout set via setMaxIdleTimeout().

I confirmed this with a CI test that connects two clients and checks maxIdleTimeout on both sessions — both report the correct value on Tomcat 9 and 11.

https://luceeserver.atlassian.net/browse/LDEV-6219

Tomcat’s own default maxIdleTimeout is 0 (no timeout), so the 60s cutoff isn’t coming from there either.

The 60s is almost certainly something in front of Tomcat. Common culprits:

  • nginx — proxy_read_timeout defaults to 60s. For WebSocket you need:
proxy_read_timeout 300s;
proxy_send_timeout 300s;
  • Apache mod_proxy — ProxyTimeout defaults to 60s
  • HAProxy — timeout tunnel defaults to whatever timeout client is
  • Cloudflare / AWS ALB / etc — each has its own idle timeout

Can you check what sits between the browser and Tomcat in your setup? If you’ve got nginx or Apache as a reverse proxy, that’s almost certainly where the 60s is coming from.

3

That is what it would be then. I have Apache reverse proxying my websocket connection, and I did not change the settings as I just got it working. It was a frustrating weekend trying to get websockets working and just chalked up the issue to the snapshot given the other errors and issues I had (noted on the forum here).

4

I hate it when that happens!, anyway thanks for taking the time to flag to this problem, I have updated the docs to help others facing the same problem