[DATALAB-1692] [GCP]: Running code for Data Engine Service does not use GPU

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Done
• Affects Version/s: v.2.5
• Fix Version/s: v.2.5
• Component/s: DataLab Main
• Labels: Debian DevOps GCP Known_issues(release2.3) Known_issues(release2.4)
• Epic Link: DataLab Compute Engine
• Sprint: Apache DataLab release 2.5
• Assignee: Mykola Bodnar
• Reporter: Vira Vitanska
• Created: 03/Apr/20 11:17
• Updated: 30/Jun/21 12:19
• Resolved: 30/Jun/21 12:19

Description

Preconditions:

1. Data Engine Service is created with shape for GPU

Steps to reproduce:

1. Run code that use GPU for Data Engine Service

Actual result:

1. GPU is not used

Expected result:

1. GPU is used

---

Code which use GPU
1. https://weeraman.com/put-that-gpu-to-good-use-with-python-e5a437168c01
2. https://www.geeksforgeeks.org/running-python-script-on-gpu/

---

If use command 'nvidia-smi' on Dataproc - the command is not found

---

Add new parameters in data_engine_service_description.json: Master/Slave GPU type/count.

 

Issue Links

• is related to: DATALAB-2232 [Back-end][GCP]: Add new parameters for Dataproc
• is related to: DATALAB-2233 [Front-end][GCP]: Add new parameters for Dataproc
• is related to: DATALAB-2430 [Back-end][GCP]: Add new parameters for Apache Standalone cluster
• is related to: DATALAB-2431 [Front-end][GCP]: Add new GPU parameters for Apache Standalone cluster
• relates to: DATALAB-1763 [Compute] Write playbook for GPU usage