Chapter 9. Using Operator Lifecycle Manager on restricted networks
When OpenShift Container Platform is installed on restricted networks, Operator Lifecycle Manager (OLM) can no longer use the default OperatorHub sources as they require full Internet connectivity. Cluster administrators can disable those default sources and create local mirrors so that OLM can install and manage Operators from the local sources instead.
9.1. Configuring OperatorHub for restricted networks
Cluster administrators can configure OLM and OperatorHub to use local content in restricted network environments.
Prerequisites
- Cluster administrator access to an OpenShift Container Platform cluster and its internal registry.
- Separate workstation without network restrictions.
- If pushing images to the OpenShift Container Platform cluster’s internal registry, the registry must be exposed with a route.
-
podman
version 1.4.4+
Procedure
Disable the default OperatorSources.
Add
disableAllDefaultSources: true
to the spec:$ oc patch OperatorHub cluster --type json \ -p '[{"op": "add", "path": "/spec/disableAllDefaultSources", "value": true}]'
This disables the default OperatorSources that are configured by default during an OpenShift Container Platform installation.
Retrieve package lists.
To get the list of packages that are available for the default OperatorSources, run the following
curl
commands from your workstation without network restrictions:$ curl https://quay.io/cnr/api/v1/packages?namespace=redhat-operators > packages.txt $ curl https://quay.io/cnr/api/v1/packages?namespace=community-operators >> packages.txt $ curl https://quay.io/cnr/api/v1/packages?namespace=certified-operators >> packages.txt
Each package in the new
packages.txt
is an Operator that you could add to your restricted network catalog. From this list, you could either pull every Operator or a subset that you would like to expose to users.Pull Operator content.
For a given Operator in the package list, you must pull the latest released content:
$ curl https://quay.io/cnr/api/v1/packages/<namespace>/<operator_name>/<release>
This example uses the etcd Operator:
Retrieve the digest:
$ curl https://quay.io/cnr/api/v1/packages/community-operators/etcd/0.0.12
From that JSON, take the digest and use it to pull the gzipped archive:
$ curl -XGET https://quay.io/cnr/api/v1/packages/community-operators/etcd/blobs/sha256/8108475ee5e83a0187d6d0a729451ef1ce6d34c44a868a200151c36f3232822b \ -o etcd.tar.gz
To pull the information out, you must untar the archive into a
manifests/<operator_name>/
directory with all the other Operators that you want. For example, to untar to an existing directory calledmanifests/etcd/
:$ mkdir -p manifests/etcd/ 1 $ tar -xf etcd.tar.gz -C manifests/etcd/
- 1
- Create different subdirectories for each extracted archive so that files are not overwritten by subsequent extractions for other Operators.
Break apart
bundle.yaml
content, if necessary.In your new
manifests/<operator_name>
directory, the goal is to get your bundle in the following directory structure:manifests/ └── etcd ├── 0.0.12 │ ├── clusterserviceversion.yaml │ └── customresourcedefinition.yaml └── package.yaml
If you see files already in this structure, you can skip this step. However, if you instead see only a single file called
bundle.yaml
, you must first break this file up to conform to the required structure.You must separate the CSV content under
data.clusterServiceVersion
(each file in the list), the CRD content underdata.customResourceDefinition
(each file in the list), and the package content underdata.Package
into their own files.For the CSV file creation, find the following lines in the
bundle.yaml
file:data: clusterServiceVersions: |
Omit those lines, but save a new file consisting of the full CSV resource content beginning with the following lines, removing the prepended
-
character:Example
clusterserviceversion.yaml
file snippetapiVersion: operators.coreos.com/v1alpha1 kind: ClusterServiceVersion [...]
For the CRD file creation, find the following line in the
bundle.yaml
file:customResourceDefinitions: |
Omit this line, but save new files consisting of each, full CRD resource content beginning with the following lines, removing the prepended
-
character:Example
customresourcedefinition.yaml
file snippetapiVersion: apiextensions.k8s.io/v1beta1 kind: CustomResourceDefinition [...]
For the package file creation, find the following line in the
bundle.yaml
file:packages: |
Omit this line, but save a new file consisting of the package content beginning with the following lines, removing the prepended
-
character, and ending with apackageName
entry:Example
package.yaml
filechannels: - currentCSV: etcdoperator.v0.9.4 name: singlenamespace-alpha - currentCSV: etcdoperator.v0.9.4-clusterwide name: clusterwide-alpha defaultChannel: singlenamespace-alpha packageName: etcd
Identify images required by the Operators you want to use.
