Scripted Deployment⚓︎

The Scripted Deployment provides a demonstration of an example deployment, and can found in the subdirectory deployment-guide/deploy of the source repository for this guide…

git clone https://github.com/EOEPCA/deployment-guide \
&& cd deployment-guide \
&& ls deploy

The script deploy/eoepca/eoepca.sh acts as an entry-point to the full system deployment. In order to tailor the deployment for your target environment, the script is configured through environment variables and command-line arguments. By default the script assumes deployment to a local minikube.

NOTE that the scripted deployment assumes that installation of the Prerequisite Tooling has been performed.

The following subsections lead through the steps for a full local deployment. Whilst minikube is assumed, minimal adaptions are required to make the deployment to your existing Kubernetes cluster.

The deployment follows these broad steps:

Configuration
Tailoring of deployment options.
Deployment
Creation of cluster and deployment of eoepca services.
Protection
Application of protection for authorized access to services.

The Protection step is split from Deployment as there are some manual steps to be performed before the Protection can be applied.

Configuration⚓︎

The script deploy/eoepca/eoepca.sh is configured by some environment variables and command-line arguments.

Environment Variables⚓︎

Variable	Description	Default
REQUIRE_<cluster-component>	A set of variables that can be used to control which CLUSTER components are deployed by the script, as follows (with defaults): `REQUIRE_MINIKUBE=true` `REQUIRE_INGRESS_NGINX=true` `REQUIRE_CERT_MANAGER=true` `REQUIRE_LETSENCRYPT=true` `REQUIRE_SEALED_SECRETS=false` `REQUIRE_MINIO=false`	see description
REQUIRE_<eoepca-component>	A set of variables that can be used to control which EOEPCA components are deployed by the script, as follows (with defaults): `REQUIRE_STORAGE=true` `REQUIRE_DUMMY_SERVICE=false` `REQUIRE_LOGIN_SERVICE=true` `REQUIRE_PDP=true` `REQUIRE_USER_PROFILE=true` `REQUIRE_ADES=true` `REQUIRE_RESOURCE_CATALOGUE=true` `REQUIRE_DATA_ACCESS=true` `REQUIRE_WORKSPACE_API=true` `REQUIRE_BUCKET_OPERATOR=true` `REQUIRE_HARBOR=true` `REQUIRE_PORTAL=true` `REQUIRE_PDE=true`	see description
REQUIRE_<protection-component>	A set of variables that can be used to control which PROTECTION components are deployed by the script, as follows (with defaults): `REQUIRE_DUMMY_SERVICE_PROTECTION=false` `REQUIRE_ADES_PROTECTION=true` `REQUIRE_RESOURCE_CATALOGUE_PROTECTION=true` `REQUIRE_DATA_ACCESS_PROTECTION=true` `REQUIRE_WORKSPACE_API_PROTECTION=true`	see description
MINIKUBE_KUBERNETES_VERSION	The Kubernetes version to be used by minikube Note that the EOEPCA development has been conducted primarily using version 1.22.5.	`v1.22.5`
MINIKUBE_MEMORY_AMOUNT	Amount of memory to allocate to the docker containers used by minikube to implement the cluster.	`12g`
USE_METALLB	Enable use of minikube’s built-in load-balancer. The load-balancer can be used to facilitate exposing services publicly. However, the same can be achieved using minikube’s built-in ingress-controller. Therefore, this option is suppressed by default.	`false`
USE_INGRESS_NGINX_HELM	Install the ingress-nginx controller using the published helm chart, rather than relying upon the version that is built-in to minikube. By default we prefer the version that is built in to minikube.	`false`
USE_INGRESS_NGINX_LOADBALANCER	Patch the built-in minikube nginx-ingress-controller to offer a service of type `LoadBalancer`, rather than the default `NodePort`. It was initially thought that this would be necessary to achieve public access to the ingress services - but was subsequently found that the default `NodePort` configuration of the ingress-controller was sufficient. This option is left in case it proves useful. Only applicable for `USE_INGRESS_NGINX_HELM=false` (i.e. when using the minikube built-in )	`false`
OPEN_INGRESS	Create ‘open’ ingress endpoints that are not subject to authorization protection. For a secure system the open endpoints should be disabled (`false`) and access to resource should be protected via ingress that apply protection	`false`
USE_TLS	Indicates whether TLS will be configured for service `Ingress` rules. If not (i.e. `USE_TLS=false`), then the ingress-controller is configured to disable `ssl-redirect`, and `TLS_CLUSTER_ISSUER=notls` is set.	`true`
TLS_CLUSTER_ISSUER	The name of the ClusterIssuer to satisfy ingress tls certificates. Out-of-the-box ClusterIssuer instances are configured in the file `deploy/cluster/letsencrypt.sh`.	`letsencrypt-staging`
LOGIN_SERVICE_ADMIN_PASSWORD	Initial password for the `admin` user in the login-service.	`changeme`
MINIO_ROOT_USER	Name of the ‘root’ user for the Minio object storage service.	`eoepca`
MINIO_ROOT_PASSWORD	Password for the ‘root’ user for the Minio object storage service.	`changeme`
HARBOR_ADMIN_PASSWORD	Password for the ‘admin’ user for the Harbor artefact registry service.	`changeme`
STAGEOUT_TARGET	Configures the ADES with the destination to which it should push processing results: `workspace` - via the Workspace API `minio` - to minio S3 object storage	`workspace`
INSTALL_FLUX	The Workspace API relies upon Flux CI/CD, and has the capability to install the required flux components to the cluster. If your deployment already has flux installed then set this value `false` to suppress the Workspace API flux install	`true`
CREODIAS_DATA_SPECIFICATION	Apply the data specification to harvest from the CREODIAS data offering into the resource-catalogue and data-access services. Can only be used when running in the CREODIAS (Cloudferro) cloud, with access to the `eodata` network.	`false`

