6

I am working with GPU supported VMs on GCP Compute Engine.

As OS I use a container optimized version (COS 89-16108.403.47 LTS), which supports simple GPU driver installation by running 'cos-extensions install gpu' via SSH (see Google doc).

This had worked perfectly so far until I started getting an error message saying that the download of some driver signature fails (see full error message below) a couple of days ago and I couldn't get it to work ever since.

Can someone either confirm that I am experiencing a bug here or help me fix this problem?

Many thanks in advance!

~ $ cos-extensions install gpu
Unable to find image 'gcr.io/cos-cloud/cos-gpu-installer:v2.0.3' locally
v2.0.3: Pulling from cos-cloud/cos-gpu-installer
419e7ae5bb1e: Pull complete 
6f6ec2441524: Pull complete 
11d24f918ba9: Pull complete 
Digest: sha256:1cf2701dc2c3944a93fd06cb6c9eedfabf323425483ba3af294510621bb37d0e
Status: Downloaded newer image for gcr.io/cos-cloud/cos-gpu-installer:v2.0.3
I0618 06:33:49.227680    1502 main.go:21] Checking if this is the only cos_gpu_installer that is running.
I0618 06:33:49.258483    1502 install.go:74] Running on COS build id 16108.403.47
I0618 06:33:49.258505    1502 installer.go:187] Getting the default GPU driver version
I0618 06:33:49.285265    1502 utils.go:72] Downloading gpu_default_version from https://storage.googleapis.com/cos-
tools/16108.403.47/gpu_default_version
I0618 06:33:49.353149    1502 utils.go:120] Successfully downloaded gpu_default_version from https://storage.google
apis.com/cos-tools/16108.403.47/gpu_default_version
I0618 06:33:49.353381    1502 install.go:85] Installing GPU driver version 450.119.04
I0618 06:33:49.353461    1502 cache.go:69] error: failed to read file /root/var/lib/nvidia/.cache: open /root/var/l
ib/nvidia/.cache: no such file or directory
I0618 06:33:49.353482    1502 install.go:120] Did not find cached version, installing the drivers...
I0618 06:33:49.353491    1502 installer.go:82] Configuring driver installation directories
I0618 06:33:49.421021    1502 installer.go:196] Updating container's ld cache
I0618 06:33:49.526673    1502 signature.go:30] Downloading driver signature for version 450.119.04
I0618 06:33:49.526712    1502 utils.go:72] Downloading 450.119.04.signature.tar.gz from https://storage.googleapis.
com/cos-tools/16108.403.47/extensions/gpu/450.119.04.signature.tar.gz
E0618 06:33:49.657028    1502 artifacts.go:106] Failed to download extensions/gpu/450.119.04.signature.tar.gz from 
public GCS: failed to download 450.119.04.signature.tar.gz, status: 404 Not Found
E0618 06:33:49.657487    1502 install.go:175] failed to download driver signature: failed to download driver signat
ure for version 450.119.04: failed to download extensions/gpu/450.119.04.signature.tar.gz
Wytrzymały Wiktor
  • 11,492
  • 5
  • 29
  • 37
user15774062
  • 63
  • 3

1 Answers1

2

This seems to be a known issue, you can find it reported here and a similar thread with workarounds here.
Looks like there is a delay between the release of new COS version and release of updated drivers.

However, I ran cos-extensions list just now, and it seems there are drivers available:

$ cos-extensions list
Available extensions for COS version 89-16108.403.47:

[gpu]
450.119.04 [default]
450.80.02

And signatures as well:

$ wget https://storage.googleapis.com/cos-tools/16108.403.47/extensions/gpu/450.119.04.signature.tar.gz
--2021-06-21 12:49:58--  https://storage.googleapis.com/cos-tools/16108.403.47/extensions/gpu/450.119.04.signature.tar.gz
Resolving storage.googleapis.com... 173.194.198.128, 64.233.191.128, 173.194.74.128, ...
Connecting to storage.googleapis.com|173.194.198.128|:443... connected.
HTTP request sent, awaiting response... 200 OK
Length: 4588 (4.5K) [application/octet-stream]
Saving to: '450.119.04.signature.tar.gz'

450.119.04.signature.tar.gz  100%[=============================================>]   4.48K  --.-KB/s    in 0s      

2021-06-21 12:49:58 (62.0 MB/s) - '450.119.04.signature.tar.gz' saved [4588/4588]
Sergiusz
  • 1,175
  • 4
  • 13