Closed diverdane closed 3 years ago
@diverdane why do we need a whole separate field for the openshift image? to me, this looks weird:
image:
repository: cyberark/conjur # https://hub.docker.com/r/cyberark/conjur/
tag: '1.11.1'
pullPolicy: Always
. . .
openshift:
image:
repository: registry.connect.redhat.com/cyberark/conjur
tag: latest
pullPolicy: Always
in practice, I'm only going to deploy this to one platform, right?
what is the purpose of a separate field for openshift
images if it uses the same sub-fields? is this a flag to trigger the OCP deployment instead of the vanilla k8s?
would it make more sense to have an optional platform
sub-field of image
, that can default to kubernetes
and run the standard flow, but can be set to openshift
if deploying to openshift? eg
image:
repository: registry.connect.redhat.com/cyberark/conjur
tag: latest
pullPolicy: Always
platform: openshift
@izgeri: The intent is to incorporate two sets of default repository/tag values, one for Kubernetes and one for OpenShift. Current users of the chart will be depending on the default value of the Conjur repository to be cyberark/conjur
, whereas any OpenShift users will want a default value of registry.connect.redhat.com/cyberark/conjur
.
What we could potentially do is just introduce new OpenShift values within the image:
section, e.g.:
image:
repository: cyberark/conjur
tag: '1.11.1'
openshiftRepository: registry.connect.redhat.com/cyberark/conjur
openshiftTag: latest
pullPolicy: Always
platform: openshift
or, probably cleaner:
image:
repository: cyberark/conjur
tag: '1.11.1'
openshift:
repository: registry.connect.redhat.com/cyberark/conjur
tag: latest
pullPolicy: Always
platform: openshift
When we're ready to make a major version bump, we can probably change it back to the way the code is now (i.e. separate kubernetes:
and openshift:
sections).
Can I challenge the idea that we need a default OpenShift image? Can't we just provide clear instructions that when you want to deploy to OCP, you need to override the repository
field to point to registry.connect.redhat.com/cyberark/conjur
(and also for the Nginx image)?
Ah, yes, that makes sense, and would keep things very simple in the chart! This can be documented both in the README.md and in the values.yaml
file itself (with commented values) to make it very clear.
Of course, the platform:
field wouldn't offer much here. The user would still need to explicitly supply the image repo:tag values for OpenShift, despite the existence of a platform: openshift
field.
Is your feature request related to a problem? Please describe.
The changes that were made to the
values.yaml
file in order to support OpenShift deployments (PR #71) represent breaking changes for anyone using the Conjur OSS Helm chart. For example, the fields for the Conjur image repository/tag now use akubernetes
andopenshift
in the values path:We would like to avoid breaking changes, and thereby limit the cadence of introducing major versions. In this case, we can avoid breaking changes by reverting the
repository
andtag
fields to their original form, with default values ofcyberark/conjur
and1.11.1
respectively. The default values to use for OpenShift would then need to be clearly documented both in the chart's README.md, and ideally in the values.yaml file itself.Describe the solution you would like
image.repository
andimage.tag
value schema is restored.README.md
and in thevalues.yaml
file.Describe alternatives you have considered
Separate default values fields for OpenShift, but this is complex and awkward looking.
Additional context