Openstack Configuration⚓︎

There are some additional environment variables that configure the BucketOperator with details of the infrastructure Openstack layer.

NOTE that this is only applicable for an Openstack deployment and has only been tested on the CREODIAS.

Variable	Description	Default
OS_DOMAINNAME	Openstack domain of the admin account in the cloud provider.	`cloud_XXXXX`
OS_USERNAME	Openstack username of the admin account in the cloud provider.	`user@cloud.com`
OS_PASSWORD	Openstack password of the admin account in the cloud provider.	`none`
OS_MEMBERROLEID	ID of a specific role (e.g. the ‘member‘ role) for operations users (to allow administration), e.g. `7fe2ff9ee5384b1894a90838d3e92bab`.	`none`
OS_SERVICEPROJECTID	ID of a project containing the user identity requiring write access to the created user buckets, e.g. `573916ef342a4bf1aea807d0c6058c1e`.	`none`
USER_EMAIL_PATTERN	Email associated to the created user within the created user project. Note: `<name>` is templated and will be replaced.	`eoepca-<name>@platform.com`

Command-line Arguments⚓︎

The eoepca.sh script is further configured via command-line arguments…

Usage: eoepca.sh <action> <cluster-name> <public-ip> <domain>

Argument	Description	Default
action	Action to perform: `apply` \| `delete` \| `template`. `apply` makes the deployment `delete` removes the deployment `template` outputs generated kubernetes yaml to stdout	`apply`
cluster-name	The name of the minikube ‘profile’ for the created minikube cluster	`eoepca`
public-ip	The public IP address through which the deployment is exposed via the ingress-controller. By default, the value is deduced from the assigned cluster minikube IP address - ref. command `minikube ip`.	`<minikube-ip>`
domain	The DNS domain name through which the deployment is accessed. Forms the stem for all service hostnames in the ingress rules - i.e. `<service-name>.<domain>`. By default, the value is deduced from the assigned cluster minikube IP address, using `nip.io` to establish a DNS lookup - i.e. `<minikube ip>.nip.io`.	`<minikube ip>.nip.io`

Public Deployment⚓︎

For simplicity, the out-of-the-box scripts assume a ‘private’ deployment - with no public IP / DNS and hence no use of TLS for service ingress endpoints.

