Managing custom catalogs
- Prerequisites
- File-based catalogs
- SQLite-based catalogs
- Adding a catalog source to a cluster
- Accessing images for Operators from private registries
- Disabling the default OperatorHub sources
- Removing custom catalogs
Cluster administrators and Operator catalog maintainers can create and managing custom catalogs packaged using the bundle format on Operator Lifecycle Manager (OLM) in OKD.
Kubernetes periodically deprecates certain APIs that are removed in subsequent releases. As a result, Operators are unable to use removed APIs starting with the version of OKD that uses the Kubernetes version that removed the API. If your cluster is using custom catalogs, see Controlling Operator compatibility with OKD versions for more details about how Operator authors can update their projects to help avoid workload issues and prevent incompatible upgrades. |
Additional resources
Prerequisites
- Install the opm CLI.
File-based catalogs
File-based catalogs are the latest iteration of the catalog format in Operator Lifecycle Manager (OLM). It is a plain text-based (JSON or YAML) and declarative config evolution of the earlier SQLite database format, and it is fully backwards compatible.
For more details about the file-based catalog specification, see Operator Framework packaging format.
Creating a file-based catalog image
You can create a catalog image that uses the plain text file-based catalog format (JSON or YAML), which replaces the deprecated SQLite database format. The opm
CLI provides tooling that helps initialize a catalog in the file-based format, render new records into it, and validate that the catalog is valid.
Prerequisites
opm
version 1.18.0+podman
version 1.9.3+A bundle image built and pushed to a registry that supports Docker v2-2
The internal registry of the OKD cluster cannot be used as the target registry because it does not support pushing without a tag, which is required during the mirroring process.
Procedure
Initialize a catalog for a file-based catalog:
Create a directory for the catalog:
$ mkdir <operator_name>-index
Create a Dockerfile that can build a catalog image:
Example
<operator_name>-index.Dockerfile
# The base image is expected to contain
# /bin/opm (with a serve subcommand) and /bin/grpc_health_probe
FROM quay.io/openshift/origin-operator-registry:4.9.0
# Configure the entrypoint and command
ENTRYPOINT ["/bin/opm"]
CMD ["serve", "/configs"]
# Copy declarative config root into image at /configs
ADD <operator_name>-index /configs
# Set DC-specific label for the location of the DC root directory
# in the image
LABEL operators.operatorframework.io.index.configs.v1=/configs
The Dockerfile must be in the same parent directory as the catalog directory that you created in the previous step:
Example directory structure
.
├── <operator_name>-index
└── <operator_name>-index.Dockerfile
Populate the catalog with your package definition:
$ opm init <operator_name> \ (1)
--default-channel=preview \ (2)
--description=./README.md \ (3)
--icon=./operator-icon.svg \ (4)
--output yaml \ (5)
> <operator_name>-index/index.yaml (6)
1 Operator, or package, name. 2 Channel that subscription will default to if unspecified. 3 Path to the Operator’s README.md
or other documentation.4 Path to the Operator’s icon. 5 Output format: JSON or YAML. 6 Path for creating the catalog configuration file. This command generates an
olm.package
declarative config blob in the specified catalog configuration file.
Add a bundle to the catalog:
$ opm render <registry>/<namespace>/<bundle_image_name>:<tag> \ (1)
--output=yaml \
>> <operator_name>-index/index.yaml (2)
1 Pull spec for the bundle image. 2 Path to the catalog configuration file. The
opm render
command generates a declarative config blob from the provided catalog images and bundle images.Channels must contain at least one bundle.
Add a channel entry for the bundle. For example, modify the following example to your specifications, and add it to your
<operator_name>-index/index.yaml
file:Example channel entry
---
schema: olm.channel
package: <operator_name>
name: preview
entries:
- name: <operator_name>.v0.1.0 (1)
1 Ensure that you include the period ( .
) after<operator_name>
but before thev
in the version. Otherwise, the entry will fail to pass theopm validate
command.Validate the file-based catalog:
Run the
opm validate
command against the catalog directory:$ opm validate <operator_name>-index
Check that the error code is
0
:$ echo $?
