Spyderisk / system-modeller-deployment

Deployment scripts for Spyderisk System Modeller
Other
0 stars 1 forks source link
deployment

Spyderisk System Modeller Deployment project

The software tools created by the Spyderisk Open Project only works in very specific circumstances. This project exists to make it easier for you to get Spyderisk running in those few circumstances.

Here you will find instructions, scripts and configuration files to deploy an instance of the open source Spyderisk System Modeller on a machine you control. The System Modeller project repository has a lot more information about Spyderisk including academic reports and papers, plus of course the computer source code that this Deployment project will get running on your system.

We welcome contributions to this project. There's a lot to be done.

Table of contents

Where will Spyderisk run?

Spyderisk is open source, therefore it is primarily written for and targeted to Linux/Unix. We have made some efforts to make it work on Windows too, and within strict limits we do support Windows users.

The scenarios we explicitly support in this document for installing Spyderisk are:

We explicitly do not support Spyderisk in any of the following scenarios:

Required knowledge

You will need to understand how to install and configure Docker, and you need to understand what Keycloak does.

We do not recommend anyone install a service live on the internet unless they are confident in security measures around domain names, IP networking and routing, and virtualisation.

There is a lower level of knowledge required to install on a laptop, assuming the laptop is protected by its own firewall, and that docker is set up to disallow access from outside the laptop.

Technical overview

The deployment is made with docker-compose executed on the server or laptop.

This project orchestrates the following Docker containers:

Two orchestration definitions are provided:

Of the containers, the only one configured to expose any ports outside of the docker network is proxy.

Prerequisites

Docker is required to orchestrate the containers. Docker is available on various host operating systems, but we recommend using some form of Linux. For example, on Debian or Ubuntu type sudo apt install docker.

If you do use Windows Desktop, install the closed-source Docker Desktop making sure you comply with the Docker Desktop license.

One of the Chrome, Chromium or Firefox browsers. We have done the most testing to date with Chrome and that is probably the best-supported, but to the best of our knowledge Chromium works just as well. We strongly prefer working promoting open source first, and our browser testing and preferences in future are likely to reflect that.

On Windows Desktop you also need to first install Windows System for Linux v2 which in turn requires enabling Microsoft Hyper V virtualisation technology.

Only Windows Desktop 10 or Windows Desktop 11 are supported.

Deployment

General method:

  1. Edit the .env file to set appropriate values.
  2. Edit the .env_adaptor file to set appropriate values.
  3. Download a Spyderisk knowledgebase zip file asset. e.g. domain-network-6a3-2-2.zip and copy it into the knowledgebases folder.
  4. Run docker-compose pull to get the latest images (otherwise the locally cached ones are used, if they exist).
  5. Run docker-compose up -d or docker-compose -f docker-compose_external_kc.yml up -d to start the containers.

See below for details.

Keycloak

The Spyderisk System Modeller uses Keycloak to authenticate users. The docker-compose.yml script will create an insecure Keycloak container for testing, preconfigured with a Keycloak admin account, and Spyderisk testuser and testadmin accounts, all using password as the password. For a production system you should configure Spyderisk to connect to a secure external Keycloak service and launch the software using the docker-compose_external_kc.yml file.

To communicate securely, the Spyderisk System Modeller and Keycloak must be configured with a shared secret. This secret is defined in the KEYCLOAK_CREDENTIALS_SECRET variable in the .env file and is inserted into the ssm container (and into the optional insecure keycloak container).

To use an external Keycloak service, edit the .env file and update:

When configuring an external Keycloak, the suggested configuration is:

Deployment on a Linux Server

Edit .env and update SERVICE_PROTOCOL, SERVICE_DOMAIN, and SERVICE_PORT values to your organisation's settings.