In the case that an external-facing public deployment is desired, then the following configuration selections should be made:

public_ip - set to the external-facing public IP of your deployment, e.g. the floating IP of your load-balancer in a cloud deployment
domain - set to the domain (as per DNS records) for your deployment
Note that the EOEPCA components typically configure their ingress with hostname prefixes applied to this domain. Thus, it is necessary that the DNS record for the domain is established as a wildcard record - i.e. *.<domain>
USE_TLS=true - to enable configuration of TLS endpoints in each component service ingress
TLS_CLUSTER_ISSUER=<issuer> - should be configured ~ e.g. using the letsencrypt-production or letsencrypt-staging (testing only) Cluster Issuer that are configured by the scripted deployment

Deployment⚓︎

The deployment is initiated by setting the appropriate environment variables and invoking the eoepca.sh script with suitable command-line arguments. You may find it convenient to do so using a wrapper script that customises the environment varaibles according to your cluster, and then invokes the eoepca.sh script.

Customised examples are provided for Simple, CREODIAS and Processing deployments.

NOTE that if a prior deployment has been attempted then, before redeploying, a clean-up should be performed as described in the Clean-up section below. This is particularly important in the case that the minikube none driver is used, as the persistence is maintained on the host and so is not naturally removed when the minikube cluster is destroyed.

Initiate the deployment…

./deploy/eoepca/eoepca.sh apply "<cluster-name>" "<public-ip>" "<domain>"

The deployment takes 10+ minutes - depending on the resources of your host/cluster. The progress can be monitored…

kubectl get pods -A

The deployment is ready once all pods are either Running or Completed. This can be further confirmed by accessing the login-service web interface at https://auth.<domain>/ and logging in as user admin using the credentials configured via LOGIN_SERVICE_ADMIN_PASSWORD.

Default Credentials⚓︎

By default, the Login Service is accessed at the URL https://auth.<domain>/ with the credentials…

username: admin
password: Chang3me!

…unless the password is overridden via the variable LOGIN_SERVICE_ADMIN_PASSWORD.

Minio Object Storage⚓︎

By default, Minio is accessed at the URL http://minio-console.<domain>/ with the credentials…

username: eoepca
password: changeme

…unless the username/password are overridden via the variables MINIO_ROOT_USER and MINIO_ROOT_PASSWORD.

Harbor Container Registry⚓︎

By default, Harbor is accessed at the URL https://harbor.<domain>/ with the credentials…

username: admin
password: changeme

…unless the password is overridden via the variable HARBOR_ADMIN_PASSWORD.

Protection⚓︎

The protection of resource server endpoints is applied with the script deploy/eoepca/eoepca-protection.sh. This script should be executed with environment variables and command-line options that are consistent with those of the main deployment (ref. script eoepca.sh).

The script eoepca-protection.sh introduces two users eric and bob to demonstrate the application of authorized access to various service endpoints: ADES, Workspace API and dummy-service (simple endpoint used for debugging).

Thus, the users must first be created in the login-service and their unique IDs passed to the protection script.

Usage: eoepca-protection.sh <action> <eric-id> <bob-id> <public-ip> <domain>

Create Test Users⚓︎

Access the login-service web interface (https://auth.<domain>/) as user admin using the credentials configured via LOGIN_SERVICE_ADMIN_PASSWORD.

Select Users -> Add person to add users eric and bob (dummy details can be used). Note the Inum (unique user ID) for each user for use with the eoepca-protection.sh script.

Apply Protection⚓︎

Apply the protection…
Ensure that the script is executed with the environment variables and command-line options that are consistent with those of the main deployment.

./deploy/eoepca/eoepca-protection.sh apply "<eric-id>" "<bob-id>" "<public-ip>" "<domain>"

Create User Workspaces⚓︎

The protection steps created the test users eric and bob. For completeness we use the Workspace API to create their user workspaces, which hold their personal resources (data, processing results, etc.) within the platform - see Workspace.

