You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Describe the bug
Currently, even in release branches, imagestream tags have referencePolicy: Source set, meaning images are always pulled from the external registry source, even when an internal openshift registry is present.
Together with imagePullPolicy: Always, this decreases stability and introduces a potential point of failure.
Q: "Does it work if the external registry is unreachable?"
E.g. during a temporary network outage or when the external registry is down?
No, because the Pod uses an external url.
Expected behavior
When internal openshift registry is present, images should be cached in the internal openshift registry, which is what happens with tag referencePolicy: Local. When it is not present, images should be pulled from external.
Workaround (if any)
Manually modified imagestreams in odh-manifests before applying to cluster.
Open Data Hub Version
1.7.0
OpenShift Version
4.10.x on-prem OCP
Additional context
The correct way with dealing with internal openshift registry being present or not present is not setting referencePolicy: Source. The one exception being tag-notation references in tag.from.name in DEV environments. There, referencePolicy: Source can make sense, possibly also in combination with regularly-updates digests instead of tags in from.name, as you currently do in master branch.
It is to use image change trigger and letting Openshift resolve the correct location of images by means of that mechanism.
That requires PR-133 kubeflow notebook controller and PR-800 odh dashboard to be merged in.
Behavior of that mechanism:
The text was updated successfully, but these errors were encountered:
shalberd
changed the title
change release branch imagestream tag referencePolicy Local instead of Source
change release branch imagestream tag referencePolicy to Local instead of Source
Jul 6, 2023
Describe the bug
Currently, even in release branches, imagestream tags have referencePolicy: Source set, meaning images are always pulled from the external registry source, even when an internal openshift registry is present.
Together with imagePullPolicy: Always, this decreases stability and introduces a potential point of failure.
https://github.com/opendatahub-io/odh-manifests/blob/v1.7/notebook-images/base/jupyter-datascience-notebook-imagestream.yaml#L27
Steps To Reproduce
Disallow traffic to quai.io in proxy server referenced at Cluster Config.
Scenario "external image" https://itnext.io/variations-on-imagestreams-in-openshift-4-f8ee5e8be633 descibed at
Q: "Does it work if the external registry is unreachable?"
E.g. during a temporary network outage or when the external registry is down?
No, because the Pod uses an external url.
Expected behavior
When internal openshift registry is present, images should be cached in the internal openshift registry, which is what happens with tag referencePolicy: Local. When it is not present, images should be pulled from external.
Workaround (if any)
Manually modified imagestreams in odh-manifests before applying to cluster.
Open Data Hub Version
1.7.0
OpenShift Version
4.10.x on-prem OCP
Additional context
The correct way with dealing with internal openshift registry being present or not present is not setting referencePolicy: Source. The one exception being tag-notation references in tag.from.name in DEV environments. There, referencePolicy: Source can make sense, possibly also in combination with regularly-updates digests instead of tags in from.name, as you currently do in master branch.
It is to use image change trigger and letting Openshift resolve the correct location of images by means of that mechanism.
That requires PR-133 kubeflow notebook controller and PR-800 odh dashboard to be merged in.
Behavior of that mechanism:
The text was updated successfully, but these errors were encountered: