You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
If a DHC user starts off w/ a pre-built container, and then moves to building from source, the difference in Python dependencies can cause issues. For example, if one install pathway for one reason or another resolves numpy to 1.24.1 instead of 1.23.5, then that install will break and not the other (see #3422).
Additionally, systems that run DHC workers natively (like DnD) need to know which sets of Python dependencies are supported and tested against, so that they can install those and have confidence in stability.
In addition to stability, if a user migrated from pure DHC containers to a system that runs DHC workers natively, they (understandably) would expect a similar experience (and therefore dependencies). This cannot be guaranteed if there isn't a central source of supported Python dependencies to begin with.
Description
For the same DHC version, e.g. 0.20.1, Python dependencies differ between pre-built Docker images via https://deephaven.io/core/docs/tutorials/quickstart/ and built-from-source images via https://deephaven.io/core/docs/how-to-guides/launch-build/.
If a DHC user starts off w/ a pre-built container, and then moves to building from source, the difference in Python dependencies can cause issues. For example, if one install pathway for one reason or another resolves numpy to 1.24.1 instead of 1.23.5, then that install will break and not the other (see #3422).
Additionally, systems that run DHC workers natively (like DnD) need to know which sets of Python dependencies are supported and tested against, so that they can install those and have confidence in stability.
In addition to stability, if a user migrated from pure DHC containers to a system that runs DHC workers natively, they (understandably) would expect a similar experience (and therefore dependencies). This cannot be guaranteed if there isn't a central source of supported Python dependencies to begin with.
Steps to reproduce
pip freeze
via Docker terminal on both containersExpected results
Equivalent Python dependencies (at least as much is feasibly possible).
Actual results
Different Python dependencies.
Pre-built image:
Built-from-source image:
Additional details and attachments
Versions
The text was updated successfully, but these errors were encountered: