add known issues for kubeconfig tokens

This commit is contained in:
kinarashah
2021-01-25 12:25:51 -08:00
parent ea91010810
commit a0fa02a665
2 changed files with 25 additions and 14 deletions
@@ -22,20 +22,30 @@ Run `kubectl cluster-info` or `kubectl get pods` successfully.
## Authentication with kubectl and kubeconfig Tokens with TTL
_**Available as of v2.4.6**_
_**Available as of v2.4.6**_
_Requirements_
If admins have [enforced TTL on kubeconfig tokens]({{<baseurl>}}/rancher/v2.x/en/api/api-tokens/#setting-ttl-on-kubeconfig-tokens), the kubeconfig file requires the [Rancher cli](../cli) to be present in your PATH when you run `kubectl`. Otherwise, you’ll see error like:
`Unable to connect to the server: getting credentials: exec: exec: "rancher": executable file not found in $PATH`.
If admins have [enforced TTL on kubeconfig tokens]({{<baseurl>}}/rancher/v2.x/en/api/api-tokens/#setting-ttl-on-kubeconfig-tokens), the kubeconfig file requires the [Rancher cli](../cli) to be present in your PATH when you run `kubectl`. Otherwise, you’ll see error like:
`Unable to connect to the server: getting credentials: exec: exec: "rancher": executable file not found in $PATH`.
This feature enables kubectl to authenticate with the Rancher server and get a new kubeconfig token when required. The following auth providers are currently supported:
This feature enables kubectl to authenticate with the Rancher server and get a new kubeconfig token when required. The following auth providers are currently supported:
1. Local
2. Active Directory
3. FreeIpa, OpenLdap
4. SAML providers - Ping, Okta, ADFS, Keycloak, Shibboleth
3. FreeIpa, OpenLdap
4. SAML providers - Ping, Okta, ADFS, Keycloak, Shibboleth
When you first run kubectl, for example, `kubectl get pods`, it will ask you to pick an auth provider and log in with the Rancher server.
The kubeconfig token is cached in the path where you run kubectl under `./.cache/token`. This token is valid till [it expires](../../api/api-tokens/#setting-ttl-on-kubeconfig-tokens-period), or [gets deleted from the Rancher server](../../api/api-tokens/#deleting-tokens)
Upon expiration, the next `kubectl get pods` will ask you to log in with the Rancher server again.
When you first run kubectl, for example, `kubectl get pods`, it will ask you to pick an auth provider and log in with the Rancher server.
The kubeconfig token is cached in the path where you run kubectl under `./.cache/token`. This token is valid till [it expires](../../api/api-tokens/#setting-ttl-on-kubeconfig-tokens-period), or [gets deleted from the Rancher server](../../api/api-tokens/#deleting-tokens)
Upon expiration, the next `kubectl get pods` will ask you to log in with the Rancher server again.
_Note_
As of CLI [v2.4.10](https://github.com/rancher/cli/releases/tag/v2.4.10), the kubeconfig token can be cached at a chosen path with `cache-dir` flag or env var `RANCHER_CACHE_DIR`.
_**Current Known Issues**_
1. If [authorized cluster endpoint]({{<baseurl>}}/rancher/v2.x/en/overview/architecture/#4-authorized-cluster-endpoint) is enabled for RKE clusters to [authenticate directly with downstream cluster]({{<baseurl>}}/rancher/v2.x/en/cluster-admin/cluster-access/kubectl/#authenticating-directly-with-a-downstream-cluster) and Rancher server goes down, all kubectl calls will fail after the kubeconfig token expires. No new kubeconfig tokens can be generated if Rancher server isn't accessible.
2. If a kubeconfig token is deleted from Rancher [API tokens]({{<baseurl>}}/rancher/v2.x/en/api/api-tokens/#deleting-tokens) page, and the token is still cached, cli won't ask you to login again until the token expires or is deleted.
`kubectl` calls will result into an error like `error: You must be logged in to the server (the server has asked for the client to provide credentials`. Tokens can be deleted using `rancher token delete`.