Example output
0
Build the catalog image:
$ podman build . \
-f <operator_name>-index.Dockerfile \
-t <registry>/<namespace>/<catalog_image_name>:<tag>
Push the catalog image to a registry:
If required, authenticate with your target registry:
$ podman login <registry>
Push the catalog image:
$ podman push <registry>/<namespace>/<catalog_image_name>:<tag>
SQLite-based catalogs
The SQLite database format for Operator catalogs is a deprecated feature. Deprecated functionality is still included in OKD and continues to be supported; however, it will be removed in a future release of this product and is not recommended for new deployments. For the most recent list of major functionality that has been deprecated or removed within OKD, refer to the Deprecated and removed features section of the OKD release notes. |
Creating a SQLite-based index image
You can create an index image based on the SQLite database format by using the opm
CLI.
Prerequisites
opm
version 1.18.0+podman
version 1.9.3+A bundle image built and pushed to a registry that supports Docker v2-2
The internal registry of the OKD cluster cannot be used as the target registry because it does not support pushing without a tag, which is required during the mirroring process.
Procedure
Start a new index:
$ opm index add \
--bundles <registry>/<namespace>/<bundle_image_name>:<tag> \(1)
--tag <registry>/<namespace>/<index_image_name>:<tag> \(2)
[--binary-image <registry_base_image>] (3)
1 Comma-separated list of bundle images to add to the index. 2 The image tag that you want the index image to have. 3 Optional: An alternative registry base image to use for serving the catalog. Push the index image to a registry.
If required, authenticate with your target registry:
$ podman login <registry>
Push the index image:
$ podman push <registry>/<namespace>/<index_image_name>:<tag>
Updating a SQLite-based index image
After configuring OperatorHub to use a catalog source that references a custom index image, cluster administrators can keep the available Operators on their cluster up to date by adding bundle images to the index image.
You can update an existing index image using the opm index add
command.
Prerequisites
opm
version 1.18.0+podman
version 1.9.3+An index image built and pushed to a registry.
An existing catalog source referencing the index image.
Procedure
Update the existing index by adding bundle images:
$ opm index add \
--bundles <registry>/<namespace>/<new_bundle_image>@sha256:<digest> \(1)
--from-index <registry>/<namespace>/<existing_index_image>:<tag> \(2)
--tag <registry>/<namespace>/<existing_index_image>:<tag> \(3)
--pull-tool podman (4)
1 A comma-separated list of additional bundle images to add to the index. 2 The existing index that was previously pushed. 3 The image tag that you want the updated index image to have. 4 A tool that pulls images for opm index add
.Push the updated index image:
$ podman push <registry>/<namespace>/<existing_index_image>:<tag>
After Operator Lifecycle Manager (OLM) automatically polls the index image referenced in the catalog source at its regular interval, verify that the new packages are successfully added:
$ oc get packagemanifests -n openshift-marketplace
Pruning a SQLite-based index image
An index image, based on the Operator bundle format, is a containerized snapshot of an Operator catalog. You can prune an index of all but a specified list of packages, which creates a copy of the source index containing only the Operators that you want.
Prerequisites
podman
version 1.9.3+grpcurl (third-party command-line tool)
opm
version 1.18.0+Access to a registry that supports Docker v2-2
The internal registry of the OKD cluster cannot be used as the target registry because it does not support pushing without a tag, which is required during the mirroring process.
Procedure
Authenticate with your target registry:
$ podman login <target_registry>
Determine the list of packages you want to include in your pruned index.
Run the source index image that you want to prune in a container. For example:
$ podman run -p50051:50051 \
-it quay.io/operatorhubio/catalog:latest
Example output
Trying to pull quay.io/operatorhubio/catalog:latest...
Getting image source signatures
Copying blob ae8a0c23f5b1 done
...
