WissKI-Distillery
WissKI-Distillery is a Docker-based server provisioning and managing for multiple WissKI instances.
The WissKI Distillery is a set of scripts, tools, and applications that allows to operate a WissKI cloud of distinct but jointly managed WissKI instances, hosted on a dedicated hardware system.
Like the WissKI system, the WissKI Distillery is open source and free to use.
This README contains only technical documentation.
For members of FAU Erlangen-Nürnberg a cloud offering based on this service known as FAUWissKICloud.
Please see https://wisski.data.fau.de/ for related documentation.
** THIS DOCUMENTATION IS ALWAYS OUTDATED **
Overview
This project consists of the following:
- this README containing mostly technical documentation
- a [NEWS.md] file containing recent technical enhancements
- a Go command
wdcli
High-Level Overview
NOTE: A list of new features can be found in NEWS
The Distillery consists of a set of instances of WissKIs and (high-level) components.
Components are implemented in the component directory.
Furthermore each instance consists of several components refered to as ingredients.
Ingredients are implemented in ingredient directory.
Each WissKI is implemented as a single docker container that talks to several components.
There are only three components that directly talk to a WissKI Instance
Furthermore to allow end-users to access the WissKIs two further components exist:
- web powered by Traefik to route web traffic to inidividual instances
- A custom ssh server that enables ssh access to individual WissKI instances
Finally two other components exist:
- A public homepage that lists all instances, and provides basic statistics
- An instance administration system called info that allows web admins to perform certain maintance tasks
Technical Overview
The go command is almost dependency free.
It only expects that docker
and docker compose
are available.
Each subcommand comes with documentation, which can be found in this readme (and the readme is always outdated), as well as via the command line when passing a --help
flag.
To bootstrap a new distillery instance, the wdcli bootstrap
command can be used.
First copy the executable onto the server, using a command similar as:
CGO_ENABLED=0 GOOS=linux GOARCH=amd64 go build -o wdcli ./cmd/wdcli && scp ./wdcli distillery.example.com:
Next, access the server and run the bootstrap
command:
$ ssh distillery.example.com
user@distillery.example.com$ sudo ./wdcli bootstrap
This will create a deployment directory (/var/www/deploy
by default).
Next, edit the configuration file /var/www/deploy/.env
and customize it to your liking.
Usually it only requires adjustment in very few places.
Next, download a GraphDB zip file, and bring the distillery online using:
/var/www/deploy/wdcli system_update /path/to/graphdb.zip
System Updates
TLDR: sudo /var/www/deploy/wdcli system_update /path/to/graphdb.zip
To run a WissKI Distillery, several core Docker Instances must be installed.
These are:
-
nginx-proxy - an automated nginx reverse proxy
-
MariaDB - an SQL server
- It is configured to run inside a docker container
- A passwordless
root
account is created, which can only be used from inside the container.
- An additional admin account (as defined per config file) is created, which is used for administration.
- A secondary management account is also created. This is configured via the distillery configuration file, and can be access from anywhere.
- A
bookkeeping
database and table is created by default, to store known WissKI instance metadata in.
- It is accsssible using
127.0.0.1:3306
- A database shell can be opened using
sudo /var/www/deploy/wdcli mysql
.
- A phpmyadmin is started on
127.0.0.1:8080
.
- See internal/component/sql for implementation details.
-
proxyssh - an ssh server that delegates client connections to different WissKIs
- It is configured to run inside a docker container.
- Uses a global configurable authorized_keys file.
- Also allows users to write their own authorized_keys files.
- See distillery/resources/compose/ssh for implementation details.
-
wdresolve - a global WissKI Distillery Resolver
- It is configured to run inside a docker container
- Uses configuration which is updated with
sudo /var/www/deploy/wdcli update_prefix_config
- Running in the browser under the
/go/
path of the main domain.
- See distillery/resources/compose/resolver for implementation details.
-
dis
- a WissKI Distillery Information Server
To manage multiple docker containers, this script makes heavy use of docker compose.
Setting up these steps is fully automatic.
In particular, after obtaining a license and the installation zip file for 'GraphDB' one can just run:
sudo /var/www/deploy/wdcli system_update /path/to/graphdb.zip
In principle this script is idempotent, meaning it can be run multiple times achieving the same effect.