Inspect the CSV files of each Operator for
image:
fields to identify the pull specs for any images required by the Operator, and note them for use in a later step.For example, in the following
deployments
spec of an etcd Operator CSV:spec: serviceAccountName: etcd-operator containers: - name: etcd-operator command: - etcd-operator - --create-crd=false image: quay.io/coreos/etcd-operator@sha256:bd944a211eaf8f31da5e6d69e8541e7cada8f16a9f7a5a570b22478997819943 1 env: - name: MY_POD_NAMESPACE valueFrom: fieldRef: fieldPath: metadata.namespace - name: MY_POD_NAME valueFrom: fieldRef: fieldPath: metadata.name
- 1
- Image required by Operator.
Create an Operator catalog image.
Save the following to a Dockerfile, for example named
custom-registry.Dockerfile
:FROM registry.redhat.io/openshift4/ose-operator-registry:v4.2.24 AS builder COPY manifests manifests RUN /bin/initializer -o ./bundles.db FROM registry.access.redhat.com/ubi7/ubi COPY --from=builder /registry/bundles.db /bundles.db COPY --from=builder /usr/bin/registry-server /registry-server COPY --from=builder /bin/grpc_health_probe /bin/grpc_health_probe EXPOSE 50051 ENTRYPOINT ["/registry-server"] CMD ["--database", "bundles.db"]
Use the
podman
command to create and tag the container image from the Dockerfile:$ podman build -f custom-registry.Dockerfile \ -t <local_registry_host_name>:<local_registry_host_port>/<namespace>/custom-registry 1
- 1
- Tag the image for the internal registry of the restricted network OpenShift Container Platform cluster and any namespace.
Push the Operator catalog image to a registry.
Your new Operator catalog image must be pushed to a registry that the restricted network OpenShift Container Platform cluster can access. This can be the internal registry of the cluster itself or another registry that the cluster has network access to, such as an on-premise Quay Enterprise registry.
For this example, login and push the image to the internal registry OpenShift Container Platform cluster:
$ podman push <local_registry_host_name>:<local_registry_host_port>/<namespace>/custom-registry
Create a CatalogSource pointing to the new Operator catalog image.
Save the following to a file, for example
my-operator-catalog.yaml
:apiVersion: operators.coreos.com/v1alpha1 kind: CatalogSource metadata: name: my-operator-catalog namespace: openshift-marketplace spec: displayName: My Operator Catalog sourceType: grpc image: <local_registry_host_name>:<local_registry_host_port>/<namespace>/custom-registry:latest
Create the CatalogSource resource:
$ oc create -f my-operator-catalog.yaml
Verify the CatalogSource and package manifest are created successfully:
# oc get pods -n openshift-marketplace NAME READY STATUS RESTARTS AGE my-operator-catalog-6njx6 1/1 Running 0 28s marketplace-operator-d9f549946-96sgr 1/1 Running 0 26h # oc get catalogsource -n openshift-marketplace NAME DISPLAY TYPE PUBLISHER AGE my-operator-catalog My Operator Catalog grpc 5s # oc get packagemanifest -n openshift-marketplace NAME CATALOG AGE etcd My Operator Catalog 34s
You should also be able to view them from the OperatorHub page in the web console.
Mirror the images required by the Operators you want to use.
Determine the images defined by the Operator(s) that you are expecting. This example uses the etcd Operator, requiring the
quay.io/coreos/etcd-operator
image.ImportantThis procedure only shows mirroring Operator images themselves and not Operand images, which are the components that an Operator manages. Operand images must be mirrored as well; see each Operator’s documentation to identify the required Operand images.
To use mirrored images, you must first create an ImageContentSourcePolicy for each image to change the source location of the Operator catalog image. For example:
apiVersion: operator.openshift.io/v1alpha1 kind: ImageContentSourcePolicy metadata: name: etcd-operator spec: repositoryDigestMirrors: - mirrors: - <local_registry_host_name>:<local_registry_host_port>/coreos/etcd-operator source: quay.io/coreos/etcd-operator
Use the
oc image mirror
command from your workstation without network restrictions to pull the image from the source registry and push to the internal registry without being stored locally:$ oc image mirror quay.io/coreos/etcd-operator \ <local_registry_host_name>:<local_registry_host_port>/coreos/etcd-operator
You can now install the Operator from the OperatorHub on your restricted network OpenShift Container Platform cluster.
Additional resources
- For details on exposing your OpenShift Container Platform cluster’s internal registry to off-cluster access, see Exposing the registry.
- For details on accessing the internal registry, see Accessing the registry.