INFO[0000] serving registry database=/database/index.db port=50051
In a separate terminal session, use the
grpcurl
command to get a list of the packages provided by the index:$ grpcurl -plaintext localhost:50051 api.Registry/ListPackages > packages.out
Inspect the
packages.out
file and identify which package names from this list you want to keep in your pruned index. For example:Example snippets of packages list
...
{
"name": "couchdb-operator"
}
...
{
"name": "eclipse-che"
}
...
{
{
"name": "etcd"
}
...
In the terminal session where you executed the
podman run
command, press Ctrl and C to stop the container process.
Run the following command to prune the source index of all but the specified packages:
$ opm index prune \
-f quay.io/operatorhubio/catalog:latest \(1)
-p couchdb-operator,eclipse-che,etcd \(2)
[-i quay.io/openshift/origin-operator-registry:4.9.0] \(3)
-t <target_registry>:<port>/<namespace>/catalog:latest (4)
1 Index to prune. 2 Comma-separated list of packages to keep. 3 Required only for IBM Power Systems and IBM Z images: Operator Registry base image with the tag that matches the target OKD cluster major and minor version. 4 Custom tag for new index image being built. Run the following command to push the new index image to your target registry:
$ podman push <target_registry>:<port>/<namespace>/catalog:latest
where
<namespace>
is any existing namespace on the registry.
Adding a catalog source to a cluster
Adding a catalog source to an OKD cluster enables the discovery and installation of Operators for users. Cluster administrators can create a CatalogSource
object that references an index image. OperatorHub uses catalog sources to populate the user interface.
Prerequisites
- An index image built and pushed to a registry.
Procedure
Create a
CatalogSource
object that references your index image.Modify the following to your specifications and save it as a
catalogSource.yaml
file:apiVersion: operators.coreos.com/v1alpha1
kind: CatalogSource
metadata:
name: my-operator-catalog
namespace: olm (1)
annotations:
olm.catalogImageTemplate: (2)
"<registry>/<namespace>/<index_image_name>:v{kube_major_version}.{kube_minor_version}.{kube_patch_version}"
spec:
sourceType: grpc
image: <registry>/<namespace>/<index_image_name>:<tag> (3)
displayName: My Operator Catalog
publisher: <publisher_name> (4)
updateStrategy:
registryPoll: (5)
interval: 30m
1 If you want the catalog source to be available globally to users in all namespaces, specify the olm
namespace. Otherwise, you can specify a different namespace for the catalog to be scoped and available only for that namespace.2 Optional: Set the olm.catalogImageTemplate
annotation to your index image name and use one or more of the Kubernetes cluster version variables as shown when constructing the template for the image tag.3 Specify your index image. 4 Specify your name or an organization name publishing the catalog. 5 Catalog sources can automatically check for new versions to keep up to date. Use the file to create the
CatalogSource
object:$ oc apply -f catalogSource.yaml
Verify the following resources are created successfully.
Check the pods:
$ oc get pods -n olm
Example output
NAME READY STATUS RESTARTS AGE
my-operator-catalog-6njx6 1/1 Running 0 28s
marketplace-operator-d9f549946-96sgr 1/1 Running 0 26h
Check the catalog source:
$ oc get catalogsource -n olm
Example output
NAME DISPLAY TYPE PUBLISHER AGE
my-operator-catalog My Operator Catalog grpc 5s
Check the package manifest:
$ oc get packagemanifest -n olm
Example output
NAME CATALOG AGE
jaeger-product My Operator Catalog 93s
You can now install the Operators from the OperatorHub page on your OKD web console.
Additional resources
See Operator Lifecycle Manager concepts and resources → Catalog source for more details on the
CatalogSource
object spec.If your index image is hosted on a private registry and requires authentication, see Accessing images for Operators from private registries.
