Add how-to troubleshoot for charm deployments

docs/src/charm/howto/index.md

@@ -29,6 +29,7 @@ custom-registry
 Upgrade patch version <upgrade-patch>
 Upgrade minor version <upgrade-minor>
 Validate the cluster <validate>
+troubleshooting
 ```
 
 ---

docs/src/charm/howto/troubleshooting.md

@@ -0,0 +1,224 @@
+# How to troubleshoot {{product}}
+
+Identifying issues in a Kubernetes cluster can be difficult, especially to new users. With {{product}} we aim to make deploying and managing your cluster as easy as possible. This how-to guide will walk you through the steps to troubleshoot your {{product}} cluster.
+
+## Common issues
+
+Maybe your issue has already been solved? Check out the [troubleshooting reference][charm-troubleshooting-reference] page to see a list of common issues and their solutions. Otherwise continue with this guide to help troubleshoot your {{product}} cluster.
+
+## Verify that the cluster status is ready
+
+Verify that the cluster status is ready by running the following command:
+
+```
+juju status
+```
+
+You should see output similar to the following:
+```
+Model        Controller           Cloud/Region         Version  SLA          Timestamp
+k8s-testing  localhost-localhost  localhost/localhost  3.6.1    unsupported  09:06:50Z
+
+App         Version  Status  Scale  Charm       Channel    Rev  Exposed  Message
+k8s         1.32.0   active      1  k8s         1.32/beta  179  no       Ready
+k8s-worker  1.32.0   active      1  k8s-worker  1.32/beta  180  no       Ready
+
+Unit           Workload  Agent  Machine  Public address  Ports     Message
+k8s-worker/0*  active    idle   1        10.94.106.154             Ready
+k8s/0*         active    idle   0        10.94.106.136   6443/tcp  Ready
+
+Machine  State    Address        Inst id        Base          AZ  Message
+0        started  10.94.106.136  juju-380ff2-0  [email protected]      Running
+1        started  10.94.106.154  juju-380ff2-1  [email protected]      Running
+```
+In this example we can glean some information. The `Workload` column will show the status of a given service. The `Message` section will show you the health of a given service in the cluster. During deployment and maintenance these workload statuses will update to reflect what a given node is doing. For example the workload may say `maintenance` while message will describe this maintenance as `Ensuring snap installation`.
+
+
+During normal operation the Workload should read `active`, the Agent column (which reflects what the Juju agent is doing) should read `idle`, and the messages will either say `Ready` or another descriptive term.
+
+## Verify that the API server is healthy
+
+Fetch the kubeconfig file for a control-plane node(unit) in the cluster by running the following command:
+
+```
+juju run k8s/0 get-kubeconfig
+```
+
+```{warning}
+When running `juju run k8s/0 get-kubeconfig` you retrieve the kubeconfig file that uses the given unit's IP address.
+```
+
+Verify that the API server is healthy and reachable by running the following command:
+
+```
+kubectl get all
+```
+
+This command lists resources that exist under the default namespace. You should see output similar to the following if the API server is healthy:
+
+```
+NAME                 TYPE        CLUSTER-IP     EXTERNAL-IP   PORT(S)   AGE
+service/kubernetes   ClusterIP   10.152.183.1   <none>        443/TCP   29m
+```
+
+A typical error message may look like this if the API server can not be reached:
+
+```
+The connection to the server 127.0.0.1:6443 was refused - did you specify the right host or port?
+```
+
+The status of the API server service can be checked by running the following command:
+
+```
+juju exec --unit k8s/0 -- systemctl status snap.k8s.kube-apiserver
+```
+
+The logs of the API server service can be accessed by running the following command:
+
+```
+juju exec --unit k8s/0 -- journalctl -u snap.k8s.kube-apiserver
+```
+
+A failure could mean that:
+* The API server is not reachable due to network issues or firewall limitations
+* The API server on the particular node is unhealthy
+* The control-plane node that's being reached is down
+
+Try reaching the API server on a different unit by retrieving the kubeconfig file with `juju run <k8s/unit#> get-kubeconfig`.
+
+## Verify that the cluster nodes are healthy
+
+Verify that the nodes in the cluster are healthy by running the following command and checking if all of the nodes have the `Ready` status:
+
+```
+kubectl get nodes
+```
+
+## Troubleshooting an unhealthy node
+
+There are certain services running on each node of a {{product}} cluster which are required. The required services depend on the type of node.
+
+Services running on both types of nodes:
+* `k8sd`
+* `kubelet`
+* `containerd`
+* `kube-proxy`
+
+Services running only on control-plane nodes:
+* `kube-apiserver`
+* `kube-controller-manager`
+* `kube-scheduler`
+* `k8s-dqlite`
+
