Convert docs to MyST

docs/conf.py

@@ -14,7 +14,8 @@
 # Enable MathJax for Math
 extensions = ['sphinx.ext.mathjax',
               'sphinx.ext.intersphinx',
-              'sphinx_copybutton']
+              'sphinx_copybutton',
+              'myst_parser']
 
 # The master toctree document.
 master_doc = 'index'

docs/configuration/cull.md

@@ -0,0 +1,7 @@
+# Culling idle servers
+
+Plasma uses the [same defaults as The Littlest JupyterHub](http://tljh.jupyter.org/en/latest/topic/idle-culler.html#default-settings)
+for culling idle servers.
+
+It overrides the `timeout` value to `3600`, which means that the user servers will be shut down if they have
+been idle for more than one hour.

docs/configuration/index.md

@@ -0,0 +1,11 @@
+# Configuration
+
+```{toctree}
+:maxdepth: 3
+
+monitoring
+persistence
+resources
+cull
+namedservers
+```

docs/configuration/monitoring.md

@@ -0,0 +1,98 @@
+# Monitoring
+
+:::{warning}
+HTTPS must be enabled to be able to access Cockpit. Refer to {ref}`install-https` for more info.
+:::
+
+## Installing Cockpit
+
+`cockpit` is not installed by default as a monitoring tool for the server.
+
+First make sure HTTPS is enabled and the `name_server` variable is specified in the `hosts` file.
+See {ref}`install-https` for more info.
+
+Then execute the `cockpit.yml` playbook:
+
+```bash
+ansible-playbook cockpit.yml -i hosts -u ubuntu
+```
+
+The Plasma TLJH plugin registers `cockpit` as a JupyterHub service. This means that
+Cockpit is accessible to JupyterHub admin users via the JupyterHub interface:
+
+```{image} ../images/configuration/cockpit-navbar.png
+:align: center
+:alt: Accessing cockpit from the nav bar
+:width: 100%
+```
+
+Users will be asked to login with their system credentials. They can then access the Cockpit dashboard:
+
+```{image} ../images/configuration/cockpit.png
+:align: center
+:alt: Cockpit
+:width: 100%
+```
+
+## Monitoring user servers with Cockpit
+
+:::{note}
+Access to Docker Containers requires access to `docker`.
+
+Make sure your user can access docker on the machine with:
+
+```bash
+sudo docker info
+```
+
+Your user should also be able to login with a password. If the user doesn't have a password yet, you can
+create a new one with:
+
+```bash
+sudo passwd <username>
+```
+
+For example if your user is `ubuntu`:
+
+```bash
+sudo passwd ubuntu
+```
+
+To add more users as admin or change permissions from the Cockpit UI, see {ref}`monitoring-permissions`.
+:::
+
+Since user servers are started as Docker containers, they will be displayed in the Cockpit interface in the
+`Docker Containers` section:
+
+```{image} ../images/configuration/cockpit-docker.png
+:align: center
+:alt: Docker Containers from Cockpit
+:width: 100%
+```
+
+The Cockpit interface shows:
+
+- The username as part of the name of the Docker container
+- The resources they are currently using
+- The environment currently in use
+
+It is also possible to stop the user server by clicking on the "Stop" button.
+
+(monitoring-permissions)=
+
+## Changing user permissions from the Cockpit UI
+
+:::{note}
+You first need to be logged in with a user that has the `sudo` permission.
+:::
+
+Cockpit makes it easy to add a specific user to a certain group.
+
+For example a user can be given the "Container Administrator" role via the UI to be able to manage Docker containers
+and images on the machine:
+
+```{image} ../images/configuration/cockpit-roles.png
+:align: center
+:alt: Manage user roles from the Cockpit UI
+:width: 100%
+```

docs/configuration/namedservers.md

@@ -0,0 +1,23 @@
+# Named Servers
+
+By default, users can run only one server at once.
+
+[Named servers functionality](https://jupyterhub.readthedocs.io/en/stable/reference/config-user-env.html#named-servers) in JupyterHub
+can be activated to let the user run several servers.
+
+To allow up to 2 simultaneous named servers (in addition to the default one), create the file `named_servers_config.py`
+in the directory `/opt/tljh/config/jupyterhub_config.d` with the following content:
+
+```text
+c.JupyterHub.allow_named_servers = True
+c.JupyterHub.named_server_limit_per_user = 2
+```
+
+Then, reload tljh:
+
+```text
+sudo tljh-config reload
+```
+
+Have a look at the [named servers documentation](https://jupyterhub.readthedocs.io/en/stable/reference/config-user-env.html#named-servers)
+for more details.

