Pods not destroyed till 3 - 4 days

ShinyProxy
1

Hi All,
I have a shinyproxy running on EKS. As far as I understand there should not be a way for an application pod to be alive till 3 or 4 days unless someone is working on it for this period of time.

my config is a below
proxy:
title:
logo-url: <>
port: 8080
container-wait-time: 60000
container-log-path: /opt/shinyproxy/container-logs
authentication: keycloak
keycloak:
realm:<>
auth-server-url: <>
resource: <>
credentials-secret: <>
ssl-required: none
admin-groups: admins
container-backend: kubernetes
kubernetes:
internal-networking: true
image-pull-policy: Always
specs:
- id: assortment_tools
display-name: Assortment Tools
description: Assortment Tools include Sortimentsrahmen as well as several other standart reportings to help improve our assortment
container-cmd: [“R”, “-e”, “shiny::runApp(’/root/’, host=‘0.0.0.0’, port=3838)”]
container-image:
access-groups: <>
kubernetes-pod-patches: |
- op: add
path: /spec/volumes
value:
- name: redshift-connection
secret:
secretName: <sec_name>
- op: add
path: /spec/containers/0/volumeMounts
value:
- name: <>
mountPath: /root//cred
logging:
file:
/opt/shinyproxy/logs/shinyproxy.log

Please guide me if I understand wrong?
If I stay idle for sometime or if I logout the container is deleted, I see only one way the container is alive is use closes the tab and the proxy has no idea about it.

2

Hi

Indeed, with this configuration, the apps/pods should automatically terminate when the user closes their browser tab.

Can you try these two steps:?

  1. enable debug logging for the heartbeats of ShinyProxy, by adding the following block to the configuration. Make sure to add the logging at the same level as the proxy property (so not below proxy).
logging:
    eu:
      openanalytics:
        containerproxy:
          service:
            hearbeat: DEBUG
  1. configure the admin panel (see the admin-groups property https://shinyproxy.io/documentation/configuration/#general ) . Then open the admin panel and look for the apps in the table that will not get terminated.

If the apps keep receiving heartbeats, this means that the user did not close their browser tab. If you don’t see any heartbeats, and the app stays alive, we might have a bug.

3

Hey @tdekoninck ,

I think we do have a bug. We are currently running the shinyproxy in kubernetes cluster where each application launched belongs to the replica set. Observation in this particular is that whenever a new replica set is created when patched shinyproxy. The previous replicaset stays till it has a websocket but when a user closes browser on a running application the replica set gets destroyed and the application pods still running as there is no running hearbeat.