[DATALAB-2436] [Dataproc] Investigate GPU usage for slave

• Type: Task
• Status: Closed
• Priority: Minor
• Resolution: Done
• Fix Version/s: v.2.5
• Component/s: DataLab Main
• Security Level: Public (Regular Issues)
• Labels: Debian DevOps GCP
• Epic Link: Investigation
• Sprint: Apache DataLab release 2.5
• Story Points: 4
• Assignee: Mykola Bodnar
• Reporter: Vira Vitanska
• Created: 10/Jun/21 11:28
• Updated: 12/Jul/21 10:27
• Resolved: 12/Jul/21 10:27

Description

1. In GCP GPU kernel is pointed for slave, but not for second worker.

But nvidia-smi command is not run successfully for slave (not second worker). Why?

datalab-user@gcp-10-06-21-proj-local-des-dp-gpu-c-w-0:~$ nvidi-smiCommand 'nvidi-smi' not found, did you mean: command 'nvidia-smi' from deb nvidia-340 command 'nvidia-smi' from deb nvidia-utils-390Try: apt install <deb name>datalab-user@gcp-10-06-21-proj-local-des-dp-gpu-c-w-0:~$

2. Running code of GPU usage (https://weeraman.com/put-that-gpu-to-good-use-with-python-e5a437168c01) does not show GPU usage for Dataproc master via ssh. Why?