yahoo-eng-team team mailing list archive

[sean mooney <1975711@xxxxxxxxxxxxxxxxxx>]

im settign this to opipion for now since this seam to be distro-specific

we have confirmed that this can happen on fedora 36 but we have also confirmed that
3.10 work on debian and nixos and we have ci jobs running non voting unit test on 3.10 on ubuntu 22.04

so in general this does not appear to be a nova but it looks like its
either a fedroa issue or its related to the tox/pip version that are
bing used.

we may be able to work around it in nova but I'm not sure we should do
that or can do that without change what we test and how we test in an
undesireable way.

** Changed in: nova
       Status: New => Opinion

-- 
You received this bug notification because you are a member of Yahoo!
Engineering Team, which is subscribed to OpenStack Compute (nova).
https://bugs.launchpad.net/bugs/1975711

Title:
  tox hangs due to pip backtracking during virtualenv generation

Status in OpenStack Compute (nova):
  Opinion

Bug description:
  Description
  ===========
  On a fresh checkout of nova, running tox -e pep8 results in the process maxing out a CPU core and seemingly getting stuck (I terminated it after 30 minutes of no progress).

  I believe this is due to pip trying to find a set of packages that
  exactly satisfy cross-requirements of all dependencies, checking
  multiple progressively older versions of each package until the tree
  becomes too complex to handle at all.

  Steps to reproduce
  ==================

  * Make a fresh checkout of nova, a shallow one works since we only need master:
  git clone --depth 1 https://opendev.org/openstack/nova.git nova
  This makes sure the tox virtualenv from an existing checkout isn't reused.

  * From within the repo, run tox pep8 with verbosity to see pip output:
  $ tox -vvv -e pep8

  Expected result
  ===============
  Tox successfully sets up its virtualenv and runs pep8.

  Actual result
  =============
  pip downloads several versions of packages, outputting a large amount of messages like these for a few packages along the way:

  INFO: pip is looking at multiple versions of certifi to determine which version is compatible with other requirements. This could take a while.
    Downloading certifi-2020.4.5-py2.py3-none-any.whl (156 kB)
       |████████████████████████████████| 156 kB 81.6 MB/s 
    Downloading certifi-2019.11.28-py2.py3-none-any.whl (156 kB)
       |████████████████████████████████| 156 kB 86.8 MB/s 
    Downloading certifi-2019.9.11-py2.py3-none-any.whl (154 kB)
       |████████████████████████████████| 154 kB 79.5 MB/s 
    Downloading certifi-2019.6.16-py2.py3-none-any.whl (157 kB)
       |████████████████████████████████| 157 kB 71.6 MB/s 
    Downloading certifi-2019.3.9-py2.py3-none-any.whl (158 kB)
       |████████████████████████████████| 158 kB 84.7 MB/s
  INFO: This is taking longer than usual. You might need to provide the dependency resolver with stricter constraints to reduce runtime. If you want to abort this run, you can press Ctrl + C to do so. To improve how pip performs, tell us what happened here: https://pip.pypa.io/surveys/backtracking

  Eventually it seems to get completely stuck after one of those
  downloads, maxing out a CPU core and seemingly making no more progress
  until terminated.

  Environment
  ===========
  This happens in dev environments no in Openstack deployments. We've reproduced it on Fedora 35 and 36, I would expect others to be similarly impacted. Some system python env info:

  $ python -V
  Python 3.10.4

  $ pip show pip
  Name: pip
  Version: 21.3.1

  $ pip show tox
  Name: tox
  Version: 3.24.5

  Logs & Configs
  ==============
  Reproduced on a fresh checkout with no altered configs.

To manage notifications about this bug go to:
https://bugs.launchpad.net/nova/+bug/1975711/+subscriptions