console
console copied to clipboard
CONSOLE-3243: Rename "master" to "control plane node" in node pages
Addresses https://issues.redhat.com/browse/CONSOLE-3243
Changes all string occurences of "master" in the console to "control plane node".
/retest
QE Approver: /assign @yapei
Docs Approver: /assign @opayne1
PX Approver: /assign @RickJWagner
Console Approver: /assign @rhamilto
/label px-approved
[APPROVALNOTIFIER] This PR is NOT APPROVED
This pull-request has been approved by: RickJWagner, tvu20
Once this PR has been reviewed and has the lgtm label, please ask for approval from rhamilto by writing /assign @rhamilto
in a comment. For more information see:The Kubernetes Code Review Process.
The full list of commands accepted by this bot can be found here.
Approvers can indicate their approval by writing /approve
in a comment
Approvers can cancel approval by writing /approve cancel
in a comment
/label docs-approved
Nice work, @tvu20! Also need to include updates to Cluster Settings:
- https://github.com/openshift/console/blob/48991b57ac9e73f352fefd3f6ce7402e7da74403/frontend/public/module/k8s/cluster-settings.ts#L309 becomes
Master = 'Control plane',
- https://github.com/openshift/console/blob/48991b57ac9e73f352fefd3f6ce7402e7da74403/frontend/public/components/cluster-settings/cluster-settings.tsx#L619-L624 becomes
const mcpName = machineConfigPool?.metadata?.name;
const machineConfigPoolIsEditable = useAccessReview({
group: MachineConfigPoolModel.apiGroup,
resource: MachineConfigPoolModel.plural,
verb: 'patch',
name: mcpName,
});
- https://github.com/openshift/console/blob/48991b57ac9e73f352fefd3f6ce7402e7da74403/frontend/public/components/cluster-settings/cluster-settings.tsx#L648 becomes
to={
/k8s/cluster/nodes?rowFilter-node-role=${mcpName}}
It strikes me as a little odd that we're changing "master" to "control plane", but the underlying resource labels aren't changing. So we still have "master" in the UI. For example,
https://user-images.githubusercontent.com/895728/183096759-0aeb6d02-9b32-4ae5-a43b-2aae66246583.mov



Do we need to engage UXD here so as to avoid creating confusion for users? cc: @megan-hall
@tvu20, I think this probably merits a wider discussion. Unless the resource names (e.g., /k8s/cluster/machineconfiguration.openshift.io~v1~MachineConfigPool/master) and associated labels are changing, there is a limit to how far we can take these changes, and I fear it may create confusion on the part of end users. We probably need to revisit the intent of the story with @jhadvig and the stakeholder.
@XiyunZhao is testing on this PR
/retest
/retest
Hi @tvu20, based on current changes, I opened two bugs for this new feature, could you help to identify whether they need to update or not?
- The name of "master" on "Filter by node type" dropdown list on cluster utilization on the overview page is not changed
https://bugzilla.redhat.com/show_bug.cgi?id=2116608
- After the changes, duplicate role name will be listed on "Nodes" page which is incorrect
https://bugzilla.redhat.com/show_bug.cgi?id=2116625
Hi @XiyunZhao, I have just pushed a new commit that addresses bug 1. I am not able to replicate the second bug on my end.
/retest
/retest
/retest
/retest
@tvu20: The following test failed, say /retest
to rerun all failed tests or /retest-required
to rerun all mandatory failed tests:
Test name | Commit | Details | Required | Rerun command |
---|---|---|---|---|
ci/prow/frontend | bf85b78ce3cfdd35a76c6767f63e1571310807a7 | link | true | /test frontend |
Full PR test history. Your PR dashboard.
Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes/test-infra repository. I understand the commands that are listed here.
@tvu20 sorry for bug 2, after checking, it is not a duplicate value. I would like to ask what's the difference between "control plane" with "control-plane"? Will it mislead customers?
@opayne1 could you please elaborate on the previous question? Thanks !
@jhadvig @XiyunZhao In our documentation, we use "control plane" with no hyphen. See the Control plane architecture docs for an example. We also have it as "control planes" no hyphen in our documentation contributor guidelines. Adding the docs-approved label as this PR looks good to me.
/label docs-approved
@tvu20, I think this probably merits a wider discussion. Unless the resource names (e.g., /k8s/cluster/machineconfiguration.openshift.io~v1~MachineConfigPool/master) and associated labels are changing, there is a limit to how far we can take these changes, and I fear it may create confusion on the part of end users. We probably need to revisit the intent of the story with @jhadvig and the stakeholder.
@tvu20 Is there any conclusion for this comment? Could you help to add some details to features description?
@tvu20, I think this probably merits a wider discussion. Unless the resource names (e.g., /k8s/cluster/machineconfiguration.openshift.iov1MachineConfigPool/master) and associated labels are changing, there is a limit to how far we can take these changes, and I fear it may create confusion on the part of end users. We probably need to revisit the intent of the story with @jhadvig and the stakeholder.
I agree with @rhamilto. We should revisit the story since it can lead to confusion between Nodes, MachineConfig and MachineConfigPool resources, which are still using the legacy master
naming. Until that is solved, this change would only lead to user confusion. We should be consistent with the naming.
/hold
I'm in favor of doing what we can to get away from this terminology, but I agree this will create confusion since the actual names and labels of the resources are still master
. You'll need to know the real value when using the CLI, writing label selectors, or editing the resource YAML. Even in the UI, we would show "control plane" in some places and "master" in others as @rhamilto points out.
Have we considered fixing this at the platform level and changing the actual resource names? (I'm sure this is easier said than done since names are immutable, this is arguably API breaking, and we probably don't want to drift from upstream...)
@deads2k @sdodson Has this come up before?
@deads2k @sdodson Has this come up before?
Poking around to see what I can find. This seems to have happened upstream via kubeadm(https://github.com/kubernetes/kubeadm/issues/2200) but we won't pick those changes up in OCP as we don't leverage kubeadm.
@sdodson has pointed out that we should at least be able to use a new node role in 4.12: https://bugzilla.redhat.com/show_bug.cgi?id=1995858 (thanks!)
OK, here is the upstream guidance:
https://github.com/kubernetes/community/blob/master/sig-architecture/naming/recommendations/001-master-control-plane.md
Based on this, I'd suggest
- We use "control plane" or "control plane nodes" in text descriptions, help, etc. rather than "master"
- We use the
node-role.kubernetes.io/control-plane
role instead ofnode-role.kubernetes.io/master
- We only use
master
when displaying a name of a resource
@spadgett @rhamilto @jhadvig So based on Sams comment above, is this PR / story still needed or is this change a docs only or help only change? If this is a code change, can we push this to 4.13?
https://github.com/kubernetes/community/blob/master/sig-architecture/naming/recommendations/001-master-control-plane.md
Yes, we'll still need console code changes, most of which are in this PR, but we need a few tweaks based on the guidance.
Replaced by https://github.com/openshift/console/pull/12209