Test ModelMesh Upgrade to v0.11

[heyselbi]

Once the upgrade to v0.11 happens, we need to test:

• Sample model deployment quickstart
• Sample Model Deployment and Autoscaler quickstart
• Sample Model Deployment by using a Persistent Volume Claim quikstart

[heyselbi]

Sample model deployment quickstart is successful with the upgrade

[heyselbi]

PVC test is successful:

[selbi@fedora pvc]$ export MODEL_NAME=isvc-pvc-storage-path
[selbi@fedora pvc]$ export HOST_URL=$(oc get route ${MODEL_NAME} -ojsonpath='{.spec.host}' -n ${TEST_MM_NS})
[selbi@fedora pvc]$ export HOST_PATH=$(oc get route ${MODEL_NAME}  -ojsonpath='{.spec.path}' -n ${TEST_MM_NS})
[selbi@fedora pvc]$ curl   --silent --location --fail --show-error --insecure https://${HOST_URL}${HOST_PATH}/infer -d  @../common_manifests/input-sklean.json

{"model_name":"isvc-pvc-storage-path__isvc-5157aa043b","outputs":[{"name":"predict","datatype":"INT64","shape":[1,1],"data":[8]}]}

[selbi@fedora pvc]$ export MODEL_NAME=isvc-pvc-storage-uri  export MODEL_NAME=isvc-pvc-storage-uri
[selbi@fedora pvc]$ export HOST_URL=$(oc get route ${MODEL_NAME} -ojsonpath='{.spec.host}' -n ${TEST_MM_NS})
export HOST_PATH=$(oc get route ${MODEL_NAME}  -ojsonpath='{.spec.path}' -n ${TEST_MM_NS})
[selbi@fedora pvc]$ curl   --silent --location --fail --show-error --insecure https://${HOST_URL}${HOST_PATH}/infer -d @../common_manifests/input-sklean.json 
{"model_name":"isvc-pvc-storage-uri__isvc-5157aa043b","outputs":[{"name":"predict","datatype":"INT64","shape":[1,1],"data":[8]}]}

[heyselbi]

HPA and gRPC tests are successful

[selbi@fedora hpa]$ export HOST_URL=$(oc get route example-onnx-mnist -ojsonpath='{.spec.host}' -n ${TEST_MM_NS})
export HOST_PATH=$(oc get route example-onnx-mnist  -ojsonpath='{.spec.path}' -n ${TEST_MM_NS})
[selbi@fedora hpa]$ curl   --silent --location --fail --show-error --insecure https://${HOST_URL}${HOST_PATH}/infer -d @../common_manifests/input-onnx.json 
{"model_name":"example-onnx-mnist__isvc-b29c3d91f3","model_version":"1","outputs":[{"name":"Plus214_Output_0","datatype":"FP32","shape":[1,10],"data":[-8.233053,-7.7497034,-3.4236815,12.3630295,-12.079103,17.266596,-10.570976,0.7130762,3.321715,1.3621228]}]}
[selbi@fedora hpa]$ oc port-forward --address 0.0.0.0 service/modelmesh-serving 8033 -n ${TEST_MM_NS}
Forwarding from 0.0.0.0:8033 -> 8033
Handling connection for 8033
^C[selbi@fedora hpa]$ oc get hpa
NAME                         REFERENCE                               TARGETS         MINPODS   MAXPODS   REPLICAS   AGE
modelmesh-serving-ovms-1.x   Deployment/modelmesh-serving-ovms-1.x   <unknown>/80%   1         1         1          3m15s
[selbi@fedora hpa]$ oc annotate servingruntime ovms-1.x serving.kserve.io/max-scale=3
servingruntime.serving.kserve.io/ovms-1.x annotated
[selbi@fedora hpa]$ oc annotate servingruntime ovms-1.x serving.kserve.io/min-scale=2
servingruntime.serving.kserve.io/ovms-1.x annotated
[selbi@fedora hpa]$  oc annotate servingruntime ovms-1.x  serving.kserve.io/targetUtilizationPercentage=50
servingruntime.serving.kserve.io/ovms-1.x annotated
[selbi@fedora hpa]$ oc get hpa
NAME                         REFERENCE                               TARGETS         MINPODS   MAXPODS   REPLICAS   AGE
modelmesh-serving-ovms-1.x   Deployment/modelmesh-serving-ovms-1.x   <unknown>/50%   2         3         1          5m45s

Observing the scaling up pods from 1 to 2 as well.

[Jooho]

live builder has some issues with building a new image now. QE team will do upgrade testing so we can skip the upgrade test this time.

[heyselbi]

Last test is successful with specific conditions. See follow up ticket.

[heyselbi]

Follow up dev ticket: opendatahub-io/opendatahub-operator#573