Accessing images for Operators from private registries
If certain images relevant to Operators managed by Operator Lifecycle Manager (OLM) are hosted in an authenticated container image registry, also known as a private registry, OLM and OperatorHub are unable to pull the images by default. To enable access, you can create a pull secret that contains the authentication credentials for the registry. By referencing one or more pull secrets in a catalog source, OLM can handle placing the secrets in the Operator and catalog namespace to allow installation.
Other images required by an Operator or its Operands might require access to private registries as well. OLM does not handle placing the secrets in target tenant namespaces for this scenario, but authentication credentials can be added to the global cluster pull secret or individual namespace service accounts to enable the required access.
The following types of images should be considered when determining whether Operators managed by OLM have appropriate pull access:
Index images
A CatalogSource
object can reference an index image, which use the Operator bundle format and are catalog sources packaged as container images hosted in images registries. If an index image is hosted in a private registry, a secret can be used to enable pull access.
Bundle images
Operator bundle images are metadata and manifests packaged as container images that represent a unique version of an Operator. If any bundle images referenced in a catalog source are hosted in one or more private registries, a secret can be used to enable pull access.
Operator and Operand images
If an Operator installed from a catalog source uses a private image, either for the Operator image itself or one of the Operand images it watches, the Operator will fail to install because the deployment will not have access to the required registry authentication. Referencing secrets in a catalog source does not enable OLM to place the secrets in target tenant namespaces in which Operands are installed.
Instead, the authentication details can be added to the global cluster pull secret in the openshift-config
namespace, which provides access to all namespaces on the cluster. Alternatively, if providing access to the entire cluster is not permissible, the pull secret can be added to the default
service accounts of the target tenant namespaces.
Prerequisites
At least one of the following hosted in a private registry:
An index image or catalog image.
An Operator bundle image.
An Operator or Operand image.
Procedure
Create a secret for each required private registry.
Log in to the private registry to create or update your registry credentials file:
$ podman login <registry>:<port>
The file path of your registry credentials can be different depending on the container tool used to log in to the registry. For the
podman
CLI, the default location is${XDG_RUNTIME_DIR}/containers/auth.json
. For thedocker
CLI, the default location is/root/.docker/config.json
.It is recommended to include credentials for only one registry per secret, and manage credentials for multiple registries in separate secrets. Multiple secrets can be included in a
CatalogSource
object in later steps, and OKD will merge the secrets into a single virtual credentials file for use during an image pull.A registry credentials file can, by default, store details for more than one registry or for multiple repositories in one registry. Verify the current contents of your file. For example:
File storing credentials for multiple registries
{
"auths": {
"registry.redhat.io": {
"auth": "FrNHNydQXdzclNqdg=="
},
"quay.io": {
"auth": "fegdsRib21iMQ=="
}
},
"https://quay.io/my-namespace/my-user/my-image": {
"auth": "eWfjwsDdfsa221=="
}
},
"https://quay.io/my-namespace/my-user": {
"auth": "feFweDdscw34rR=="
}
},
"https://quay.io/my-namespace": {
"auth": "frwEews4fescyq=="
}
}
}
Because this file is used to create secrets in later steps, ensure that you are storing details for only one registry per file. This can be accomplished by using either of the following methods:
Use the
podman logout <registry>
command to remove credentials for additional registries until only the one registry you want remains.Edit your registry credentials file and separate the registry details to be stored in multiple files. For example:
File storing credentials for one registry
{
"auths": {
"registry.redhat.io": {
"auth": "FrNHNydQXdzclNqdg=="
}
}
}
File storing credentials for another registry
{
"auths": {
"quay.io": {
"auth": "Xd2lhdsbnRib21iMQ=="
}
}
}
Create a secret in the
openshift-marketplace
namespace that contains the authentication credentials for a private registry:$ oc create secret generic <secret_name> \
-n openshift-marketplace \
--from-file=.dockerconfigjson=<path/to/registry/credentials> \
--type=kubernetes.io/dockerconfigjson
Repeat this step to create additional secrets for any other required private registries, updating the
--from-file
flag to specify another registry credentials file path.
