[RHELC-1130] Fix failing yum transaction validation

[hosekadam]

The problem is related to how we remove the excluded packages. When “rpm -e –nodeps” is used, yum doesn’t know that the package was removed. On the removed package can be depending some other package (pkg1 depends on removed pkg). So then during processing the yum operations (_perform_operations), specifically update operation (reinstall is fine), the dependency (which was satisfied before conversion and then removed by us manually using rpm -e –nodeps) is not checked and results in missing dependency.

This problem with not checking the dependency is solved in later versions, but in the one present in our converted systems. So we need to swap the packages mainly on CentOS 7 (and similar) systems.

The solution uses “yum swap” command, which is based on reinstall and install the package using yum api. The “yum swap” is mostly reimplemented on our side, using the api calls (_swap_problematic_packages).

We can use the swap command when we are removing a package from the original system vendor (like centos-logos) and this package has its replacement from rhel (in this case redhat-logos). This requires the mapping in the configuration file.

Jira Issues: RHELC-1130

Checklist

• PR has been tested manually in a VM (either author or reviewer)
• Jira issue has been made public if possible
• [RHELC-] is part of the PR title
• GitHub label has been added to help with Release notes
• PR title explains the change from the user's point of view
• Code and tests are documented properly
• The commits are squashed to as few commits as possible (without losing data)
• When merged: Jira issue has been updated to Release Pending if relevant

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Comparison is base (790db4b) 94.89% compared to head (a295f98) 94.97%.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1035      +/-   ##
==========================================
+ Coverage   94.89%   94.97%   +0.08%     
==========================================
  Files          50       50              
  Lines        4565     4601      +36     
  Branches      812      819       +7     
==========================================
+ Hits         4332     4370      +38     
+ Misses        156      155       -1     
+ Partials       77       76       -1     

Flag	Coverage Δ
centos-linux-7	90.04% <77.77%> (-0.06%)	⬇️
centos-linux-8	91.03% <77.77%> (-0.07%)	⬇️
centos-linux-9	91.08% <77.77%> (-0.07%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[r0x0d]

Just a reminder that we can get rid of this later.

We don't need the info log for update/reinstall/downgrade.

[hosekadam]

I will remove it from this PR. But I'm wondering if this cannot be used for better debugging of the yum transaction. If yes, then it could be part of https://issues.redhat.com/browse/RHELC-999 What do you think? @r0x0d

[r0x0d]

So, I guess this is not related. The log message we used here to debug is mainly for us to identify which package is being added to the transaction and the operation related to the package.

What the ticket https://issues.redhat.com/browse/RHELC-999 asks for is to capture error messages and have a better handling for that. We could potentially solve it by improving our callback classes and try to capture and output better messages there.

Also, the errors about dependency solving only happen when we reach the part where try to resolve the dependencies.

---

Said that, there might be one benefit to log those packages as a DEBUG. That would help us to understand the operation tied to the package, potentially saving us some time in the analysis when a report is opened.

But I don't think we need to leave this for now, again, improving our callback classes might make more sense than placing those logs here.

[has-bot]

/packit test --labels tier0

---

@oamg/conversions-qe please review results and provide ack.

Log | Bot Usage

[hosekadam]

Finished the part with YUM swap (tests, comments). I have a few questions:

• Do you agree with | symbol in config file for mapping the packages?
• Do we want to implement the swap for other system than CentOS 7 (and similar, like Oracle 7)?
  The bug was occurring only on system with YUM package manager, on DNF wasn't any problem. My suggestion is to add to the config file note, that the swap_pkgs variable is used only on systems with YUM. And don't implement the swap for dnf.

What do you think? @r0x0d @bookwar @bocekm ? I added a short summary to the description of this PR.

I also added mapping for the packages which I found has equivalent of them in rhel repositories. Mostly, I finished up with packages centos-* mapped to redhat-* for now (so added just centos-bookmarks from the previous state). I have one more candidate libreport-centos, but I'm not sure about this one if it couldn't cause a problem. I need to test them more (all of them). So the mapping is still in progress.

