ns describe quota used not update

[j0rGeT]

I found that when the hrq quota is used, the quota information used in describe ns is not updated. After the quota used in the child ns is updated, the value of used in the parent ns is still 0。

parent-ns:

root@saar-virtual-machine:~# kubectl describe ns kubecube-workspace-workspacez5
Name:         kubecube-workspace-workspacez5

Resource Quotas
 Name:                    hrq.hnc.x-k8s.io
 Resource                 Used  Hard
 --------                 ---   ---
 limits.cpu               0     2
 limits.memory            0     500Mi
 requests.cpu             0     2
 requests.memory          0     500Mi
 requests.nvidia.com/gpu  0     0`

sub-ns:

root@saar-virtual-machine:~# kubectl describe ns kubecube-project-projectz1
Name:         kubecube-project-projectz1
Status:       Active

Resource Quotas
 Name:                    hrq.hnc.x-k8s.io
 Resource                 Used  Hard
 --------                 ---   ---
 limits.cpu               1     2
 limits.memory            100M  500Mi
 requests.cpu             1     2
 requests.memory          100M  500Mi
 requests.nvidia.com/gpu  0     0

[zfrhv]

true, also mentioned in #274
when describing the ns it displays the ResourceQuota and not the HierarchicalResourceQuota
so the values in the ResourceQuota are not updated

[mochizuki875]

HierarchicalResourceQuota seems to be implemented by the combination of ResourceQuota and Admission Webhook.
ResourceQuota is namespaced resource and just control the namespace's resource consumption.
So if child ns is updated, it is not reflected in parent ns.
It is reflected to HierarchicalResourceQuota's status field.

[adrianludwin]

Yeah, hrq.hnc.x-k8s.io is just the per-namespace implementation of the HRQ - it has no visibility into what's happening in the rest of the hierarchy. But @mochizuki875 's amazing PR (#283) will at least make it easy to understand what's there.

Perhaps we should give that RQ a more descriptive name - e.g. local-impl.hnc.x-k8s.io would more clearly emphasize that you shouldn't trust the data in here. wdyt?

(Sorry for the delays, just some maintainer burnout)

[zfrhv]

@adrianludwin
thats a very good idea

[keisukesakasai]

@adrianludwin
Hi, I will try to fix this. Is "local-impl." OK?

/assign @keisukesakasai

[adrianludwin]

Maybe it's nice to get the word "hrq" in there as well - e.g. if local-impl-hrq lgty, then I'm good with it too. Thanks!

[zfrhv]

@adrianludwin
@keisukesakasai
maybe its better to include the original namespace name in which hrq is created, or the hrq resource name?
for example: <namespace name>-hrq-local-imp
just in case there are multiple hrqs in the namespace tree.

for example in this case:

grandpa (has hrq)
├─ father1 (has hrq)
│  ├─ son1
│  ├─ son2
├─ father2

there should be 2 ResourceQuotas in namespace son1, and they cant have the same name
(if i correctly understand how it works)

[adrianludwin]

HNC combines the limits from grandpa and father1 already, so we can't name it after the "real" object's namespace because there could be multiple of them.

[mochizuki875]

For this issue, we should add some kind annotation to document as discussed #301 (comment).

[mochizuki875]

Close this issue as #340 has been merged.
/close

[k8s-ci-robot]

@mochizuki875: Closing this issue.

In response to this:

Close this issue as #340 has been merged.
/close

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.