On 4/5/15 5:23 AM, c.whalley wrote:
> On 02/04/2015 16:28, c.whalley wrote:
>> I've got pretty much the example from:
> I've just proven this is caused by the mod_caucho config by tweaking the
> setup so it uses port 8080 throughout, i.e. with the same deployed app
> but "pure" resin we're good to go.
> I've seen there is an apache mod_proxy_wstunnel which can upgrade the
> incoming connection to a websocket connection but so far still haven't
> been able to get this going, in any case why is this needed? I thought
> all incoming traffic on port 80 was forwarded to 8080 once a virtual
> server had the appropriate entries to do so for the servets etc in its
Because HTTP and websockets are very different, a HTTP proxy isn't
powerful enough to tunnel websockets messages. Basically, a websocket
proxy needs two threads, one to listen for server messages and a second
thread to listen for client messages. But HTTP only needs a single
thread to listen for server messages, because once the client POST is
done, there are no more messages for HTTP.
mod_caucho is designed around HTTP only (and Apache itself was designed
around HTTP only).
>> This runs fine in my local env with resin on 8080 and Apache on 80
>> (Win), same config on a Fedora 21 QA server but when deployed to
>> production, which uses mod_caucho, I see
> resin-interest mailing list
resin-interest mailing list