jenkinsci / pyenv-pipeline-plugin

Execute commands in Python virtualenvs in Jenkins Pipeline DSL
https://plugins.jenkins.io/pyenv-pipeline/
MIT License
33 stars 15 forks source link

Need an option to start with clean new Env every build #10

Open tal5ab opened 6 years ago

tal5ab commented 6 years ago

Right now i am forced to uninstall or use --upgrade --force-reinstall to get my latest build installed for testings Will be nice to have an option that starts a new env every build

cstarner commented 6 years ago

This would be a nice feature. At the very least, I think that we would need a setting that determines whether or not the workspace is cleaned up after the build; one that defaults to True.

As I see it, we can go about the actual cleanup in a couple of ways.

  1. We can hook into the Jenkins Build Process, and cleanup after ourselves automatically.
  2. We can add another DSL Step to call and cleanup the virtualenvs.
  3. We can track all of the virtualenvs we create during a build; if we run into an instance of a virtualenv that already exists, but we didn't create, we recreate it.

2 I think can be ruled out right away; this project has already had to create more than enough steps and I am reluctant to add more. 1 would be the cleanest, but the approach of 3 would ensure idempotency, which is more in line with good CI practices.

Do you have a preference either way? Or do you see a different approach to this problem other than the ones I outlined?

tal5ab commented 6 years ago

i would keep it simple , add a flag , and if exist remove the virtualenv if it exist before the script starts

cjw296 commented 5 years ago

Would love this too, but what's the Jenkins Pipeline equivalent of the workspace cleanup plugin?

cstarner commented 5 years ago

cleanWs()

timblaktu commented 3 years ago

I found this plugin thinking its main utility was to clean up pipelines using pyenv and pyenv-virtualenv, replacing instances of sh pyenv virtualenv <python_version> <venv_name> and sh pyenv activate <venv_name> with a single withPythonEnv(venv_name) {}.

Scanning issues I found this one which seems pretty critical for adoption. It's been a while, but I wanted to check.. are you (@cstarner, @cjw296, @tal5ab) all still using this plugin for managing venvs on Jenkins projects, and if so, what are you currently doing to force a clean slate and recreate the venv when you need to?

My initial tests with this plugin have taught me that it does not actually behave as I described above. The docs make it clear that the string arg to withPythonEnv must be a python executable or a pre-existing venv name.

When I correct this and pass it the path to a python interpreter installed by pyenv, e.g. PYTHON_INTERPRETER_PATH:

        PYTHON_VERSION = "3.9.0"
        PYTHON_INTERPRETER_PATH = "/home/jenkins/.pyenv/versions/${PYTHON_VERSION}/bin/python"

I see from this plugin's output that it is using python -m venv to generate the venv, not pyenv-virtualenv, and it sticks the venv at the root of the workspace:

10:17:51  [Pipeline] withPythonEnv
10:16:08  $ /home/jenkins/.pyenv/versions/3.9.0/bin/python --version
10:16:08  $ /home/jenkins/.pyenv/versions/3.9.0/bin/python -m venv /home/jenkins/.jenkins/workspace/eering-services-vsphere-template/.pyenv-home-jenkins-.pyenv-versions-3.9.0-bin-python

None of this is very useful for me, as I really just wanted to replace this pattern:

                    steps{
                        sh """#!/bin/bash -l
                            pyenv virtualenv ${PYTHON_VERSION} ${PYTHON_VENV_NAME} || true
                            pyenv activate ${PYTHON_VENV_NAME}
                            pip3 blah blah
                            python blah blah
                            """
                    }

with something like:

                    withPythonEnv("${PYTHON_VERSION}", "${PYTHON_VENV_NAME}") {
                        sh """#!/bin/bash -l
                            pip3 blah blah
                            python blah blah
                            """
                    }

I appreciate that this plugin must be useful to others, or it wouldn't exist, so I'm just curious if those still actively using this could identify their use case, and/or provide some feedback on how/whether I can achieve my simple desire with this plugin. Perhaps I should just functionalize pyenv_create_and_activate in a shared library instead of using this plugin?