[FEATURE REQUEST] salt-pip - reinstall pip packages on upgrades - like a local-requirements.txt file in /etc/salt/

[ITJamie]

Is your feature request related to a problem? Please describe.
When you upgrade salt packages sometimes the salt-pip packages stop working.

Describe the solution you'd like
a local-requirements.txt file which exists in /etc/salt . that file should automatically have salt-pip install run against it after package upgrades.

Describe alternatives you've considered
us os package hooks but thats a pain to maintain, having a regular cron which just re-installs packages...

Please Note
If this feature request would be considered a substantial change or addition, this should go through a SEP process here https://github.com/saltstack/salt-enhancement-proposals, instead of a feature request.

[dwoz]

@ITJamie The intent of extras-3.10 is that it won't get touched by the package manager and everything there should work provided the python minor version does not change. Are you able to elaborate on what is causing the breakage?

[ITJamie]

this happened again recently when the python version was updated in 3008.2 > 3008.3 and 3006.26 > 3006.27

this is quite painful when you have custom grains which require packages or when your using things like pyjsonb returner on the master

[ITJamie]

To clarify the proposed behaviour: support an optional /etc/salt/local-requirements.txt, with an equivalent file under the Windows Salt configuration directory.
After installing the updated onedir runtime, the package/installer should run the new runtime’s salt-pip install -r before starting or restarting Salt services.
This would reinstall dependencies for the new Python version, including compatible compiled extensions, rather than reuse the previous interpreter’s extras directory. This timing matters for custom grains and master returners whose dependencies must already be available when Salt starts.
If the file is absent, existing behaviour should remain unchanged. If installation fails, the upgrade should report that failure clearly before restarting affected services.

[dwoz]

This feature request was moved out of the issue tracker as part of a backlog cleanup — there is no active development commitment behind it. If you intend to implement this, please leave a comment and a maintainer can convert the discussion back to an issue to track the PR.

[ITJamie]

@dwoz i did have a pr open against the previous issue that this discussion was

PR: #70426
Issue: #68219