Project

General

Profile

Actions

Bug #47448

open

mgr/diskprediction_local: add unittest's to ensure correct python dependencies

Added by Karanraj Chauhan over 3 years ago. Updated over 3 years ago.

Status:
Fix Under Review
Priority:
Normal
Category:
diskprediction_local
Target version:
-
% Done:

0%

Source:
Tags:
Backport:
Regression:
No
Severity:
3 - minor
Reviewed:
Affected Versions:
ceph-qa-suite:
Pull request ID:
Crash signature (v1):
Crash signature (v2):

Description

Currently, there are no unit tests that ensure that models in the diskprediction_local module work correctly. In some cases, users installed versions of scikit-learn that are do not comply with the requirements.txt in this module. As a result the ProphetStor models (default for this module) threw errors, because these were not compatible with the user installed version of scikit-learn. To prevent such issues in the future, we could add
1. unittest to ensure all models work without errors with versions as specified in the requirements.txt
2. unittest to ensure users install dependencies only as specified in the requirements.txt

Actions #1

Updated by Yaarit Hatuka over 3 years ago

  • Assignee set to Karanraj Chauhan
Actions #2

Updated by Karanraj Chauhan over 3 years ago

To address point 1 above, I'm thinking of adding python "unittest"s as done for other mgr modules such as dashboard, insights, etc.

For point 2 above, should it also just be a unittest inside the diskprediction_local module? Or is there some way of enforcing dependencies get installed just like in requirements.txt?

WDYT?

Actions #3

Updated by Yaarit Hatuka over 3 years ago

Hi Karan,

I started looking more in depth at the second type of test, so I'll add it myself (will link here).

Actions #4

Updated by Yaarit Hatuka over 3 years ago

  • Status changed from New to Fix Under Review
  • Pull request ID set to 37513
Actions

Also available in: Atom PDF