Using Workspace Swagger UI⚓︎

The Workspace API provides a Swagger UI that facilitates interaction with the API - at the URL https://workspace-api.<domain>/docs#. Access to The Workspace API is protected, such that the necessary access tokens must be supplied in requests, which is most easily achieved by logging in via the ‘portal’.

The portal is accessed at https://portal.<domain>/. It is a rudimentary web service that facilitates establishing the appropriate tokens in the user’s browser context. Login to the portal as the admin user, using the configured credentials.

Access the Workspace Swagger UI at https://workspace-api.<domain>/docs. Workspaces are created using POST /workspaces (Create Workspace). Expand the node and select Try it out. Complete the request body, such as…

{
  "preferred_name": "eric",
  "default_owner": "d95b0c2b-ea74-4b3f-9c6a-85198dec974d"
}

…where the default_owner is the user ID (Inum) for the user - thus protecting the created workspace for the identified user.

Using `curl`⚓︎

The same can be achieved with a straight http request, for example using curl…

curl -X 'POST' \
  'https://workspace-api.192.168.49.2.nip.io/workspaces' \
  -H 'accept: application/json' \
  -H 'Content-Type: application/json' \
  -H 'X-User-Id: <admin-id-token>' \
  -d '{
  "preferred_name": "<workspace-name>",
  "default_owner": "<user-inum>"
}'

Values must be provided for:

admin-id-token - User ID token for the admin user
workspace-name - name of the workspace, typically the username
user-inum - the ID of the user for which the created workspace will be protected

The ID token for the admin user can be obtained with a call to the token endpoint of the Login Service - supplying the credentials for the admin user and the pre-registered client…

curl -L -X POST 'https://auth.<domain>/oxauth/restv1/token' \
  -H 'Cache-Control: no-cache' \
  -H 'Content-Type: application/x-www-form-urlencoded' \
  --data-urlencode 'scope=openid user_name is_operator' \
  --data-urlencode 'grant_type=password' \
  --data-urlencode 'username=admin' \
  --data-urlencode 'password=<admin-password>' \
  --data-urlencode 'client_id=<client-id>' \
  --data-urlencode 'client_secret=<client-secret>'

A json response is returned, in which the field id_token provides the user ID token for the admin user.

Using `create-workspace` helper script⚓︎

As an aide there is a helper script create-workspace. The script is available in the deployment-guide repository, and can be obtained as follows…

git clone git@github.com:EOEPCA/deployment-guide
cd deployment-guide

The create-workspace helper script requires some command-line arguments…

Usage:
  create-workspace <domain> <user> <user-inum> [<client-id> <client-secret>]

For example…

./deploy/bin/create-workspace 192.168.49.2.nip.io eric d95b0c2b-ea74-4b3f-9c6a-85198dec974d

The script prompts for the password of the admin user.

By default <client-id> and <client-secret> are read from the client.yaml file that is created by the deployment script, which auto-registers a Login Service client. Thus, these args can be ommited to use the default client credentials.

Clean-up⚓︎

Before initiating a fresh deployment, if a prior deployment has been attempted, then it is necessary to remove any persistent artefacts of the prior deployment. This includes…

Minikube cluster
Delete the minikube cluster…
minikube delete
If necessary specify the cluster (profile)…
minikube -p <profile> delete
Persistent Data
In the case that the minikube none driver is used, the persistence is maintained on the host and so is not naturally removed when the minikube cluster is destroyed. In this case, the minikube standard StorageClass is fulfilled by the hostpath provisioner, whose persistence is removed as follows…
sudo rm -rf /tmp/hostpath-provisioner
Client Credentials
During the deployment a client of the Authorisation Server is registered, and its credentials stored for reuse in the file client.yaml. Once the cluster has been destroyed, then these client credentials become stale and so should be removed to avoid polluting subsequent deployments…
rm -rf ./deploy/eoepca/client.yaml

There is a helper script clean that can be used for steps 2 and 3 above, (the script does not delete the cluster).

./deploy/cluster/clean