CFE-1162: Updates enhancement to reflect hcp usecase and with latest information on aws tags support #1700
Conversation
Signed-off-by: Trilok Geer <[email protected]>
Signed-off-by: Trilok Geer <[email protected]>
|
[APPROVALNOTIFIER] This PR is NOT APPROVED This pull-request has been approved by: The full list of commands accepted by this bot can be found here.
Needs approval from an approver in each of these files:
Approvers can indicate their approval by writing |
TrilokGeer
changed the title
TrilokGeer
changed the title
openshift-ci-robot
added
the
jira/valid-reference
|
@TrilokGeer: This pull request references CFE-1162 which is a valid jira issue. Warning: The referenced jira issue has an invalid target version for the target branch this PR targets: expected the story to target the "4.18.0" version, but no target version was set. In response to this:
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 openshift-eng/jira-lifecycle-plugin repository. |
|
|
||
| A new field `propagateUserTags` is added in GA release version. The `experimentalPropagateUserTags` field will be deprecated. In future release versions, `experimentalPropagateUserTags` will be removed. | ||
| When both fields are set, `experimentalPropagateUserTags` takes precedence. | ||
| If the userTags field is changed post-install, there is no guarantee about how an in-cluster operator will respond to the change. Some operators may reconcile the change and change tags on the AWS resource. Some operators may ignore the change. However, if tags are removed from userTags, the tag will not be removed from the AWS resource. |
There was a problem hiding this comment.
You moved this paragraph, but did you mean to remove it entirely (or perhaps keep it for historical reference, but mark it as outdated)? We've run into an obstacle with cluster-ingress-operator and cloud-provider-aws (see openshift/cluster-ingress-operator#1148 (comment)), so I want to understand whether (1) handling updates and (2) handling removals in particular are hard requirements or not.
There was a problem hiding this comment.
In contrast to cluster-ingress-operator and cloud-provider-aws, I checked the behaviour with aws-load-balancer-operator#137, and it can update and remove the tags.
oc get infrastructures.config.openshift.io cluster -oyaml | yq .status.platformStatus
aws:
region: us-east-2
resourceTags:
- key: conflict-key1
value: plat-value2
- key: conflict-key2
value: plat-value3
type: AWS
aws elbv2 describe-tags --resource-arns arn:aws:elasticloadbalancing:us-east-*******
{
"TagDescriptions": [
{
"ResourceArn": "arn:aws:elasticloadbalancing:us-east-*******",
"Tags": [
{
"Key": "service.k8s.aws/resource",
"Value": "LoadBalancer"
},
{
"Key": "service.k8s.aws/stack",
"Value": "aws-load-balancer-test-default-lb-class/echoserver"
},
{
"Key": "conflict-key2",
"Value": "plat-value3"
},
{
"Key": "conflict-key1",
"Value": "plat-value2"
},
{
"Key": "elbv2.k8s.aws/cluster",
"Value": "ckyal-20241021-2e71f8-pmfc8"
}
]
}
]
}
Updated infrastructure status: removed conflict-key1 and updated conflict-key2
oc get infrastructures.config.openshift.io cluster -oyaml | yq .status.platformStatus
aws:
region: us-east-2
resourceTags:
- key: conflict-key2
value: new-plat-value3
type: AWS
aws elbv2 describe-tags --resource-arns arn:aws:elasticloadbalancing:us-east-*******
{
"TagDescriptions": [
{
"ResourceArn": "arn:aws:elasticloadbalancing:us-east-*******",
"Tags": [
{
"Key": "service.k8s.aws/resource",
"Value": "LoadBalancer"
},
{
"Key": "service.k8s.aws/stack",
"Value": "aws-load-balancer-test-default-lb-class/echoserver"
},
{
"Key": "conflict-key2",
"Value": "new-plat-value3"
},
{
"Key": "elbv2.k8s.aws/cluster",
"Value": "ckyal-20241021-2e71f8-pmfc8"
}
]
}
]
}
So, there is a bit of inconsistency between upstream cloud-provider-aws and aws-load-balancer-controller on their way of handling tags.
From OpenShift (cluster-ingress-operator and aws-load-balancer-operator) side, I think we can only control
aws-load-balancer-additional-resource-tagsannotation forcluster-ingress-operator, and--default-tagscontainer arg foraws-load-balancer-operator
to propagate relevant tags to its consumers, which in this case are cloud-provider-aws and aws-load-balancer-controller, respectively. If we need to standardise the behaviour, then we might need to change the upstream logic.
There was a problem hiding this comment.
You moved this paragraph, but did you mean to remove it entirely (or perhaps keep it for historical reference, but mark it as outdated)? We've run into an obstacle with cluster-ingress-operator and cloud-provider-aws (see openshift/cluster-ingress-operator#1148 (comment)), so I want to understand whether (1) handling updates and (2) handling removals in particular are hard requirements or not.
Ah!, I think it is a residue of the initial version, thanks @Miciah. In summary, all redhat supported operators (in-cluster or day2) will reconcile the tags on resources created and managed by them. Removal of tag is not supported, as it'd intervene and remove the tags un-intentionally for tags added externally by the user, if any.
There was a problem hiding this comment.
Removal of tag is not supported, as it'd intervene and remove the tags un-intentionally for tags added externally by the user, if any.
aws-load-balancer-controller provides a command line flag called --external-managed-tags through which a list of tag keys can be ignored to be reconciled by the controller.
xRef :
- https://kubernetes-sigs.github.io/aws-load-balancer-controller/latest/deploy/configurations/
- add support for external-managed-tags & prefer defaultTags kubernetes-sigs/aws-load-balancer-controller#1970
external-managed-tags stringList AWS Tag keys that will be managed externally. Specified Tags are ignored during reconciliation
This means if a user wants to manage a tag externally, it has to be added in external-managed-tags list. If not, aws-load-balancer-controller will remove unsolicited tags, and apply tags defined in --default-tags.
@alebedev87 As I can see, aws-load-balancer-operator doesn't provide a field to specify ExternalManagedTags. Was there any discussion around this, or it's an existing gap?
Signed-off-by: Trilok Geer <[email protected]>
| For the resources created and managed by hosted control plane, cluster api provider for aws reconciles the user tags on AWS resources. The hosted control plane updates the `infrastructure.config.openshift.io` resource to reflect new tags in `resourceTags`. The OpenShift operators, both core and non-core (managed by RedHat), reconcile the respective AWS resources created and managed by them. | ||
| Given that, there is no universal controller to update all resources created by OpenShift, the day2 updates of tags is not supported for standalone OpenShift deployments. |
There was a problem hiding this comment.
What does it mean "not supported for standalone OpenShift deployments"? What will happen if a cluster admin changes resourceTags in infrastructure? Line 95 ("If the userTags field is changed post-install,") says it's going to be applied to the cloud objects.
There was a problem hiding this comment.
At present, in standalone Openshift deployments, the resourceTags in infrastructure is handled as immutable and not a supported operation. The userTags update is allowed for hcp usecases, where, hcp control plane updates the infrastructure status based on edits in hostedCluster and nodepool objects.
There was a problem hiding this comment.
resourceTags in infrastructure is handled as immutable and not a supported operation
This is not true, status.platformStatus.aws.resourceTags is editable on a standalone cluster.
Something will happen on the standalone cluster. I think with the current PR, the volumes just get re-tagged, just as in HCP.
|
/cc @alebedev87 |
Signed-off-by: Trilok Geer <[email protected]>
|
@TrilokGeer: all tests passed! 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-sigs/prow repository. I understand the commands that are listed here. |
The PR updates enhancement to reflect hcp usecase and with latest information on aws tags support.