Details
-
New Feature
-
Status: Resolved
-
Major
-
Resolution: Fixed
-
None
-
None
-
Minor
-
Not applicable
Description
Try to use gpu with nvidia driver on Azure.
Azure VMSize: Standard_nc6s_v3
Required field in pod/job:
resources:
limits:
nvidia.com/gpu: 1
hpcc/values.yaml additional setting:
thor:
- name: thor
prefix: thor
numWorkers: 2
maxJobs: 4
maxGraphs: 2
workerResources:
nvidia.com/gpu: 1
Result yaml:
thor:
logging:
detail: 80
maxGraphs: 2
maxJobs: 4
name: thor
numWorkers: 2
prefix: thor
vaults:
ecl: null
ecl-user: null
storage: null
workerResources:
nvidia.com/gpu: 1.......
image: hpccsystems/platform-gnn-gpu:8.2.4-az
imagePullPolicy: IfNotPresent
name: thorworker-HPCC_JOBNAME
resources:
limits:
nvidia.com/gpu: 1
error :
kubectl logs thor-eclagent-79c66db78d-wzslt
00000000 USR ERR 2021-07-27 01:37:04.325 1 1 UNK "99: agentexec.cpp(398) : Error processing config file
: Error loading configuration file /etc/config/thor.yaml (invalid yaml): IPropertyTree: UNKNOWN ERROR CODE: -1
Invalid xml attribute: @nvidia.com/gpu"
Two students will try to test GNN with GPU on Azure. But it is optional