github/kubernetes
1
0
Fork 0
mirror of https://github.com/k3s-io/kubernetes.git synced 2025-07-29 14:37:00 +00:00
Code Issues Projects Releases Wiki Activity

Putting the nvidia-smi command in a try catch to avoid errors.

Browse Source
This commit is contained in:
Matt Bruzek 2017-04-05 16:09:02 -05:00 committed by George Kraft
parent d50cf1e499
commit 258ee22858
1 changed files with 9 additions and 5 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

14
cluster/juju/layers/kubernetes-worker/reactive/kubernetes_worker.py
View File

@ -669,6 +669,15 @@ def enable_gpu():
return
hookenv.log('Enabling gpu mode')
try:
# Not sure why this is necessary, but if you don't run this, k8s will
# think that the node has 0 gpus (as shown by the output of
# `kubectl get nodes -o yaml`
check_call(['nvidia-smi'])
except CalledProcessError as cpe:
hookenv.log('Unable to communicate with the NVIDIA driver.')
hookenv.log(cpe)
return
kubelet_opts = FlagManager('kubelet')
if get_version('kubelet') < (1, 6):
@ -682,11 +691,6 @@ def enable_gpu():
_apply_node_label('gpu=true', overwrite=True)
_apply_node_label('cuda=true', overwrite=True)
# Not sure why this is necessary, but if you don't run this, k8s will
# think that the node has 0 gpus (as shown by the output of
# `kubectl get nodes -o yaml`
check_call(['nvidia-smi'])
set_state('kubernetes-worker.gpu.enabled')
set_state('kubernetes-worker.restart-needed')