Expose generic OpenShift routes for KServe InferenceServices #84

vaibhavjainwiz · 2023-09-18T19:09:22Z

Description

With #59 we had implemented a hot fix by exposing predictor URL directly to user but to provide proper fix we need to implement a generic openshift route for kserve inference service.

How Has This Been Tested?

Create a new cluster
Setup the cluster with below Installation step
https://github.com/opendatahub-io/caikit-tgis-serving/blob/main/demo/kserve/install-manual.md
Before executing Step 11, specify custom manifests

apiVersion: dscinitialization.opendatahub.io/v1
kind: DSCInitialization
metadata:
  name: example
spec:
  devFlags:
    manifestsUri: https://github.com/vaibhavjainwiz/odh-manifests/tarball/generic_route

Apply Step 11.
Deploy/validate a Kserve model with below Step.
https://github.com/opendatahub-io/caikit-tgis-serving/blob/main/demo/kserve/deploy-remove.md

If everything work as mentioned in above step then we are good to go :)

Expected behavior

A new route should be created in istio-system namespace with name same as InferenceService. If new route is created then we are good to go :)

Merge criteria:

The commits are squashed in a cohesive manner and have meaningful messages.
Testing instructions have been added in the PR body (for PRs involving changes that are not immediately obvious).
The developer has manually tested the changes and verified that the changes work

Signed-off-by: Vaibhav Jain <vajain@redhat.com>

Jooho · 2023-09-18T20:31:33Z

I am testing this PR but I have a couple of questions.

what images should be used for this test?
when I followed your test steps, it deployed quay.io/opendatahub/odh-model-controller:stable
I don't think it is your test image for this PR.
What is the expected result from these test steps?
Where should I check? Route?

controllers/processors/deltaProcessor.go

israel-hdez

Please, add unit tests.

controllers/kserve_inferenceservice_controller.go

controllers/reconcilers/kisvc_route_reconciler.go

controllers/kserve_inferenceservice_controller.go

vaibhavjainwiz · 2023-09-19T12:06:56Z

what images should be used for this test?
when I followed your test steps, it deployed quay.io/opendatahub/odh-model-controller:stable

To test this PR you need to use odh-model-controller image created by openshift-ci for this corresponding PR.
quay.io/opendatahub/odh-model-controller:pr-84
I am not sure how we could provide overriden value of odh-model-controller image through v2 operator. @Jooho Could you please with this configuration.

What is the expected result from these test steps?
Where should I check? Route?

With this PR, new more generic route would be created in istio-system namespace which should be use to query model.

Signed-off-by: Vaibhav Jain <vajain@redhat.com>

Jooho · 2023-09-20T12:54:39Z

To test this PR you need to use odh-model-controller image created by openshift-ci for this corresponding PR.
quay.io/opendatahub/odh-model-controller:pr-84
I am not sure how we could provide overriden value of odh-model-controller image through v2 operator. @Jooho Could you please with this configuration.

I'm not entirely certain about the testing approach you ultimately used, but if testers or reviewers are expected to rely on the PR image, it would greatly help if you could provide clear and accurate test steps in the description. This would make it much more straightforward to conduct testing using the provided information.

With this PR, new more generic route would be created in istio-system namespace which should be use to query model.

As testers, it's essential that the PR description provides clear guidance on what needs to be verified using the 'Expected result' or 'Expected behavior' sections. Testing can be a time-consuming process for reviewers and testers, so it's crucial for PR senders to provide detailed clarification in this regard.

vaibhavjainwiz · 2023-09-20T14:43:48Z

On my local, I follow following step for testing :

Install the Kserve controller using odh operator
Remove ODH operator
Set deployment replicas of odh-model-controller to 0
Run odh-model-controller from local in debug mode.

