websockets behind apache somehow limited?

114 views
Skip to first unread message

Tom Eicher

unread,
Jul 28, 2023, 7:11:06 AM7/28/23
to WildFly
Hello,

I have s standard websocket endpoint

@ServerEndpoint("/mySocket")
...
@OnOpen
public void open(Session session) {
session.setMaxIdleTimeout(WS_IDLEMTIMEOUT_MS);
...


running on (still) Wildfly 26.0.1 on http port 8888

behind an Apache 2.4.51 with a
<VirtualHost *:443>
ProxyPass /mytSocket "ws://localhost:8888/mySocket" upgrade=websocket
ProxyPass / http://localhost:8888/


and it's working nicely!
But somehow I noticed I never see more than a hundred connected
clients, even if I know more than 400 clients are online
(client is an android app).

And now I found in my apache log "bursts" of
x.x.x.x - - [20/Jul/2023:00:01:07 +0200] "GET /mySocket?param=0119 HTTP/1.1" 408 - "-" "-"

and 408 is of course timeout.
But the server is not really loaded, and the wildfly is relaxed.
So maybe it means we are waiting for a free slot / connection or something?

sometimes (often enough) it works, though
x.x.x.x - - [20/Jul/2023:22:57:35 +0200] "GET /mySocket?param=0119 HTTP/1.1" 101 - "-" "-"

maxconnections in wildfly is not limited at all, apache is running
as prefork with
ServerLimit 650
MaxRequestWorkers 650
MaxRequestsPerChild 10000


I am not sure about the websocket tunneling now working differently
since Apache 2.4.47. (proxy_wstunnel module no longer required?)

I also suspect we have a higher server CPU load since one of the
last Apache upgrades, but can't nail it down.

so I would like to ask for any ideas what to check, where
to tune, ... any ideas welcome.

Thanks Tom.



Bartosz Baranowski

unread,
Jul 28, 2023, 7:17:08 AM7/28/23
to WildFly
Sorry, dont have much time to look through right now, but  something rings a bell: https://www.mastertheboss.com/jbossas/jboss-performance/wildfly-performance-tuning/  - does this help in any way?

Tom Eicher

unread,
Jul 28, 2023, 7:21:22 AM7/28/23
to WildFly
Hmm, I don't think so.
All other (regular) requests are going fine 100%, at the same point in time where the websockets get 408,
so it must be something specific to websockets I'd say...
Reply all
Reply to author
Forward
0 new messages