-
Notifications
You must be signed in to change notification settings - Fork 121
"Python: Command Not Found" in Ubuntu 20.04LTS During Node Preparation #363
Comments
Thanks for opening this issue. 20.04 isn't supported yet, but will be in the next version. |
- Resolves #363 - Add CentOS 7.8 support - Remove CentOS 8 support - Add Datacenter-Core-20H2 - Unblock -gs, -g2, -gensecond support - Update Nvidia Tesla driver to 470.57.02 and CUDA 11.4 - Update Nvidia Grid driver to 470.63.01 and CUDA 11.4 - Update LIS driver to 3.4.5
@emattiza @alfpark @emattiza |
@coolstate I just used an older Ubuntu image. We now use a different azure service for our needs after a refactor. Good luck! |
If you enable unverified versions you can select the dvsm from Windows which is referenced in the Microsoft tutorials and the commands will work on startup. |
- Resolves #363 - Add CentOS 7.8 support - Remove CentOS 8 support - Add Datacenter-Core-20H2 - Unblock -gs, -g2, -gensecond support - Update Nvidia Tesla driver to 470.57.02 and CUDA 11.4 - Update Nvidia Grid driver to 470.63.01 and CUDA 11.4 - Update LIS driver to 3.4.5
- Resolves #363 - Add CentOS 7.8 support - Remove CentOS 8 support - Add Datacenter-Core-20H2 - Unblock -gs, -g2, -gensecond support - Update Nvidia Tesla driver to 470.57.02 and CUDA 11.4 - Update Nvidia Grid driver to 470.63.01 and CUDA 11.4 - Update LIS driver to 3.4.5
Problem Description
When initiating a pool with shipyard pool add, using the microsoft-azure-batch publisher 20-04-lts ubuntu image, the python executable is not available in the newly created node's $PATH per Ubuntu's new policy of handling the system python executable name and the 2-to-3 migration. Because nodeprep.sh references "python" as the executable name in a couple spots, the executable fails to be found, and the node preparation start task fails.
Batch Shipyard Version
shipyard, version 3.9.1
Steps to Reproduce
generate a pool with an ubuntu-server-container, microsoft-azure-batch, 20-04-lts image
(shipyard pool add with a simple pool config for a single dedicated node)
Expected Results
Successful pool deploy
Actual Results
stderr.txt contains python:command not found (nodeprep.sh:385)
Redacted Configuration
pool.yaml
Additional Logs
Additional Comments
This can be worked around by specifying a system python symlink in the additional_node_prep > pre commands. Happy to submit a pull request with changes to nodeprep.sh that test and direct to the appropriate python executable name to run so that the additional_node_prep pre-commands are no longer needed. I can always, of course, just pick a different image as well!
The text was updated successfully, but these errors were encountered: