|
1 | | -### L2 overview |
2 | | -This layer consists of workloads for resource provisioning after Intel Device Plugins Operator is installed and custom resources are created. |
3 | | - |
4 | | -#### dGPU testing workload |
5 | | -The workload used is [clinfo](https://github.com/Oblomov/clinfo), which displays the related information of dGPU card. The OCP buildconfig is leveraged to build clinfo container image and push it to the embedded repository through OCP imagestream. |
6 | | -The Job pod is scheduled on a node with dGPU card and the resource ```gpu.intel.com/i915``` is registered by the dGPU device plugin. |
7 | | -Below operations are verified on OCP-4.11 bare-metal cluster. |
8 | | -To deploy the workload: |
9 | | -``` |
10 | | -oc apply -f clinfo_build.yaml |
11 | | -oc apply -f clinfo_job.yaml |
12 | | -``` |
13 | | -To check the clinfo pod logs: |
14 | | -``` |
15 | | -oc get pods | grep clinfo |
16 | | -oc logs <clinfo_pod_name> |
17 | | -``` |
18 | | - |
19 | | -A sample result for clinfo detecting dGPU card: |
20 | | -``` |
| 1 | +# Verifying Intel Hardware Feature Provisioning |
| 2 | +## Introduction |
| 3 | +After provisioning Intel hardware features on RHOCP, the respective hardware resources are exposed to the RHOCP cluster. The workload containers can request these resources. The following sample workloads help verify if these resources can be used as expected. These sample workloads container images are built and packaged on-premises through [RHOCP BuildConfig](https://docs.openshift.com/container-platform/4.12/cicd/builds/understanding-buildconfigs.html) and pushed to the embedded repository through [RHOCP ImageStream](https://docs.openshift.com/container-platform/4.12/openshift_images/image-streams-manage.html). |
| 4 | +## Prerequisites |
| 5 | +• Provisioned RHOCP 4.12 cluster. Follow steps [here](https://github.com/intel/intel-technology-enabling-for-openshift#provisioning-rhocp-cluster). |
| 6 | +• Provisioning Intel HW features on RHOCP. Follow steps [here](https://github.com/intel/intel-technology-enabling-for-openshift#provisioning-intel-hardware-features-on-rhocp) |
| 7 | +### Verify Intel® Software Guard Extensions (Intel® SGX) Provisioning |
| 8 | +This [SampleEnclave](https://github.com/intel/linux-sgx/tree/master/SampleCode/SampleEnclave) application workload from the Intel SGX SDK runs an Intel SGX enclave utilizing the EPC resource from the Intel SGX provisioning. |
| 9 | +* Build the container image. |
| 10 | +```$ oc apply -f https://raw.githubusercontent.com/intel/intel-technology-enabling-for-openshift/ main/tests/l2/sgx/sgx_build.yaml``` |
| 11 | +* Deploy and run the workload. |
| 12 | +```$ oc apply -f https://raw.githubusercontent.com/intel/intel-technology-enabling-for-openshift/main/tests/l2/sgx/sgx_job.yaml``` |
| 13 | +* Check the results. |
| 14 | +```$ oc get pods |
| 15 | + intel-sgx-job-4tnh5 0/1 Completed 0 2m10s |
| 16 | + intel-sgx-workload-1-build 0/1 Completed 0 30s |
| 17 | +``` |
| 18 | +``` |
| 19 | +$ oc logs intel-sgx-job-4tnh5 |
| 20 | + Checksum(0x0x7fffac6f41e0, 100) = 0xfffd4143 |
| 21 | + Info: executing thread synchronization, please wait... |
| 22 | + Info: SampleEnclave successfully returned. |
| 23 | + Enter a character before exit ... |
| 24 | +``` |
| 25 | +### Verify Intel® Data Center GPU provisioning |
| 26 | +This workload runs [clinfo](https://github.com/Oblomov/clinfo) utilizing the i915 resource from GPU provisioning and displays the related GPU information. |
| 27 | +* Build the workload container image. |
| 28 | +```$ oc apply -f https://raw.githubusercontent.com/intel/intel-technology-enabling-for-openshift/main/tests/l2/dgpu/clinfo_build.yaml ``` |
