Open sambles opened 2 years ago
Update: testing by @RolandASc points to a connection problem internal to the cluster. External UI apps connecting via HTTPS don't show this problem.
Follow up is to look at the ingress nginx configuration
Something is wonky with the websocket connection
https://www.civo.com/learn/using-websockets-with-ingress-controller https://stackoverflow.com/questions/52883653/nginx-ingress-controller-websocket-support#52884066
The above happens with firefox after an connection Upgrade request to the websocket:
x.x.x.x - - [15/Sep/2022:08:47:56 +0000] "GET /websocket/ HTTP/1.1" 101 8387 "-" "Mozilla/5.0 (X11; Linux x86_64; rv:103.0) Gecko/20100101 Firefox/103.0" 651 262.311 [default-oasis-ui-8080] [] 10.240.0.21:3838 0 262.312 101 eada5b797fb74856da7abec2d04afc42
10.240.0.30 - - [15/Sep/2022:08:48:01 +0000] "GET /auth/realms/oasis/protocol/openid-connect/userinfo HTTP/1.1" 200 226 "-" "python-requests/2.26.0" 1773 0.007 [default-keycloak-8080] [] 10.240.0.12:8080 226 0.008 200 913bbd6c200667a9dddd7859bac7dc15
Seems similar to, the following and only happens when the Shiny app is behind a proxy server. https://github.com/rstudio/rstudio/issues/2940 https://github.com/rstudio/shiny/issues/2925
hello,
could you think about using traefik instead of nginx ?
idle session crashes also behind a ALB on AWS.
happy to help out on this issue if needs be?
Hi @dholdaway thanks for the pointer, i'll give traefik a try. My other thought was placing shinyproxy between the UI application and ingress.
Darren has had luck switching ingress to traefk. Rodney to try
Issue Description
An authentication issue with keycloak causes the UI app to crash after a short period of time ~ 3-4 mins
Steps to Reproduce (Bugs only)
Version / Environment information
UI 1.11.3 Newer platform build (platform-2.0)