The problem

When trying to run the integration tests for the slurmd charmed operator, our CI pipeline fails with the following stacktrace:

Traceback (most recent call last):
  File "/home/runner/work/slurmd-operator/slurmd-operator/tests/integration/test_charm.py", line 48, in test_build_and_deploy
    await asyncio.gather(
  File "/home/runner/work/slurmd-operator/slurmd-operator/.tox/integration/lib/python3.10/site-packages/juju/model.py", line 1820, in deploy
    resources = await self.add_local_resources(application_name,
  File "/home/runner/work/slurmd-operator/slurmd-operator/.tox/integration/lib/python3.10/site-packages/juju/model.py", line 2052, in add_local_resources
    data = p.read_text() if p.exists() else ''
  File "/usr/lib/python3.10/pathlib.py", line 1135, in read_text
    return f.read()
  File "/usr/lib/python3.10/codecs.py", line 322, in decode
    (result, consumed) = self._buffer_decode(data, self.errors, final)
UnicodeDecodeError: 'utf-8' codec can't decode byte 0x8b in position 1: invalid start byte

This error is raised when we try to attach our resource lbnl-nhc-1.4.3.tar.gz to the slurmd application. NHC (Node Health Check) is a utility for measuring the health of compute nodes in your cluster; it is commonly used to collect observability metrics. When we try to attach NHC, pathlib.Path(...).read_text(...) tries to decode the .tar.gz file as an utf-8 text file rather than a binary file. This causes python-libjuju to raise the above UnicodeDecodeError which then torpedoes our CI pipeline. The block of code below is what is responsible for this issue:

This could potentially be fixed by switch the implementation to pathlib.Path(...).read_bytes(...)

Steps to reproduce

1. Assume that you already have a cloud and Juju controller set up...
2. git clone -b jaime/drop-focal-support [email protected]:jaimesouza/slurmd-operator.git
3. cd slurmd-operator
4. tox run -e integration
5. Wait a few minutes...
6. See stacktrace with UnicodeDecodeError