-
Feature Request
-
Resolution: Done
-
Minor
-
None
-
None
-
False
-
False
-
-
-
-
- Proposed title of this feature request
- Llimiting network bandwidth with namespace level.
- What is the nature and description of the request?
- Current OpenShift4 can limits network bandwidth with pod level.https://docs.openshift.com/container-platform/4.9/nodes/pods/nodes-pods-configuring.html#nodes-pods-configuring-bandwidth_nodes-pods-configuring
However, each ordinary user can modify their pod's yaml freely.
It means that it's impossible to limit network bandwidth from admin side.
So customer wants to limit network bandwidth with namespace level.
They think it's good if network bandwidth can be limited with ResourceQuota as well as CPU or Memory.
- Why does the customer need this? (List the business requirements here)
- Let's say one cluster with multiple users by creating multiple projects on the cluster.
For example, UserA,B,C have ProjectA,B,C respectively.
Now if UserA deployed a Pod that consumes all of the network bandwidth, UserB and C are no longer able to run their Pods.
It's a big problem if it happened in production.
All the admin of their cluster can do is just to say "Hey! Don't deploy such a Pod!" to users.
It's just a caution so any user can break the rule if they really want to.
It's bothering the admin.
The customer has no idea to manage their cluster properly without the limiting network bandwidth feature.
It makes them hesitate to use OpenShift in production.
- There is no way to prevent the situation that network bandwidth is occupied by just one user, To implement the feature which can limit network bandwidth with namespace level is the only way to prevent such a situation.
- List any affected packages or components.
- Node
- Namespace/Project
Note:
CU expectation is as follows.
~~~
apiVersion: v1
kind: ResourceQuota
metadata:
name: quota
spec:
hard:
cpu: "1000"
memory: 200Gi
pods: "10"
networkBandwidth: 100M <=== Add this parameter
After admin configured ResourceQuota as the above, user can split it to their pods as follows:
- PodA: 50M
- PodB: 30M
- PodC: 20M
~~~
Thanks.
- duplicates
-
RFE-1833 [CNI] Support Bandwidth plugin in OCP 4.x
- Rejected