+Services running only on worker nodes:
+* `k8s-apiserver-proxy`
+
+SSH to the unhealthy node by running the following command:
+
+```
+juju ssh <k8s/unit#>
+```
+
+Check the status of these services on the failing node by running the following command:
+
+```
+sudo systemctl status snap.k8s.<service>
+```
+
+The logs of a failing service can be checked by running the following command:
+
+```
+sudo journalctl -xe -u snap.k8s.<service>
+```
+
+If the issue indicates a problem with the configuration of the services on the node, examine the arguments used to run these services.
+
+The arguments of a service on the failing node can be examined by reading the file located at `/var/snap/k8s/common/args/<service>`
+
+## Verify that the system pods are healthy
+
+Verify that the cluster's pods are healthy by running the following command and checking if all of the pods are `Running` and `Ready`:
+
+```
+kubectl get pods -n kube-system
+```
+
+The pods in the `kube-system` namespace belong to {{product}}' features such as `network`. Unhealthy pods could be related to configuration issues or nodes not meeting certain requirements.
+
+## Troubleshooting a failing pod
+
+Events on a failing pod can be checked by running the following command:
+
+```
+kubectl describe pod <pod-name> -n <namespace>
+```
+
+Logs on a failing pod can be checked by running the following command:
+
+```
+kubectl logs <pod-name> -n <namespace>
+```
+
+## Using the built-in inspection script
+
+{{product}} ships with a script to compile a complete report on {{product}} and the system which it is running on. This is an essential tool for bug reports and for investigating whether a system is (or isn’t) working and collecting all the relevant data in one place.
+
+Inspection script can be executed on a specific unit by running the following commands:
+
+```
+juju exec --unit <k8s/unit#> -- sudo /snap/k8s/current/k8s/scripts/inspect.sh /home/ubuntu/inspection-report.tar.gz
+juju scp <k8s/unit#>:/home/ubuntu/inspection-report.tar.gz ./
+```
+
+The command output is similar to the following:
+```
+Collecting service information
+Running inspection on a control-plane node
+ INFO:  Service k8s.containerd is running
+ INFO:  Service k8s.kube-proxy is running
+ INFO:  Service k8s.k8s-dqlite is running
+ INFO:  Service k8s.k8sd is running
+ INFO:  Service k8s.kube-apiserver is running
+ INFO:  Service k8s.kube-controller-manager is running
+ INFO:  Service k8s.kube-scheduler is running
+ INFO:  Service k8s.kubelet is running
+Collecting registry mirror logs
+Collecting service arguments
+ INFO:  Copy service args to the final report tarball
+Collecting k8s cluster-info
+ INFO:  Copy k8s cluster-info dump to the final report tarball
+Collecting SBOM
+ INFO:  Copy SBOM to the final report tarball
+Collecting system information
+ INFO:  Copy uname to the final report tarball
+ INFO:  Copy snap diagnostics to the final report tarball
+ INFO:  Copy k8s diagnostics to the final report tarball
+Collecting networking information
+ INFO:  Copy network diagnostics to the final report tarball
+Building the report tarball
+ SUCCESS:  Report tarball is at /home/ubuntu/inspection-report.tar.gz
+```
+
+This script confirms that the services are running, and the resulting report file can be viewed to get a detailed look at every aspect of the system.
+
+## Collecting debug information
+
+To collect comprehensive debug output from your {{product}} cluster, install and run [juju-crashdump][] on a computer that has the Juju client installed, with the current controller and model pointing at your {{product}} deployment.
+
+```
+sudo snap install juju-crashdump --classic --channel edge
+juju-crashdump -a debug-layer -a config
+```
+
+Running the `juju-crashdump` script will generate a tarball of debug information that includes systemd unit status and logs, Juju logs, charm unit data, and Kubernetes cluster information. It is recommended that you include this tarball when filing a bug.
+
+## Reporting a bug
+If you cannot solve your issue and believe the fault may lie in {{product}}, please [file an issue on the project repository][].
+
+Help us deal effectively with issues by including the report obtained from the inspect script, the tarball obtained from `juju-crashdump`, as well as any additional logs, and a summary of the issue.
+
+<!-- Links -->
+
+[file an issue on the project repository]: https://github.com/canonical/k8s-operator/issues/new/choose
+[charm-troubleshooting-reference]: ../reference/troubleshooting
+[juju-crashdump]: https://github.com/juju/juju-crashdump

docs/src/charm/reference/index.md

@@ -18,6 +18,7 @@ architecture
 Ports and Services <ports-and-services>
 charm-configurations
 Community <community>
+troubleshooting
 
 ```
 

docs/src/charm/reference/troubleshooting.md

@@ -0,0 +1,4 @@
+# Troubleshooting
+
+This page provides techniques for troubleshooting common {{product}}
+issues.