When running on a heterogeneous cluster, a cluster with nodes of different
architectures, the DataImportCronTemplates (DICTs; also known as golden
images), in the hyperconverged cluster operator (HCO) should be annotated with
the ssp.kubevirt.io/dict.architectures annotation, where the value is the
list of the architectures supported by the image, that is defined in each DICT.
For pre-defined DICTs, this annotation is already set, but for custom DICTs (user defined DICTs), this annotation must be set by the user in the HyperConverged custom resource (CR).
For each DICT, if the ssp.kubevirt.io/dict.architectures annotation is
missing, Then HCO will trigger the HCOGoldenImageWithNoArchitectureAnnotation
alert for this specific DICT.
Note: This alert is only triggered, if the
enableMultiArchBootImageImportfeature gate is enabled in the HyperConverged CR.
When this alert is triggered, it means that the golden image created for this DICT, is with undefined architecture. There is a risk that when this image will be used as a boot image for a virtual machine, and the virtual machine will be scheduled on a node with a CPU architecture different than the image architecture,the virtual machine will fail to start.
Read the HyperConverged CR:
# Get the namespace of the HyperConverged CR
$ NAMESPACE="$(kubectl get hyperconverged -A --no-headers | awk '{print $1}')"
#Read the HyperConverged CR
$ kubectl get hyperconverged -n "${NAMESPACE}" -o yaml
Go over the output of the command. If there are DICT objects under the
spec.dataImportCronTemplates field in the HyperConverged CR, then for each
one of them, check if the ssp.kubevirt.io/dict.architectures annotation is
set. If the annotation is not set, then this alert is triggered.
Below is an example for a HyperConverged CR with a valid DICT with the
ssp.kubevirt.io/dict.architectures annotation set:
apiVersion: hco.kubevirt.io/v1beta1
kind: HyperConverged
...
spec:
...
dataImportCronTemplates:
- metadata:
annotations:
...
ssp.kubevirt.io/dict.architectures: amd64
name: the-name-of-the-dict
spec:
...
The status.nodeInfo.workloadsArchitectures shows the list of architectures
that are supported by the cluster.
User-defined DICTs are defined in the HyperConverged CR, in the
spec.dataImportCronTemplates field.
First, check what architectures are supported by the image. You can use the following command:
$ podman manifest inspect your-registry/your-image:latest
See here for the podman manifest inspect documentation.
If the image is multi architecture manifest (fat manifest), it will include the
manifests field, which is a list of architectures supported by the image. If
the image is not a multi architecture manifest, you will need to find out what
is its architecture.
Then, edit the HyperConverged CR, to add the missing ssp.kubevirt.io/dict.architectures
annotation.
The format of the annotation is a comma-separated list of architectures;
e.g., amd64,arm64,s390x.
If the image does not support any of the architectures supported by the cluster, you will need to either rebuild the image for one or more of the architectures supported by the cluster, or remove the DICT from the HyperConverged CR.
Find some more information about building multi-architecture images, see the podman documentation.
If you cannot resolve the issue, see the following resources: