> On Aug. 22, 2014, 9:20 a.m., John Dilley wrote:
> >
All the review comments are incorporated excpet one which was getting the
pciDevice value from the host. I have given my reasons below.
Also, one more change was made. In the function "check_for_vGPU_resource" now
we are passing the hostid of the VM also, reason being :
Previously we were doing ssh to the last host in the list we get in setup
class which may not be the host on which vm is deployed or it maybe in a
different cluster than the VM.
So, now we are creating a ssh connection with the host on which VM is
deployed from inside this function so that we get the right information we need.
> On Aug. 22, 2014, 9:20 a.m., John Dilley wrote:
> > test/integration/component/test_deploy_vgpu_vm.py, line 190
> > <https://reviews.apache.org/r/24420/diff/5/?file=665972#file665972line190>
> >
> > Don't hardcode "Group of NVIDIA Corporation GK104GL [GRID K1] GPUs"
> > into the test, run gpu-group-list on the host instead
I think we can stick with this because CS itself hardcodes the pciDevice. When
we try to create a service offering from CCP UI, we don't pass the pciDevice,
we just pass the card type.
If we pass NVIDIA GRID K2, it itslef picks up the pciDevice as "Group of NVIDIA
Corporation GK104GL [GRID K2] GPUs"
If we pass NVIDIA GRID K1, it picks "Group of NVIDIA Corporation GK107GL [GRID
K1] GPUs"
So, if the host has pciDevice value anything other than this then the VM
deployment will fail by itself.
Thats why i think we can ignore this in here.
> On Aug. 22, 2014, 9:20 a.m., John Dilley wrote:
> > test/integration/component/test_deploy_vgpu_vm.py, line 215
> > <https://reviews.apache.org/r/24420/diff/5/?file=665972#file665972line215>
> >
> > Don't hardcode "Group of NVIDIA Corporation GK104GL [GRID K2] GPUs"
> > into the test, run gpu-group-list on the host instead
I think we can stick with this because CS itself hardcodes the pciDevice. When
we try to create a service offering from CCP UI, we don't pass the pciDevice,
we just pass the card type.
If we pass NVIDIA GRID K2, it itslef picks up the pciDevice as "Group of NVIDIA
Corporation GK104GL [GRID K2] GPUs"
If we pass NVIDIA GRID K1, it picks "Group of NVIDIA Corporation GK107GL [GRID
K1] GPUs"
So, if the host has pciDevice value anything other than this then the VM
deployment will fail by itself.
Thats why i think we can ignore this in here.
- abhinav
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/24420/#review51272
-----------------------------------------------------------
On Aug. 22, 2014, 2:09 p.m., abhinav roy wrote:
>
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/24420/
> -----------------------------------------------------------
>
> (Updated Aug. 22, 2014, 2:09 p.m.)
>
>
> Review request for cloudstack, Doug Clark, John Dilley, sailaja mada, and
> Sanjay Tripathi.
>
>
> Repository: cloudstack-git
>
>
> Description
> -------
>
> This Diff includes vGPU VM lifecycle tests as well as a function to validate
> the vGPU resources present in a vGPU VM deployed on CS. It has the following
> tests :
>
> 1. A function to validate vGPU resources in a VM. It does validation on both
> the host as well as on CS.
> It includes following lifecycle tests.
>
>
> 2. Deploy VM
> 3. Stop VM
> 4. Start VM
> 5. Restore VM
> 6. Reboot VM
> 7. Destroy VM
> 8. Recover VM
>
>
> NOTE : VM lifecycle test cases were included in this script after the 1st
> rouund of review with Doug. We decided to include those tests here just to
> remove the overhead of registering template again. so, now there is no use of
> this review request https://reviews.apache.org/r/24425/
>
>
> Diffs
> -----
>
> test/integration/component/test_deploy_vgpu_vm.py fd3f374
>
> Diff: https://reviews.apache.org/r/24420/diff/
>
>
> Testing
> -------
>
> Testing :
>
> 1. Executed the script on non GPU test setup and ensured tests being skipped.
> 2. Executed on K2 GPU drivers installed setup and ensured all lifecycle test
> cases are working fine and the function is checking for the vGPU resources in
> a vGPU VM.
>
>
> Thanks,
>
> abhinav roy
>
>
