Pulp 2 15 3 Test Result Summary¶

Summary

Started On: March 13, 2018
Finished On: March 19, 2018
Version: 2.15.3
Result: Success
Build URL: N/A

Summary

This release included bug fixes. The QE tasks in the release process include making sure automation jobs pass and upgrade automation jobs pass. 2.15.2 was released on F27.

Pulp QE approve the release of pulp 2.15.3

Beta History
Beta1 : March 13, 2018

Issues Addressed
================

Docker Support
3286 Sync logic should not assume on schema1 by default existence
3357 docker_distributor_web is racy, not atomic
3258 Update with new recipe for v2s1 manifest upload

Pulp
3246 Update branching.rst to match pup-0003
3347 Advise users to use `setsebool` to set pulp_manage_rsync
3386 PEP8 Pulp's docs/conf.py

RPM Support
3307 import_upload: improve data validation
3339 Missing docs on how to import RPM package which has rich dependencies in Requires
3353 Missing docs on importing RPM repository module metadata
3411 Document Implications that Pulp2 does not support metalink for rpm syncing

Summary:

Most of the issues are documentation-related. The candidates for automation are:
Issue #3286: Sync logic should not assume on schema1 by default existence
Issue #3357: docker_distributor_web is racy, not atomic
Issue #3307: import_upload: improve data validation
#3286 might be automatable. To automate this issue, we need a test that does the following:

Create a Docker repository. Let its feed reference a Docker image that only has a v2 representation, and not a v1 representation.
Sync the repository. Verify that the the sync succeeded.
Publish the repository. Verify that:
The v1 version of the docker image may not be downloaded.
The v2 version of the docker image may be downloaded.
I'm not sure how hard this is. The main impediment is my own poor knowledge of Docker. I think it's doable with our existing tooling.

#3357 has a reproducer that requires patching source code. I don't want to do this, and have asked for an alternate reproducer - say, one that requires setting filesystem permissions. The response has basically been "that's hard. Trust that the people affected have verified the fix."

#3307 can be tested quite easily. We already have tests that upload files to Pulp. To automate this issue, we need a test that starts and upload, but provides some invalid parameters. Not too hard. I could bang this out in a day.