Provisioning a new WissKI instance -- 'wdcli provision'
TLDR: sudo /var/www/deploy/wdcli provision name-of-website
A new WissKI instance consists of several components:
- A Drupal instance inside a lightweight php runtime container (a
barrel
in which to store WissKI)
- An entry in the SQL bookkeeping table that stores instance meta-data
- An SQL database and user for Drupal
- A GraphDB repository and user as SPARQL endpoint
Each WissKI instance is identified by a ``slug''.
This is a preferably short name that is used to form a domain name for the WissKI instance.
The WissKI distillery assumes that each instance is a subdomain of a given domain.
For example, if the given domain is 'wisskis.example.com' and the slug of a particular instance is 'blue', the subdomain used by this instance would be 'blue.wisskis.example.com'.
The given domain can be configured within the '.env' file.
We use the following process to provision a new instance:
1. Create a new docker-compose.yml file
In this step we first create a directory on the real system to hold all files relating to this instance.
By default, this takes place inside /var/www/deploy/instances/$DOMAIN
, but this can be configured.
We then create a docker-compose file in this directory that is ready for running the barrel
container.
2. Create an appropriate SQL database and user
We create a new SQL database to eventually store Drupal-related data in.
The user and database names are generated from the slug.
The database password is randomly generated and only made available directly to the Drupal instance later.
3. Create a GraphDB repository and user
Next, we create a dedicated GraphDB repository for the WissKI instance.
We also create a new GraphDB user with access to this repository.
4. Provision the instance inside the container
We start the container in provisioning mode.
This does the following:
- Creates a new composer project that requires drupal/recommended-project`.
- Installs
drush
into this project.
- Runs the
drush site-install
command to configure the Drupal instance. Generates a random password to use.
- Adds and enables WissKI-specific modules for this instance.
- Sets up a WissKI Salz Adapter to use the GraphDB Repository.
6. Start the Docker Container
Finally, we can start the docker container.
These steps can be performed automatically.
To do so, use:
sudo /var/www/deploy/wdcli provision SLUG
Rebuild an instance -- 'wdcli rebuild'
Sometimes it becomes necessary (because of changes to this project) to rebuild the docker image running a certain docker instance.
To do so, use:
sudo /var/www/deploy/wdcli rebuild SLUG
Note that rebuilding an instance does restart the docker container resulting in a small (typical < 1 second) interruption to the website in question.
Furthermore, while the container recreated, the old image stays on the host.
To delete all instances, run:
sudo docker image prune --all
To automatically rebuild all instances, use the rebuild command without any arguments:
sudo /var/www/deploy/wdcli rebuild
Reserving an instance -- 'wdcli reserve'
Sometimes it is useful to reserve a particular instance name.
This is done by hosting a placeholder website at the domain.
To do so, use:
sudo /var/www/deploy/wdcli reserve SLUG
To un-reserve a website, manually stop the docker stack and remove the folder.
Purge an existing WissKI instance -- 'wdcli purge'
Sometimes it is required to remove a given WissKI instance.
In particular all parts belonging to it should be removed.
To use it, run:
sudo /var/www/deploy/wdcli purge SLUG
This cannot be undone (expect for manually re-installing a backup or snapshot).
Therefore it typically requires explicit confirmation.
Open a shell -- 'wdcli shell'
Sometimes manual changes to a given WissKI instance are required.
For this purpose, you can use:
sudo /var/www/deploy/wdcli shell SLUG
This will open a shell in the provided WissKI instance.
List all instances -- 'wdcli ls'
To list all instances, the following command can be used:
sudo /var/www/deploy/wdcli ls
Backups & Snapshots -- 'wdcli backup' and 'wdcli snapshot'
Backup the entire Distillery
This project comes with a backup script.
To make a backup of all instances, run:
sudo /var/www/deploy/wdcli backup
Backups may temporarily shutdown individual instances to ensure data consistency.
Typical backup times are a minute or less.
Backups are stored in the /var/www/deploy/snapshots/archives
directory.
They contain:
- a snapshot of every single instance (see below)
- a complete backup of the SQL database
- nquads of all the GraphDB repositories
- a backup of the configuration + data file(s)
Files are .tar.gz
ipped.
By default, backups are kept for up to thirty days, after which they are removed.