Create or update an existing
CatalogSource
object to reference one or more secrets:apiVersion: operators.coreos.com/v1alpha1
kind: CatalogSource
metadata:
name: my-operator-catalog
namespace: openshift-marketplace
spec:
sourceType: grpc
secrets: (1)
- "<secret_name_1>"
- "<secret_name_2>"
image: <registry>:<port>/<namespace>/<image>:<tag>
displayName: My Operator Catalog
publisher: <publisher_name>
updateStrategy:
registryPoll:
interval: 30m
1 Add a spec.secrets
section and specify any required secrets.If any Operator or Operand images that are referenced by a subscribed Operator require access to a private registry, you can either provide access to all namespaces in the cluster, or individual target tenant namespaces.
To provide access to all namespaces in the cluster, add authentication details to the global cluster pull secret in the
openshift-config
namespace.Cluster resources must adjust to the new global pull secret, which can temporarily limit the usability of the cluster.
Extract the
.dockerconfigjson
file from the global pull secret:$ oc extract secret/pull-secret -n openshift-config --confirm
Update the
.dockerconfigjson
file with your authentication credentials for the required private registry or registries and save it as a new file:$ cat .dockerconfigjson | \
jq --compact-output '.auths["<registry>:<port>/<namespace>/"] |= . + {"auth":"<token>"}' \(1)
> new_dockerconfigjson
1 Replace <registry>:<port>/<namespace>
with the private registry details and<token>
with your authentication credentials.Update the global pull secret with the new file:
$ oc set data secret/pull-secret -n openshift-config \
--from-file=.dockerconfigjson=new_dockerconfigjson
To update an individual namespace, add a pull secret to the service account for the Operator that requires access in the target tenant namespace.
Recreate the secret that you created for the
openshift-marketplace
in the tenant namespace:$ oc create secret generic <secret_name> \
-n <tenant_namespace> \
--from-file=.dockerconfigjson=<path/to/registry/credentials> \
--type=kubernetes.io/dockerconfigjson
Verify the name of the service account for the Operator by searching the tenant namespace:
$ oc get sa -n <tenant_namespace> (1)
1 If the Operator was installed in an individual namespace, search that namespace. If the Operator was installed for all namespaces, search the openshift-operators
namespace.Example output
NAME SECRETS AGE
builder 2 6m1s
default 2 6m1s
deployer 2 6m1s
etcd-operator 2 5m18s (1)
1 Service account for an installed etcd Operator. Link the secret to the service account for the Operator:
$ oc secrets link <operator_sa> \
-n <tenant_namespace> \
<secret_name> \
--for=pull
Additional resources
See What is a secret? for more information on the types of secrets, including those used for registry credentials.
See Updating the global cluster pull secret for more details on the impact of changing this secret.
See Allowing pods to reference images from other secured registries for more details on linking pull secrets to service accounts per namespace.
Disabling the default OperatorHub sources
Operator catalogs that source content provided by Red Hat and community projects are configured for OperatorHub by default during an OKD installation. As a cluster administrator, you can disable the set of default catalogs.
Procedure
Disable the sources for the default catalogs by adding
disableAllDefaultSources: true
to theOperatorHub
object:$ oc patch OperatorHub cluster --type json \
-p '[{"op": "add", "path": "/spec/disableAllDefaultSources", "value": true}]'
Alternatively, you can use the web console to manage catalog sources. From the Administration → Cluster Settings → Configuration → OperatorHub page, click the Sources tab, where you can create, delete, disable, and enable individual sources. |
Removing custom catalogs
As a cluster administrator, you can remove custom Operator catalogs that have been previously added to your cluster by deleting the related catalog source.
Procedure
In the Administrator perspective of the web console, navigate to Administration → Cluster Settings.
Click the Configuration tab, and then click OperatorHub.
Click the Sources tab.
Select the Options menu for the catalog that you want to remove, and then click Delete CatalogSource.