Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

K8s dashboard showing "no healthy upstream"; remove K8s dashboard links and services #1699

Closed
richardsliu opened this issue Oct 2, 2018 · 2 comments

Comments

@richardsliu
Copy link
Contributor

This is configured as:

// This service adds a rule to our reverse proxy for accessing the K8s dashboard.
local k8sDashboard = {
  local isDashboardTls = if params.cloud == "acsengine" || params.cloud == "aks" then
    "false"
  else
    "true",
  // Due to https://github.com/ksonnet/ksonnet/issues/670, escaped characters in
  // jsonnet files are not interpreted correctly by ksonnet, which causes runtime
  // parsing failures. This is fixed in ksonnet 0.12.0, so we can merge this back
  // to the jsonnet file when we take a dependency on ksonnet 0.12.0 or later.
  local annotations = function(isDashboardTls) {
    "getambassador.io/config":
      std.join("\n", [
        "---",
        "apiVersion: ambassador/v0",
        "kind:  Mapping",
        "name: k8s-dashboard-ui-mapping",
        "prefix: /k8s/ui/",
        "rewrite: /",
        "tls: " + isDashboardTls,
        // We redirect to the K8s service created for the dashboard
        // in namespace kube-system. We don't use the k8s-dashboard service
        // because that isn't in the kube-system namespace and I don't think
        // it can select pods in a different namespace.
        "service: kubernetes-dashboard.kube-system",
      ]),
  },

However there is no such service in the kube-system namespace:

ricliu@ricliu:~/kubeflow/kubeflow$ kubectl -n kube-system get services
NAME TYPE CLUSTER-IP EXTERNAL-IP PORT(S) AGE
default-http-backend NodePort 10.3.254.180 80:30631/TCP 12h
heapster ClusterIP 10.3.250.231 80/TCP 12h
kube-dns ClusterIP 10.3.240.10 53/UDP,53/TCP 12h
metrics-server ClusterIP 10.3.250.4 443/TCP 12h

@jcurlier
Copy link

jcurlier commented Oct 4, 2018

+1

@jlewi
Copy link
Contributor

jlewi commented Oct 9, 2018

Should we just get rid of the link to the K8s dashboard and let users take care of it using the appropriate solution for their K8s distribution.

This has never worked particularly well because the dashboard is not easy to authenticate to (on GKE you had to print an access token and then past it in). I also think the dashboard has been deprecated and removed on GKE and possibly others.

/area 0.4.0

/cc @swiftdiaries

@jlewi jlewi changed the title K8s dashboard showing "no healthy upstream" K8s dashboard showing "no healthy upstream"; remove K8s dashboard links and services Oct 9, 2018
jlewi added a commit to jlewi/kubeflow that referenced this issue Oct 17, 2018
* The K8s dashboard is not always installed; so just adding a link
  to it leads to confusion because it may not actually be present.

* Furthermore, even when the dashboard is running there might not be
  a convenient way to authenticate to it short of having users generate
  and paste access tokens.

Fix kubeflow#1699
k8s-ci-robot pushed a commit that referenced this issue Oct 17, 2018
* The K8s dashboard is not always installed; so just adding a link
  to it leads to confusion because it may not actually be present.

* Furthermore, even when the dashboard is running there might not be
  a convenient way to authenticate to it short of having users generate
  and paste access tokens.

Fix #1699
@carmine carmine added this to the 0.4.0 milestone Nov 6, 2018
kunmingg pushed a commit to kunmingg/kubeflow that referenced this issue Dec 6, 2018
* The K8s dashboard is not always installed; so just adding a link
  to it leads to confusion because it may not actually be present.

* Furthermore, even when the dashboard is running there might not be
  a convenient way to authenticate to it short of having users generate
  and paste access tokens.

Fix kubeflow#1699
saffaalvi pushed a commit to StatCan/kubeflow that referenced this issue Feb 11, 2021
* The K8s dashboard is not always installed; so just adding a link
  to it leads to confusion because it may not actually be present.

* Furthermore, even when the dashboard is running there might not be
  a convenient way to authenticate to it short of having users generate
  and paste access tokens.

Fix kubeflow#1699
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

No branches or pull requests

5 participants