keep collecting status when apply failed

[Garrybest]

Signed-off-by: Garrybest [email protected]

What type of PR is this?
/kind bug

What this PR does / why we need it:
If there is something wrong with applied condition, we should not erase the status in binding status. Applied information is from execution controller whereas status is from work status controller.

Which issue(s) this PR fixes:
Fixes #

Special notes for your reviewer:

Does this PR introduce a user-facing change?:

Keep aggregating status from work to binding even if application from binding to work fails.

[karmada-bot]

[APPROVALNOTIFIER] This PR is NOT APPROVED

This pull-request has been approved by:
To complete the pull request process, please assign kevin-wangzefeng after the PR has been reviewed.
You can assign the PR to them by writing /assign @kevin-wangzefeng in a comment when ready.

The full list of commands accepted by this bot can be found here.

Needs approval from an approver in each of these files:

• OWNERS

Approvers can indicate their approval by writing /approve in a comment
Approvers can cancel approval by writing /approve cancel in a comment

[codecov-commenter]

Codecov Report

Merging #2991 (bcae500) into master (c8954de) will increase coverage by 0.44%.
The diff coverage is 72.00%.

@@            Coverage Diff             @@
##           master    #2991      +/-   ##
==========================================
+ Coverage   38.58%   39.02%   +0.44%     
==========================================
  Files         206      206              
  Lines       18798    18807       +9     
==========================================
+ Hits         7253     7340      +87     
+ Misses      11117    11032      -85     
- Partials      428      435       +7     

Flag	Coverage Δ
unittests	39.02% <72.00%> (+0.44%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

Impacted Files	Coverage Δ
pkg/controllers/status/workstatus_controller.go	0.00% <0.00%> (ø)
pkg/util/helper/workstatus.go	41.90% <75.00%> (+31.29%)	⬆️
pkg/search/proxy/store/util.go	92.89% <0.00%> (-1.43%)	⬇️
pkg/util/names/names.go	96.38% <0.00%> (-0.17%)	⬇️
pkg/util/namespace.go	100.00% <0.00%> (ø)
...scheduler/core/spreadconstraint/select_clusters.go	86.04% <0.00%> (+0.33%)	⬆️
pkg/controllers/binding/common.go	26.47% <0.00%> (+0.45%)	⬆️
.../scheduler/core/spreadconstraint/group_clusters.go	95.20% <0.00%> (+2.01%)	⬆️
pkg/util/serviceaccount.go	90.27% <0.00%> (+2.55%)	⬆️
... and 1 more

Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.

[jwcesign]

Can we add ut for this? Also, we need release notes.

[jwcesign]

/lgtm cc @XiShanYongYe-Chang for checking

[XiShanYongYe-Chang]

/assign

[RainbowMango]

If there is something wrong with applied condition, we should not erase the status in binding status.

Can you explain it in more detail? Where do we erase the status? and why we shouldn't do that?

[Garrybest]

Where do we erase the status?

If failed to apply, aggregated status will not be set even though status exists in work status and this function just returns.

karmada/pkg/util/helper/workstatus.go

Lines 186 to 196 in 3d06bb8

	if !applied {
	aggregatedStatus := workv1alpha2.AggregatedStatusItem{
	ClusterName: clusterName,
	Applied: applied,
	AppliedMessage: appliedMsg,
	Health: workv1alpha2.ResourceUnknown,
	}
	statuses = append(statuses, aggregatedStatus)
	continue
	}

Why we shouldn't do that?

Actually, apply is sort of synchronization from 'karmada' to 'member clusters', it only take effects on resource metadata and spec. However, status collection is a synchronization from 'member clusters' to 'karmada'. So it is unresonable to skip status aggregation when failed to apply because these two actions have no identical relationship.

[XiShanYongYe-Chang]

Hi @Garrybest, I think the modification is reasonable.

Do you encounter this problem? After a resource fails to be applied, the status of the work exists.

In addition, we have a similar logic. Please confirm whether to modify it:

karmada/pkg/controllers/federatedresourcequota/federated_resource_quota_status_controller.go

Line 183 in a20dd6e

if !applied {

[XiShanYongYe-Chang]

This one looks more like a cleanup, what do you think @Garrybest?

[Garrybest]

Do you encounter this problem? After a resource fails to be applied, the status of the work exists.

Yes, I got a validation webhook in member cluster which rejects the modification whereas it had no effects on status collection.

In addition, we have a similar logic.

Cool.

This one looks more like a cleanup, what do you think @Garrybest?

I'm not sure about the category, cc @RainbowMango for help😄.

[RainbowMango]

So it is unresonable to skip status aggregation when failed to apply because these two actions have no identical relationship.

The reason why skip status aggregation is the status is probably an ambiguous one. It's not the right status of current resource template.
For instance:

1. We deployed a Deployment(v1), and get the status back, it works as expected.
2. Then, we update the Deployment to V2, if we can't apply it to member clusters, then we still get the V1 status. It will bring ambiguity.

[Garrybest]

For instance:

1. We deployed a Deployment(v1), and get the status back, it works as expected.
2. Then, we update the Deployment to V2, if we can't apply it to member clusters, then we still get the V1 status. It will bring ambiguity.

Got it, it makes sense. Well, in this case, deployment of status will be erased here. So users will see a deployment with 0 ready replicas and 0 available replicas from status, maybe this situation brings a panic for users.

These two different logic represents different consideration and scenarios, which one is more sensible? Glad to here more opinions😊.

[RainbowMango]

So users will see a deployment with 0 ready replicas and 0 available replicas from status, maybe this situation brings a panic for users.

Yes, I agree that just erasing the previous status is not ideal behavior, but I think we have sent the event to the resource template, isn't it? it should be a relief.

[Garrybest]

Actually I'm just concerned that users are likely to have their own controllers in karmada control plane which depend on object status to reconcile and the handling process could be disrupted since status does not collect correctly under this circumstance.

Well, it's just some ideas, sort of design tradeoff😄. We may need more evidences and scenario about this patch.

[RainbowMango]

Well, it's just some ideas, sort of design tradeoff😄. We may need more evidences and scenario about this patch.

Yes. kind of tradeoff here.

[RainbowMango]

I'm trying to figure out which PR/Issue should be included in the coming v1.7 release which is planned at the end of this month.
I guess we don't have enough time for this feature, so I'm moving this to v1.8.

@XiShanYongYe-Chang Could you please confirm if we still facing this issue? I feel we solved a similar issue somewhere. Not sure about that.