tanishqgandhi1908 opened a new issue, #8466:
URL: https://github.com/apache/texera/issues/8466

   ### Feature Summary
   
   Every computing unit runs the same image, fixed when the cluster is 
installed. ML models needing a different Python version, a system package, or a 
library built from source cannot run — a Python virtual environment holds pip 
packages and nothing else. AlphaFold 3 fails all three ways at once: it needs 
Python ≥ 3.12 where the engine image has 3.10, it is not on PyPI and compiles a 
C++ extension, and its data pipeline shells out to jackhmmer and nhmmer, which 
are executables.
   
   This lets an administrator register an image reference from a public 
registry. Texera checks it is a Texera computing-unit image and records the 
digest behind the reference. The image then appears in a dropdown when creating 
a computing unit; that unit runs it, and nobody else is affected.
   
   ### Proposed Solution or Design
   
   An administrator registers a reference — a plain owner/name:tag or a Docker 
Hub page address — on an admin page. Texera reads the image's manifest and 
config blob to check its start command runs computing-unit-master, which means 
it was built FROM the Texera computing-unit image, and resolves the digest the 
reference points at. A user then picks the image when creating a computing unit.
   
   Four decisions worth stating:
   
   - Nothing is copied and no registry is added. Units pull the reference the 
same way the default computing-unit image is already pulled, so no new 
infrastructure and no node configuration is required.
   - The reference is pinned to the digest resolved at registration, so a tag 
its owner moves later cannot change what a unit runs.
   - Validation happens at registration, in front of the administrator who 
typed the reference, rather than when a user's unit fails to start.
   - The image is chosen when a unit is created and fixed for its lifetime; a 
running unit's image cannot be swapped.
   
   Out of scope: building images, private registries, per-operator images.
   
   
   ### Affected Area
   
   Deployment / Infrastructure, Workflow UI


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]

Reply via email to