opendesk/docs/debugging.md

<!--
SPDX-FileCopyrightText: 2023 Bundesministerium des Innern und für Heimat, PG ZenDiS "Projektgruppe für Aufbau ZenDiS"
SPDX-License-Identifier: Apache-2.0
-->

<h1>Debugging</h1>

<!-- TOC -->
* [Disclaimer](#disclaimer)
* [Enable debugging](#enable-debugging)
* [Adding containers to a pod for debugging purposes](#adding-containers-to-a-pod-for-debugging-purposes)
  * [Adding a container to a pod/deployment - Dev/Test only](#adding-a-container-to-a-poddeployment---devtest-only)
  * [Temporary/ephemeral containers](#temporaryephemeral-containers)
* [Components](#components)
  * [Helmfile](#helmfile)
  * [MariaDB](#mariadb)
  * [Nextcloud](#nextcloud)
  * [OpenProject](#openproject)
  * [PostgreSQL](#postgresql)
  * [Keycloak](#keycloak)
<!-- TOC -->

# Disclaimer

This document collects information on how to deal with debugging an openDesk deployment.

It will be extended over time as we deal with debugging cases.

We for sure do not want to reinvent the wheel, so we might link to external sources that contain helpful
information where available.

> **Warning**<br>
> You should never enable the debug option in production environments! By looking up `debug.enabled` in the deployment, you
will find the various places changes are applied when enabling debugging. So, outside of development and test
environments, you should use them thoughtfully and carefully if needed.

# Enable debugging

Check the openDesk [`debug.yaml.gotmpl`](../helmfile/environments/default/debug.yaml.gotmpl) and set for your deployment
```
debug:
  enabled: true
```

This will result in:
- setting most component's log level to debug
- making the Keycloak admin console available by default at `https://id.<your_domain>/admin/`
- configured the ingress for `http://minio-console.<your_domain>`

> **Note**<br>
> When enabling debug mode and updating your deployment, you must manually delete all jobs before updating. In debug mode, we keep the jobs, and some job fields are immutable, leading to a deployment failure.

> **Note**<br>
> All containers should write their log output to STDOUT; if you find (valuable) logs inside a container, please let us know!

# Adding containers to a pod for debugging purposes

During testing or development, you may need to execute tools, browse, or even change things in the filesystem of another container.

This can be a challenge the more security-hardened the container images are because there are no debugging tools available, and sometimes, there is not even a shell.

Adding a container to a Pod can ease the pain.

Below are some wrap-up notes on debugging openDesk by adding debug containers. Of course, there are many more detailed resources out there.

## Adding a container to a pod/deployment - Dev/Test only

You can add a container by editing and updating an existing deployment, which is quite comfortable with tools like [Lens](https://k8slens.dev/).

- Select the container you want to make use of as a debugging container; in the example below, it is `registry.opencode.de/bmi/opendesk/components/platform-development/images/opendesk-debugging-image:latest`.
- Ensure the `shareProcessNamespace` option is enabled for the Pod.
- Reference the selected container within the `containers` array of the deployment.
- If you want to access another container's filesystem, ensure both containers' user/group settings match.
- Save & update the deployment.

The following example can be used to debug the `openDesk-Nextcloud-PHP` container; if you want to modify files, remember to set `readOnlyRootFilesystem` to `true` on the PHP container.

```yaml
      shareProcessNamespace: true
      containers:
        - name: debugging
          image: registry.opencode.de/bmi/opendesk/components/platform-development/images/opendesk-debugging-image:latest
          command: ["/bin/bash", "-c", "while true; do echo 'This is a temporary container for debugging'; sleep 5 ; done"]
          securityContext:
            capabilities:
              drop:
                - ALL
            privileged: false
            runAsUser: 65532
            runAsGroup: 65532
            runAsNonRoot: true
            readOnlyRootFilesystem: false
            allowPrivilegeEscalation: false
            seccompProfile:
              type: RuntimeDefault
```

- After the deployment has been reloaded, open the shell of the debugging container.
- When you've succeeded, you will see the processes of both/all containers in the Pod when doing a `ps aux`.
- To access other containers' filesystems, select the PID of a process from the other container and do a `cd /proc/<selected_process_id>/root`.

## Temporary/ephemeral containers

An interesting read we picked most of the details below from: https://iximiuz.com/en/posts/kubernetes-ephemeral-containers/

Sometimes, you do not want to add a container permanently to your existing deployment. In that case, you could use [ephemeral containers](https://kubernetes.io/docs/concepts/workloads/pods/ephemeral-containers/).