To support testing for reviewer, I had created a image of odh-model-controller from local and push it in quay.io.
I had also updated the image reference in odh-manifest repo so if deploy their cluster with https://github.com/vaibhavjainwiz/odh-manifests/tarball/generic_route in DSCInitialization then everything should work.

vaibhavjainwiz · 2023-09-20T14:53:25Z

@Jooho
I had added Expected behavior section in Issue Description.

vaibhavjainwiz · 2023-09-20T14:59:54Z

/test

openshift-ci · 2023-09-20T14:59:57Z

@vaibhavjainwiz: The /test command needs one or more targets.
The following commands are available to trigger required jobs:

/test fvt
/test images
/test pr-image-mirror
/test unit

Use /test all to run all jobs.

In response to this:

/test

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.

vaibhavjainwiz · 2023-09-20T15:00:10Z

/test all

Signed-off-by: Vaibhav Jain <vajain@redhat.com>

vaibhavjainwiz · 2023-09-24T17:34:53Z

@israel-hdez I had the unit testcases. Could you please review.

israel-hdez

@vaibhavjainwiz I'm happy with the code.

I still want to test before approving. I didn't have time today.

If you want an approval sooner, maybe work with @rpancham and test it together?
Otherwise, I can try on my Tuesday.

controllers/inferenceservice_controller.go

controllers/kserve_inferenceservice_controller.go

Code looks OK.

Signed-off-by: Vaibhav Jain <vajain@redhat.com>

israel-hdez · 2023-09-26T19:26:00Z

/retest

israel-hdez

I tested. So far, looks good, with some issues:

It seems RBAC is missing in this PR. After installing and creating an ISVC I got spec.host: Forbidden: you do not have permission to set the host field of the route, which I fixed by modifying the Role to add something like this: Route's new admission routes/custom-host doesn't work for legacy API openshift/origin#14904 (comment).
- I remember this RBAC would need to go in some kubebuilder comments, to let it generate them for you. I don't know how to do this.
Similar permission denied error when trying to delete the ISVC and, thus, trying to delete the route.
Name conflict. Having two ISVCs with same name on different namespaces, will lead to a conflict in the route.
Schema mismatch: The ISVC can state that protocol is plain HTTP, while the route will be created for HTTPS.
- Perhaps for our demo/sample setup, this is error config on KServe side (i.e. is config error on my testing). Yet, I think we want to honor the schema in the ISVC, just in case some user wants to use plain HTTP.

vaibhavjainwiz · 2023-09-28T11:23:11Z

I tested. So far, looks good, with some issues:

It seems RBAC is missing in this PR. After installing and creating an ISVC I got spec.host: Forbidden: you do not have permission to set the host field of the route, which I fixed by modifying the Role to add something like this: Route's new admission routes/custom-host doesn't work for legacy API openshift/origin#14904 (comment).

I remember this RBAC would need to go in some kubebuilder comments, to let it generate them for you. I don't know how to do this.

Similar permission denied error when trying to delete the ISVC and, thus, trying to delete the route.

Strange these RBAC are already present in code. You should not see this error. I will debug it more.

odh-model-controller/main.go

Line 68 in bf310d9

    
           // +kubebuilder:rbac:groups=route.openshift.io,resources=routes,verbs=get;list;watch;create;update;patch

odh-model-controller/config/rbac/role.yaml

Line 135 in 8ccb4cc

- apiGroups:

Name conflict. Having two ISVCs with same name on different namespaces, will lead to a conflict in the route.

I updated route name to include isvc.Name + isvc.Namespace so that It would not cause.

Schema mismatch: The ISVC can state that protocol is plain HTTP, while the route will be created for HTTPS.

Perhaps for our demo/sample setup, this is error config on KServe side (i.e. is config error on my testing). Yet, I think we want to honor the schema in the ISVC, just in case some user wants to use plain HTTP.