This can be configured in the WissKI Distillery Configuration File.
Snapshot a single instance
To snapshot a single instance, you can sudo /var/www/deploy/wdcli snapshot SLUG
.
It takes either 1 or 2 arguments:
# snapshot a single instance and pick a new file in /snapshots/archives
sudo /var/www/deploy/wdcli snapshot SLUG
# backup a single instance into a specific file
sudo /var/www/deploy/wdcli snapshot SLUG /path/to/snapshot.tar.gz
The snapshot proceeeds as follows:
- make a copy of the instance configuration
- shutdown the running instance
- make a dump of the triplestore and mysql databases
- make a copy of the file system
- export all pathbuilders
- start the instance again
- package the data into the final
.tar.gz
file
When uptime is critical, it is possible to skip shutting down a running instance.
This might result in inconsistent backup data.
To do so, run the script with the --keepalive
flag:
sudo /var/www/deploy/wdcli snapshot SLUG --keepalive
SSH Access
The distillery exposes an ssh daemon for users to access individual WissKI Shells.
It is running on port 2222 by default.
To access a shell in a particular barrel set the username equal to the slug.
For instance, to gain access to a shell inside a WissKI instance with a slug porcelain
use the following command line:
ssh -p 2222 porcelain@localhost
Replace localhost
with the hostname of the WissKI Distillery.
Inside the container, normal shell acess is provided.
Both drush
and composer
are available.
No technical reasons using sudo
or switching to root
is not possible.
Authentication
Authentication is performed using SSH Keys.
They are associated o distillery user accounts.
Within each instance, ssh keys can be added to the file /var/www/.ssh/authorized_keys
using the default OpenSSH authorized_keys
format.
Furthermore, global ssh Keys (that have access to every instance) can be added to a GLOBAL_AUTHORIZED_KEYS_FILE
. This is set in the Distillery .env
file, and defaults to /distillery/authorized_keys/
.
Port Forwarding
In order to access the GraphDB Workbench or phpmyadmin ssh port forwarding can be used.
GraphDB is running on the host triplestore
on port 7200
.
PhpMyAdmin is running on the host phpmyadmin
on port 8080
.
To forward both you can use a command such as:
ssh -p 2222 -L localhost:7200:triplestore:7200 -L localhost:8080:phpmyadmin:8080 porcelain@localhost
This will make GraphDB and PhpMyAdmin available at localhost:7200
and localhost:8080
for the duration of the connection.
Resolver
In order to resolve WissKI URIs globally, we make use of wdresolve.
This can be queried with a single URI, and will be redirected to the page of the corresponding WissKI Entity.
This is deployed under /go/
path of the top-level domain.
For example, if the domain name of the distillery instance is wisski.example.com
, then the resolver would respond to queries like https://wisski.example.com/go/?uri=https://first.wisski.example.com/content/123
.
The resolver configuration is automatically updated by the update_prefix_config.sh
script.
It should not be neccessary to reload this configuration manually, as it is automatically called during system_update.sh
.
It is also possible to manually add a URI prefix to an instance.
For this purpose, add a file named prefixes
to the base directory of the instance, with one prefix per line.
Furthermore, you can also exclude a specific instance from URL prefix resolving.
This should be the case for cloned or backup instances.
For this purpose, add a file named prefixes.skip
to the base directory of the instance.
This will casuse the instance to be skipped entirely.
License
This project and associated files in this repository are licensed as follows:
WissKI-Distillery - A docker-based WissKI instance server
Copyright (C) 2020-22 CDI <https://www.cdi.fau.de/>
This program is free software: you can redistribute it and/or modify
it under the terms of the GNU Affero General Public License as
published by the Free Software Foundation, either version 3 of the
License, or (at your option) any later version.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU Affero General Public License for more details.
You should have received a copy of the GNU Affero General Public License
along with this program. If not, see <https://www.gnu.org/licenses/>.
Please see LICENSE
for a legally binding license text.
The short summary of the license is:
- You may use this software for any purpose, including commerical.
- You may create derivative works, and use those for any purpose, including commerical.
if you follow the following conditions:
- You provide the end-user with a copy of this license.
- You make the source code of any derivative works available.
- Any derivative works clearly list changes made.
- You license any derivative works under the same license.
This also applies if you only run a backend service based on this software.