For the commands further down this section, we set some environment variables first:
- `NAMESPACE`: The namespace in which the Pod you want to inspect is running.
- `DEPLOYMENT_NAME`: The deployment's name responsible for spawning the Pod you want to inspect within the pre-mentioned namespace.
- `POD_NAME`: The name of the Pod you want to inspect within the pre-mentioned namespace.
- `EPH_CONTAINER_NAME`: Choose the name for the container, and "debugging" seems obvious.
- `DEBUG_IMAGE`: The image you want to use for debugging purposes.

e.g.

```shell
export EPH_CONTAINER_NAME=debugging
export NAMESPACE=my_test_deployment
export DEPLOYMENT_NAME=opendesk-nextcloud-php
export POD_NAME=opendesk-nextcloud-php-6686d47cfb-7642f
export DEBUG_IMAGE=registry.opencode.de/bmi/opendesk/components/platform-development/images/opendesk-debugging-image:latest
```

You still need to ensure that your deployment supports process namespace sharing:

```shell
kubectl -n ${NAMESPACE} patch deployment ${DEPLOYMENT_NAME} --patch '
spec:
  template:
    spec:
      shareProcessNamespace: true'
```

Now, you can add the ephemeral container with:
```shell
kubectl -n ${NAMESPACE} debug -it --attach=false -c ${EPH_CONTAINER_NAME} --image={DEBUG_IMAGE} ${POD_NAME}
```
and open its interactive terminal with
```shell
kubectl -n ${NAMESPACE} attach -it -c ${EPH_CONTAINER_NAME} ${POD_NAME}
```

# Components

## Helmfile

When refactoring the Helmfile structure you want to ensure that there are not unintended mistakes by e.g. `diff`
comparing the output of Helmfile from before and after the change by calling:

```shell
helmfile template -e dev >output_to_compare.yaml
```

## MariaDB

When using the openDesk bundled MariaDB, you can explore the database(s) using the MariaDB interactive terminal from the Pod's command line: `mariadb -u root -p`. On the password prompt, provide the value for `MARIADB_ROOT_PASSWORD` found in the Pod's environment.

While you will find all the details for the CLI tool in [the online documentation](https://mariadb.com/kb/en/mariadb-command-line-client/), some quick commands are:

- `help`: Get help on the psql command set
- `show databases`: Lists all databases
- `use <databasename>`: Connect to `<databasename>`
- `show tables`: Lists tables within the currently connected database
- `quit`: Quit the client

## Nextcloud

`occ` is the CLI for Nextcloud; all the details can be found in the [upstream documentation](https://docs.nextcloud.com/server/latest/admin_manual/configuration_server/occ_command.html).

You can run occ commands in the `opendesk-nextcloud-php` pod like this: `php /var/www/html/occ config:list`

## OpenProject

OpenProject is a Ruby on Rails application. Therefore, you can make use of the Rails console from the Pod's command line `bundle exec rails console`
and run debug code like this:

```
uri = URI('https://nextcloud.url/apps/integration_openproject/check-config')
Net::HTTP.start(uri.host, uri.port,
 :use_ssl => uri.scheme == 'https') do |http|
 request = Net::HTTP::Get.new uri
 response = http.request request # Net::HTTPResponse object
end
```

## PostgreSQL

Using the openDesk bundled PostgreSQL, you can explore database(s) using the PostgreSQL interactive terminal from the Pod's command line: `psql -U postgres`.

While you will find all details in the [psql subsection](https://www.postgresql.org/docs/current/app-psql.html)) of the PostgreSQL documentation, some quick commands are:

- `\?`: Get help on the psql command set
- `\l`: Lists all databases
- `\c <databasename>`: Connect to `<databasename>`
- `\dt`: List (describe) tables within the currently connected database
- `\q`: Quit the client

## Keycloak

Keycloak is the gateway to integrate other authentication management systems or applications. It can be desired to
avoid enabling debug mode for the whole platform when you just need to look into Keycloak.

That can easily be achieved in two steps:

1. Updating the value for `KC_LOG_LEVEL` in the related configmap `ums-keycloak`.
```shell
export NAMESPACE=<your_namespace>
export CONFIGMAP_NAME=ums-keycloak
kubectl patch -n ${NAMESPACE} configmap ${CONFIGMAP_NAME} --type merge -p '{"data":{"KC_LOG_LEVEL":"DEBUG"}}'
```

2. Restart the Keycloak Pod(s).

> **Note**<br>
> As the `ums-keycloak-extensions-handler` is performing frequent (one per second) requests to Keycloak for retrieval of the Keycloak event history, you might want to stop/remove the deployment while debugging/analysing Keycloak to not get your debug output spammed by these requests.