docs/configuration/persistence.md

@@ -0,0 +1,91 @@
+# Data Persistence
+
+(persistence-user-data)=
+
+## User Data
+
+The user servers are started using JupyterHub's [SystemUserSpawner](https://github.com/jupyterhub/dockerspawner#systemuserspawner).
+
+This spawner is based on the [DockerSpawner](https://github.com/jupyterhub/dockerspawner#dockerspawner), but makes it possible
+to use the host users to start the notebook servers.
+
+Concretely this means that the user inside the container corresponds to a real user that exists on the host.
+Processes will be started by that user, instead of the default `jovyan` user that is usually found in the regular
+Jupyter Docker images and on Binder.
+
+For example when the user `foo` starts their server, the list of processes looks like the following:
+
+```bash
+foo@9cf23d669647:~$ ps aux
+USER       PID %CPU %MEM    VSZ   RSS TTY      STAT START   TIME COMMAND
+root         1  1.1  0.0  50944  3408 ?        Ss   11:17   0:00 su - foo -m -c "$0" "$@" -- /srv/conda/envs/notebook/bin/jupyterhub-singleuser --ip=0.0.0.0 --port=8888 --NotebookApp.default_url=/lab --ResourceUseDisplay.track_cpu_percent=True
+foo         32  5.4  0.8 399044 70528 ?        Ssl  11:17   0:01 /srv/conda/envs/notebook/bin/python /srv/conda/envs/notebook/bin/jupyterhub-singleuser --ip=0.0.0.0 --port=8888 --NotebookApp.default_url=/lab --ResourceUseDisplay.track_cpu_percent=True
+foo         84  0.0  0.0  20312  4036 pts/0    Ss   11:17   0:00 /bin/bash -l
+foo        112 29.0  0.5 458560 46448 ?        Ssl  11:17   0:00 /srv/conda/envs/notebook/bin/python -m bash_kernel -f /home/foo/.local/share/jupyter/runtime/kernel-9a7c8ad3-4ac2-4754-88cc-ef746d1be83e.json
+foo        126  0.5  0.0  20180  3884 pts/1    Ss+  11:17   0:00 /bin/bash --rcfile /srv/conda/envs/notebook/lib/python3.8/site-packages/pexpect/bashrc.sh
+foo        140  0.0  0.0  36076  3368 pts/0    R+   11:17   0:00 ps aux
+```
+
+The following steps happen when a user starts their server:
+
+1. Mount the user home directory on the host into the container. This means that the file structure in the container reflects what is on the host.
+2. A new directory is created in the user home directory for each new environment (i.e for each Docker image).
+   For example if a user starts the `2020-python-course` environment, there will be a new folder created under `/home/user/2020-python-course`.
+   This folder is then persisted to disk in the user home directory on the host. Any file and notebook created from the notebook interface are also persisted to disk.
+3. On server startup, the entrypoint script copies the files from the base image that are initially in `/home/jovyan` to `/home/user/2020-python-course` in the container.
+   They are then persisted in `/home/user/2020-python-course` on the host.
+
+```{image} ../images/configuration/persistence.png
+:align: center
+:alt: Mounting user's home directories
+:width: 80%
+```
+
+- The files highlighted in blue correspond to the files initially bundled in the environment. These files are copied to the environment subdirectory in the user home directory on startup.
+- The other files are examples of files created by the user.
+
+## User server startup
+
+The user server is started from the environment directory:
+
+```{image} ../images/configuration/user-server-rootdir.png
+:align: center
+:alt: User servers are started in the environment directory
+:width: 50%
+```
+
+The rest of the user files are mounted into the container, see {ref}`persistence-user-data`.
+
+A user can for example open a terminal and access their files by typing `cd`.
+
+They can then inspect their files:
+
+```text
+foo@3e29b2297563:/home/foo$ ls -lisah
+total 56K
+ 262882 4.0K drwxr-xr-x  9 foo  foo  4.0K Apr 21 16:53 .
+6205024 4.0K drwxr-xr-x  1 root root 4.0K Apr 21 16:50 ..
+ 266730 4.0K -rw-------  1 foo  foo   228 Apr 21 14:41 .bash_history
+ 262927 4.0K -rw-r--r--  1 foo  foo   220 May  5  2019 .bash_logout
+ 262928 4.0K -rw-r--r--  1 foo  foo  3.7K May  5  2019 .bashrc
+1043206 4.0K drwx------  3 foo  foo  4.0K Apr 21 09:26 .cache
+ 528378 4.0K drwx------  3 foo  foo  4.0K Apr 17 17:36 .gnupg
+1565895 4.0K drwxrwxr-x  2 foo  foo  4.0K Apr 21 09:55 .ipynb_checkpoints
+1565898 4.0K drwxr-xr-x  5 foo  foo  4.0K Apr 21 09:27 .ipython
+1565880 4.0K drwxrwxr-x  3 foo  foo  4.0K Apr 21 09:26 .local
+ 262926 4.0K -rw-r--r--  1 foo  foo   807 May  5  2019 .profile
+1050223 4.0K drwxrwxr-x 12 foo  foo  4.0K Apr 20 10:44 2020-python-course
+1043222 4.0K drwxrwxr-x 13 foo  foo  4.0K Apr 20 17:07 r-intro
+ 258193 4.0K -rw-rw-r--  1 foo  foo   843 Apr 21 09:56 Untitled.ipynb
+```
+
+## Shared Data
+
+In addition to the user data, the plugin also mounts a shared data volume for all users.
+
+The shared data is available under `/srv/data` inside the user server, as pictured in the diagram above.
+
+On the host machine, the shared data should be placed under `/srv/data` as recommended in the
+[TLJH documentation](http://tljh.jupyter.org/en/latest/howto/content/share-data.html#option-2-create-a-read-only-shared-folder-for-data).
+
+The shared data is **read-only**.