Are you talking about below config?
https://github.com/opendatahub-io/kserve/blob/1529b712ae536286fe939e7b6ad426305bf168f1/config/configmap/inferenceservice.yaml#L452

Signed-off-by: Vaibhav Jain <vajain@redhat.com>

israel-hdez · 2023-09-28T16:27:46Z

Strange these RBAC are already present in code. You should not see this error. I will debug it more.

odh-model-controller/main.go

Line 68 in bf310d9

// +kubebuilder:rbac:groups=route.openshift.io,resources=routes,verbs=get;list;watch;create;update;patch

odh-model-controller/config/rbac/role.yaml

Line 135 in 8ccb4cc

- apiGroups:

Yes, some RBAC is there. Although, it is missing delete privileges.
It is also missing the privilege to set custom hosts, which is a different resource.

Schema mismatch: The ISVC can state that protocol is plain HTTP, while the route will be created for HTTPS.

Perhaps for our demo/sample setup, this is error config on KServe side (i.e. is config error on my testing). Yet, I think we want to honor the schema in the ISVC, just in case some user wants to use plain HTTP.

Are you talking about below config? https://github.com/opendatahub-io/kserve/blob/1529b712ae536286fe939e7b6ad426305bf168f1/config/configmap/inferenceservice.yaml#L452

Yes. That configuration is reflected in the ISVC. So, you don't need to look at the ConfigMap. Just check the schema set in the ISVC.

Signed-off-by: Vaibhav Jain <vajain@redhat.com>

vaibhavjainwiz · 2023-10-05T19:42:09Z

@israel-hdez I had incoporated all review comments. Could you please review again.
I had made the change in odh-model-controller to honor user specified url scheme but some change are also required in kserve as well because url in InferenceService status is set by KServe controller itself. I will create a seperate ticket in KServe to fix it.

israel-hdez · 2023-10-06T05:16:57Z

/lgtm

openshift-ci · 2023-10-06T05:18:31Z

[APPROVALNOTIFIER] This PR is APPROVED

This pull-request has been approved by: israel-hdez, vaibhavjainwiz

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

The pull request process is described here

Needs approval from an approver in each of these files:

~~OWNERS~~ [israel-hdez]

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

Partial revert of opendatahub-io/odh-manifests#916, because opendatahub-io/odh-model-controller#84 has been completed. Signed-off-by: Edgar Hernández <23639005+israel-hdez@users.noreply.github.com>

Expose generic OpenShift routes for KServe InferenceServices

c89daf8

Signed-off-by: Vaibhav Jain <vajain@redhat.com>

vaibhavjainwiz requested review from Jooho, VedantMahabaleshwarkar, israel-hdez and heyselbi September 18, 2023 19:09

openshift-ci bot requested a review from Xaenalt September 18, 2023 19:09

Jooho reviewed Sep 18, 2023

View reviewed changes

controllers/processors/deltaProcessor.go Show resolved Hide resolved

israel-hdez previously requested changes Sep 19, 2023

View reviewed changes

openshift-ci bot assigned israel-hdez Sep 19, 2023

israel-hdez reviewed Sep 19, 2023

View reviewed changes

controllers/kserve_inferenceservice_controller.go Outdated Show resolved Hide resolved

Incorporate review comments

24356ec

Signed-off-by: Vaibhav Jain <vajain@redhat.com>

vaibhavjainwiz mentioned this pull request Sep 20, 2023

Error on reconciliation of odh-model-controller #86

Closed

heyselbi linked an issue Sep 20, 2023 that may be closed by this pull request

Expose generic OpenShift routes for KServe InferenceServices #67

Closed

vaibhavjainwiz added 2 commits September 20, 2023 23:34

Fetch ServiceHost from isvc.Status.URL directly

b8be39c

Signed-off-by: Vaibhav Jain <vajain@redhat.com>

Add Unit test cases

b970010

Signed-off-by: Vaibhav Jain <vajain@redhat.com>

israel-hdez reviewed Sep 25, 2023

View reviewed changes

controllers/inferenceservice_controller.go Outdated Show resolved Hide resolved

controllers/kserve_inferenceservice_controller.go Outdated Show resolved Hide resolved

