nvidia-gpu 2.0 make compatible for k8s

[CodeJuan]

Ⅰ. Describe what this PR did

k8s-device-plugin uses environment variables to specify a GPU accelerated container, so we should support it.
In order to be compatible with k8s, pouchd should set nvidia prestart hook if nvidia environment variable was set by user.

Ⅱ. Does this pull request fix one issue?

Ⅲ. Why don't you add test cases (unit test/integration test)? (你真的觉得不需要加测试吗？)

Ⅳ. Describe how to verify it

1. Prerequisites:
   • An instance with a NVIDIA GPU
   • The appropriate GPU driver is installed
2. Run GPU container with nvidia env.

pouch run -it -e NVIDIA_VISIBLE_DEVICES=all -e NVIDIA_DRIVER_CAPABILITIES=all centos:7 bash

3. Exec nvidia-smi in container.
4. If nvidia-smi return error NVIDIA-SMI couldn't find libnvidia-ml.so library in your system. Please make sure that the NVIDIA Display Driver is properly installed and present in your system, then add a link to libnvidia-ml.so

version=`nvidia-smi --help | head -n 1 | awk -F "-- v" '{print $2}'`
ln -s  /usr/lib64/libnvidia-cfg.so.1 /usr/lib64/libnvidia-cfg.so
ln -s /usr/lib64/libnvidia-cfg.so."$version" /usr/lib64/libnvidia-cfg.so.1
ln -s  /usr/lib64/libnvidia-ml.so.1 /usr/lib64/libnvidia-ml.so
ln -s /usr/lib64/libnvidia-ml.so."$version" /usr/lib64/libnvidia-ml.so.1

Ⅴ. Special notes for reviews

[codecov]

Codecov Report

❗ No coverage uploaded for pull request base (master@401e132). Click here to learn what that means.
The diff coverage is 72.72%.

@@            Coverage Diff            @@
##             master    #2434   +/-   ##
=========================================
  Coverage          ?   68.88%           
=========================================
  Files             ?      277           
  Lines             ?    18221           
  Branches          ?        0           
=========================================
  Hits              ?    12552           
  Misses            ?     4250           
  Partials          ?     1419

Flag	Coverage Δ
#criv1alpha1test	31.62% <31.81%> (?)
#criv1alpha2test	35.78% <31.81%> (?)
#integrationtest	40.13% <31.81%> (?)
#nodee2etest	33.16% <31.81%> (?)
#unittest	26.73% <72.72%> (?)

Impacted Files	Coverage Δ
daemon/mgr/spec_hook.go	22.72% <0%> (ø)
daemon/mgr/spec_nvidia_hook.go	76.19% <76.19%> (ø)

[CLAassistant]

All committers have signed the CLA.

[HusterWan]

How about adding defer os.Remove(fullname) here? delete the test-nvidia-container-runtime-hook after test finished

[CodeJuan]

Actually, there is defer function at line22.
Sometimes test was broken unexpected, we should delete the mock file at first.

[HusterWan]

Sorry master Xiong, my fault

[CodeJuan]

Take it easy man 😁

[allencloud]

please add more description to this variable. Is this is local file when executing hook?

[CodeJuan]

Both nvidia-container-runtime-hook and nvidia-container-cli was packaged in pouch rpm.

[allencloud]

Before the graph, I think we should add a general introduction of the usage of PouchContainer with GPU.

[CodeJuan]

done

[zhuangqh]

UT fails @CodeJuan

=== RUN   Test_setNvidiaHook
--- FAIL: Test_setNvidiaHook (0.00s)
	spec_nvidia_hook_test.go:108: setNvidiaHook = exec: "test-nvidia-container-runtime-hook": executable file not found in $PATH, want <nil>
	spec_nvidia_hook_test.go:111: setNvidiaHook = [], want [{ [ prestart] [] <nil>}]
	spec_nvidia_hook_test.go:108: setNvidiaHook = exec: "test-nvidia-container-runtime-hook": executable file not found in $PATH, want <nil>
	spec_nvidia_hook_test.go:111: setNvidiaHook = [], want [{ [ prestart] [] <nil>}]
FAIL
coverage: 10.8% of statements
FAIL	github.com/alibaba/pouch/daemon/mgr	0.593s
make: *** [unit-test] Error 1

[zhuangqh]

LGTM