| 29 | +* Deploy and execute the workload. |
| 30 | +```$ oc apply -f https://raw.githubusercontent.com/intel/intel-technology-enabling-for-openshift/main/tests/l2/dgpu/clinfo_job.yaml``` |
| 31 | +* Check the results. |
| 32 | +``` |
| 33 | + $ oc get pods |
| 34 | + intel-dgpu-clinfo-1-build 0/1 Completed 0 3m20s |
| 35 | + intel-dgpu-clinfo-56mh2 0/1 Completed 0 35s |
| 36 | +``` |
| 37 | +``` |
| 38 | +$ oc logs intel-dgpu-clinfo-56mh2 |
21 | 39 | Platform Name Intel(R) OpenCL HD Graphics |
22 | | -Number of devices 1 |
23 | | - Device Name Intel(R) Graphics [0x56c1] |
| 40 | + Number of devices 1 |
| 41 | + Device Name Intel(R) Data Center GPU Flex Series 140 [0x56c1] |
24 | 42 | Device Vendor Intel(R) Corporation |
25 | 43 | Device Vendor ID 0x8086 |
26 | 44 | Device Version OpenCL 3.0 NEO |
27 | | - Driver Version 22.23.23405 |
| 45 | + Device UUID 86800000-c156-0000-0000-000000000000 |
| 46 | + Driver UUID 32322e34-332e-3234-3539-352e33350000 |
| 47 | + Valid Device LUID No |
| 48 | + Device LUID 80c6-4e56fd7f0000 |
| 49 | + Device Node Mask 0 |
| 50 | + Device Numeric Version 0xc00000 (3.0.0) |
| 51 | + Driver Version 22.43.24595.35 |
28 | 52 | Device OpenCL C Version OpenCL C 1.2 |
29 | | - Device Type GPU |
30 | | - Device Profile FULL_PROFILE |
31 | | - Device Available Yes |
32 | | - Compiler Available Yes |
33 | | - Linker Available Yes |
34 | | - Max compute units 128 |
35 | | - Max clock frequency 2100MHz |
36 | | - Device Partition (core) |
37 | | - Max number of sub-devices 0 |
38 | | - Supported partition types None |
39 | | - Supported affinity domains (n/a) |
40 | | - Max work item dimensions 3 |
41 | | - Max work item sizes 1024x1024x1024 |
42 | | - Max work group size 1024 |
43 | | - Preferred work group size multiple 64 |
44 | | - Max sub-groups per work group 128 |
45 | | - Sub-group sizes (Intel) 8, 16, 32 |
46 | | - Preferred / native vector sizes |
47 | | -``` |
| 53 | + Device OpenCL C all versions OpenCL |
| 54 | +``` |
48 | 55 |
|
49 | | -#### SGX test case |
50 | | -The test case used is SGX SDK [Sample Enclave App](https://github.com/intel/linux-sgx/tree/master/SampleCode/SampleEnclave), which launches a simple SGX enclave. Similar to dGPU, OCP buildconfig and imagestream are leveraged for the container image. |
51 | | -The job pod is scheduled on a node enabled with SGX requesting enclave memory resource ```sgx.intel.com/epc```. The resource is created by the SGX device plugin. |
52 | | -Below operations are verified on OCP 4.11 bare-metal cluster. |
53 | | -To build the test case: |
54 | | -``` |
55 | | -oc apply -f sgx_build.yaml |
56 | | -``` |
57 | | -To deploy the job: |
58 | | -``` |
59 | | -oc apply -f sgx_job.yaml |
60 | | -``` |
61 | | -To check the pod logs: |
62 | | -``` |
63 | | -oc get pods | grep sgx |
64 | | -oc logs <sgx_pod_name> |
65 | | -``` |
66 | | -Sample pod result: |
67 | | -``` |
68 | | -Checksum(0x0x7fffac6f41e0, 100) = 0xfffd4143 |
69 | | -Info: executing thread synchronization, please wait... |
70 | | -Info: SampleEnclave successfully returned. |
71 | | -Enter a character before exit ... |
72 | | -``` |
73 | | -On the node, the updated resources are: |
74 | | -``` |
75 | | -oc describe <node name> | grep sgx.intel.com |
76 | | -sgx.intel.com/enclave 1 1 |
77 | | -sgx.intel.com/epc 5Mi 5Mi |
78 | | -sgx.intel.com/provision 0 0 |
79 | | -``` |
| 56 | +## See Also |
| 57 | +For Intel SGX demos on vanilla Kubernetes, refer to [link](https://github.com/intel/intel-device-plugins-for-kubernetes/tree/main/demo/sgx-sdk-demo) |
| 58 | +For GPU demos on vanilla Kubernetes, refer to [link](https://github.com/intel/intel-device-plugins-for-kubernetes/tree/main/demo/intel-opencl-icd) |
0 commit comments