Update log level

aaa5e73

Signed-off-by: Vaibhav Jain <vajain@redhat.com>

israel-hdez reviewed Sep 26, 2023

View reviewed changes

Update route name to include namespace into it

518d3ae

Signed-off-by: Vaibhav Jain <vajain@redhat.com>

Jooho mentioned this pull request Oct 3, 2023

Change in domain templates opendatahub-io/kserve#96

Closed

vaibhavjainwiz added 2 commits October 5, 2023 21:49

Fix RBAC

caf432b

Signed-off-by: Vaibhav Jain <vajain@redhat.com>

honor user specified url scheme

f3569a5

Signed-off-by: Vaibhav Jain <vajain@redhat.com>

openshift-ci bot added the lgtm label Oct 6, 2023

israel-hdez approved these changes Oct 6, 2023

View reviewed changes

openshift-ci bot added the approved label Oct 6, 2023

openshift-ci bot merged commit f076018 into opendatahub-io:main Oct 6, 2023
6 checks passed

Jooho added the odh-1.11 label Oct 16, 2023

vaibhavjainwiz mentioned this pull request Oct 17, 2023

Expose generic OpenShift routes for KServe InferenceServices #67

Closed

israel-hdez mentioned this pull request Oct 24, 2023

Enable Istio Virtual Host opendatahub-io/kserve#115

Merged

vaibhavjainwiz deleted the generic_route branch December 1, 2023 09:44

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Expose generic OpenShift routes for KServe InferenceServices #84

Expose generic OpenShift routes for KServe InferenceServices #84

vaibhavjainwiz commented Sep 18, 2023 •

edited

Jooho commented Sep 18, 2023

israel-hdez left a comment

vaibhavjainwiz commented Sep 19, 2023 •

edited

Jooho commented Sep 20, 2023

vaibhavjainwiz commented Sep 20, 2023 •

edited

vaibhavjainwiz commented Sep 20, 2023

vaibhavjainwiz commented Sep 20, 2023

openshift-ci bot commented Sep 20, 2023

vaibhavjainwiz commented Sep 20, 2023

vaibhavjainwiz commented Sep 24, 2023

israel-hdez left a comment

israel-hdez commented Sep 26, 2023

israel-hdez left a comment •

edited

vaibhavjainwiz commented Sep 28, 2023 •

edited

israel-hdez commented Sep 28, 2023

vaibhavjainwiz commented Oct 5, 2023

israel-hdez commented Oct 6, 2023

openshift-ci bot commented Oct 6, 2023

Expose generic OpenShift routes for KServe InferenceServices #84

Expose generic OpenShift routes for KServe InferenceServices #84

Conversation

vaibhavjainwiz commented Sep 18, 2023 • edited

Description

How Has This Been Tested?

Expected behavior

Merge criteria:

Jooho commented Sep 18, 2023

israel-hdez left a comment

Choose a reason for hiding this comment

vaibhavjainwiz commented Sep 19, 2023 • edited

Jooho commented Sep 20, 2023

vaibhavjainwiz commented Sep 20, 2023 • edited

vaibhavjainwiz commented Sep 20, 2023

vaibhavjainwiz commented Sep 20, 2023

openshift-ci bot commented Sep 20, 2023

vaibhavjainwiz commented Sep 20, 2023

vaibhavjainwiz commented Sep 24, 2023

israel-hdez left a comment

Choose a reason for hiding this comment

israel-hdez commented Sep 26, 2023

israel-hdez left a comment • edited

Choose a reason for hiding this comment

vaibhavjainwiz commented Sep 28, 2023 • edited

israel-hdez commented Sep 28, 2023

vaibhavjainwiz commented Oct 5, 2023

israel-hdez commented Oct 6, 2023

openshift-ci bot commented Oct 6, 2023

vaibhavjainwiz commented Sep 18, 2023 •

edited

vaibhavjainwiz commented Sep 19, 2023 •

edited

vaibhavjainwiz commented Sep 20, 2023 •

edited

israel-hdez left a comment •

edited

vaibhavjainwiz commented Sep 28, 2023 •

edited