To start the service and connect to an external Keycloak:

sudo -E docker-compose -f docker-compose_external_kc.yml up -d

or to connect to an insecure local Keycloak:

sudo -E docker-compose up -d

Multiple Deployments on the same Linux Server

Multiple deployments of the dockerised Spyderisk System Modeller can co-exist on the same server. Each deployment requires its own folder, and adjusted PORT settings.

Start the second deployment with or without a secure Keycloak service as described above. The second service can be access/proxied from the port defined by PROXY_EXTERNAL_PORT value.

Deployment on a Test Server

Sometimes we need to deploy onto a docker host server which does not have an FQDN and where we cannot open a public port. In this case we need to invent an FQDN and make it so that both the SSM container and the client web browser both (a) resolve to the docker host server when looking up the FQDN and (b) can access the server on the appropriate port.

For example, if the host server is fiab.altostrat.com and we are using the default public-facing port for the SSM of 8089, then to route from the web browser (client machine) to the server:

  1. Make up an FQDN for the service, e.g. example.com
  2. On the web browser (client) host, add a line to the hosts file (either /etc/hosts or C:\Windows\System32\drivers\etc\hosts) so that the client interprets the made up FQDN as being on the local machine:

    127.0.0.1 example.com

  3. From the local (client) host, make an SSH tunnel from the local localhost:8089 to localhost:8089 on the server (in this example, the host machine is fiab.altostrat.com and the SSH port is a non-standard 17248):
ssh -nNT -L 8089:localhost:8089 username@fiab.altostrat.com:17248 -v

This way, when the URL http://example.com:8089/whatever is put into the web browser on the client machine, the FQDN is looked up in the local hosts file and localhost (127.0.0.1) is returned. The browser connects on port 8089 on localhost which actually goes through the SSH tunnel to the server and connects to port 8089 there instead, where the reverse proxy is listening on port 8089.

To get the SSM container to also understand the made up FQDN (so that it can contact Keycloak using the FQDN) we need to:

  1. Add an extra_hosts clause to the docker-compose.yml file in the ssm definition, e.g.:
extra_hosts:
      - "example.com:host-gateway"
  1. Use the same made up FQDN in the .env file:
SERVICE_DOMAIN=example.com

The IP address in the extra_hosts clause needs to be the (internal) IP address of the host. When the SSM tries to connect to Keycloak on the URL http://example.com:8089/auth/something the extra_hosts clause means it will connect to the host machine and from there back into our reverse proxy and then to Keycloak. Without this the SSM container will try to use the resolv.conf file from the docker host which passes on to some DNS which does not know about the made up FQDN.

Deployment on a Personal Machine, both Linux and Windows

The software is designed to be deployed on a server which has an externally accessible domain name. It can be made to work on a personal machine but some extra configuration is required, and this is not our highest testing priority.

The Spyderisk software must be configured so that there is a single URL used to access Keycloak. The docker-compose.yml file sets the address to be ${SERVICE_PROTOCOL}://${SERVICE_DOMAIN}:${SERVICE_PORT}/auth/.

This URL must work (a) from the ssm container and (b) from the web browser of a user. On Windows (using Docker Desktop) an appropriate hostname for SERVICE_DOMAIN is host.docker.internal which Docker Desktop automatically inserts into the Windows hosts file (C:\Windows\System32\drivers\etc\hosts) to point at the docker host IP address. For Linux there is no equivalent host.docker.internal entry in the hosts file (/etc/hosts), so it has to be added manually, e.g. <ip address of docker gateway> host.docker.internal, the default docker gateway is 172.17.0.1.

An alternative is to manually edit the hosts file to add in your own made-up FQDN e.g. spyderisk.example.com along with the Docker gateway IP address and use that as the SERVICE_DOMAIN domain name.

To start the service with the insecure local Keycloak:

docker-compose up -d

or to connect to an external Keycloak:

docker-compose -f docker-compose_external_kc.yml up -d

The port exposed to the host machine is by default 8089 but this can be changed by editing value of PROXY_EXTERNAL_PORT in the .env file.

