tests_l2: Update readme #104
Conversation
|
@mregmi @uMartinXu please review |
|
LGTM |
| @@ -28,6 +28,10 @@ $ oc logs intel-sgx-job-4tnh5 | |||
| ``` | |||
| ### Verify Intel® Data Center GPU provisioning | |||
| This workload runs [clinfo](https://github.com/Oblomov/clinfo) utilizing the i915 resource from GPU provisioning and displays the related GPU information. | |||
| * Before deploying any workload, please ensure to set access for ```/dev/``` folder on the GPU node, with the command below: | |||
|
|
|||
There was a problem hiding this comment.
Looks like we need a selinux policy for the workload pod? @mregmi? if it is the case, do we have an existing one can be used by the workload pod?
There was a problem hiding this comment.
We can use container_device_t selinux policy. Works with a custom ocp scc.
I wonder how feasible it is to for users to declare it for every workload/e2e with other operators/dependencies. @mregmi @uMartinXu thoughts?
There was a problem hiding this comment.
#107 with container_device_t does not include enough permissions to run. Might require more permissions. For this release, I suggest we use this PR. Will keep the other PR open and verify with correct permissions. @mregmi @uMartinXu
There was a problem hiding this comment.
Yea for now lets go with bool. we will have to give extra permissions later
|
@vbedida79, can you add the target milestone for this PR? |
|
We now are working on the issue #107 and need more time to figure out the proper solution. So we will close this PR |
tests/l2/README.md
Outdated
| @@ -28,6 +28,10 @@ $ oc logs intel-sgx-job-4tnh5 | |||
| ``` | |||
| ### Verify Intel® Data Center GPU provisioning | |||
| This workload runs [clinfo](https://github.com/Oblomov/clinfo) utilizing the i915 resource from GPU provisioning and displays the related GPU information. | |||
| * Before deploying any workload, please ensure to set access for ```/dev/``` folder on the GPU node, with the command below: | |||
There was a problem hiding this comment.
Can we change it to:
To work around issue, please run the below command on the node with the GPU where your workload will run.
vbedida79
force-pushed
the
patch-310723-1
branch
from
9c6d90d
to
53ad807
Compare
Signed-off-by: vbedida79 <[email protected]>
vbedida79
force-pushed
the
patch-310723-1
branch
from
53ad807
to
035b277
Compare
|
Looks good to me! |
Updated readme for tests/l2/gpu- set boolean value of
setsebool container_use_devices onfor the GPU node.This is needed for GPU workloads to run and access host devices. Will be removed after #107 resolved
Signed-off-by: vbedida79 [email protected]