When the service is running, access e.g. http://host.docker.internal:8089/system-modeller in your web browser.

Inspecting an Existing Deployment

Accessing the Logs

From within a deployment's folder, the docker logs command can be used to inspect the log files:

#### get the log for all the containers:
docker-compose logs
#### get the log for the SSM:
docker-compose logs ssm
#### get the log for the SSM and "follow" the log file to see new entries as they arrive:
docker-compose logs ssm -f
#### tail the ssm log but starting from just the last few lines:
docker-compose logs ssm -f --tail=100

Finding the Spyderisk System Modeller Version

To discover what version of the SSM an existing deployment is using, the docker container inspect command can be used.

From within a deployment's folder, find the names of the containers:

docker-compose ps

       Name                     Command               State          Ports
----------------------------------------------------------------------------------
example_adaptor_1    gunicorn -b 0.0.0.0:8000 - ...   Up                                                 
example_keycloak_1   /tmp/import/entrypoint.sh        Up (healthy)   8080/tcp, 8443/tcp                  
example_mongo_1      docker-entrypoint.sh mongod      Up             27017/tcp                           
example_proxy_1      /tmp/import/entrypoint_tra ...   Up             0.0.0.0:8086->80/tcp,:::8086->80/tcp
example_ssm_1        /var/lib/tomcat/bin/catali ...   Up                                                 

The use, e.g.:

docker container inspect example_ssm_1 | grep 'image.revision'
                "org.opencontainers.image.revision": "426e694d0c86499505745a4418655d3ad641c9c1",

The long number is the git commit ID. It can be pasted into the search bar in GitHub (for instance) to jump to that specific commit. The build timestamp is also available with the key org.opencontainers.image.created.

Monitoring Resource Usage

To see CPU, memory and network usage (one off):

docker ps -q|xargs docker stats --no-stream

To keep monitoring it (like top):

docker ps -q | xargs docker stats

Upgrading a Deployment

It is sometimes possible to upgrade the SSM container in a deployment while keeping the user accounts and system models. This will only work if the new SSM software is compatible with the databases of the previous version.

  1. Go to the deployment's folder.
  2. Stop all the containers in the deployment:
$ docker-compose stop
Stopping system-modeller-deployment_proxy_1    ... done
Stopping system-modeller-deployment_ssm_1      ... done
Stopping system-modeller-deployment_mongo_1    ... done
Stopping system-modeller-deployment_keycloak_1 ... done
Stopping system-modeller-deployment_adaptor    ... done
  1. Remove the SSM container (using the name from the list in the previous step): docker container rm system-modeller-deployment_ssm_1
  2. The container is created from an underlying image described in the docker-compose.yml file. If the name/tag of the new image is the same as the old (for instance if "latest" is being used), then just do docker-compose pull to update the local image registry. Or, you might want to change the image reference in docker-compose.yml file to a different SSM image (e.g. change "dev" to "master" or change a tagged version).
  3. Bring the system back up again: docker-compose up -d

Backing up Restoring Spyderisk Deployment Contents

It is possible to back up the contents of an existing Spyderisk deployment initiated with docker-compose up -d. A utility bash shell script located in the utils directory facilitates this process.

What Can Be Backed Up?

The backup script preserves the following data:

Important Note

The SSM container is stopped during both the backup and restore operations. The container is restarted at the end of these processes.

Common Use Cases

This backup functionality is particularly useful in scenarios such as:

Performing a Backup

To create a backup, run the following command from the root of your project directory:

./utils/backup_spyderisk.sh backup

Upon completion, a new folder named in the format backup_YY-mm-DD_HH-MM will be generated, containing all the backed-up data.

Restoring from a Backup

To restore Spyderisk from a previously created backup, use the following command:

./utils/backup_spyderisk.sh restore -b <backup_folder_name>

Replace <backup_folder_name> with the actual name of your backup folder.