[r0x0d]

I liked the extra safety here! Pretty clever!

[r0x0d]

Finished the part with YUM swap (tests, comments). I have a few questions:

* Do you agree with `|` symbol in config file for mapping the packages?

* Do we want to implement the swap for other system than CentOS 7 (and similar, like Oracle 7)?
  The bug was occurring only on system with YUM package manager, on DNF wasn't any problem. My suggestion is to add to the config file note, that the `swap_pkgs` variable is used only on systems with YUM. And don't implement the swap for dnf.

What do you think? @r0x0d @bookwar @bocekm ? I added a short summary to the description of this PR.

I also added mapping for the packages which I found has equivalent of them in rhel repositories. Mostly, I finished up with packages centos-* mapped to redhat-* for now (so added just centos-bookmarks from the previous state). I have one more candidate libreport-centos, but I'm not sure about this one if it couldn't cause a problem. I need to test them more (all of them). So the mapping is still in progress.

---

Do you agree with | symbol in config file for mapping the packages?

I don't mind having the |.

Do we want to implement the swap for other system than CentOS 7 (and similar, like Oracle 7)?

I do think that we have to propagate the configuration file changes to the other systems to keep it consistent across the variants.

Without that, the user might not know that this option is supported by us. I'm not quite sure if we need to implement the solution to DNF as well, although, it would be nice to have, no one asked for that yet and it seems that, at least in DNF, the tool deals with it more reliably.

But I do see that we have a bunch of packages that could benefit from this swap and be removed from the exclude list. For example, maybe this is a good candidate: https://github.com/hosekadam/convert2rhel/blob/pr/hosekadam/1035/convert2rhel/data/8/x86_64/configs/oracle-8-x86_64.cfg#L12-L13.

Ideally, we want to maintain consistency between the two versions.

[hosekadam]

The all swap pkgs lists for each system were tested on the latest version of it. I've tested by installing the packages to swap prior to the conversion, then run conversion and then checked if the target package is present.

There are the results with links to the logs. The centos7 was tested twice, first attempt for the centos* pkgs and second for the libreport.

• centos7_all_without_libreport - OK http://artifacts.osci.redhat.com/testing-farm/1c32b178-5112-4749-85bf-45cf57cf65d0/work-packages_in_system_and_config_centos7_all_without_libreportvw39w5ne/log.txt
• centos7_libreport - libreport-plugin-mantisbt removed since rhel libreport doesn't need it, OK http://artifacts.osci.redhat.com/testing-farm/1c32b178-5112-4749-85bf-45cf57cf65d0/work-packages_in_system_and_config_centos7_libreport92n17s89/log.txt
• oracle7_all - OK http://artifacts.osci.redhat.com/testing-farm/10aafd54-0587-4362-a990-1ac6c0d9ced7/work-packages_in_system_and_config_oracle7_allataibbkn/log.txt
• centos8_all - OK http://artifacts.osci.redhat.com/testing-farm/1f5d08a6-7feb-46a6-9c00-5d2b8372de0d/work-packages_in_system_and_config_centos8_allf2ltqbdw/log.txt
• oracle8_all - OK http://artifacts.osci.redhat.com/testing-farm/c84992e1-a021-457b-836a-e69b5966769f/work-packages_in_system_and_config_oracle8_allbwk1g7ry/log.txt
• alma8_all - OK http://artifacts.osci.redhat.com/testing-farm/63ff1ff9-8fa1-42f7-8cb6-91155f402283/work-packages_in_system_and_config_alma8_allggf28yse/log.txt
• rocky8_all - OK http://artifacts.osci.redhat.com/testing-farm/8caa6f2b-39dd-43e6-ae00-f461baf445a6/work-packages_in_system_and_config_rocky8_allsh_ghn_p/log.txt

So from my point of view, this is finally ready (after removing the comment from the centos7 config file, I will do it in a moment)!

[r0x0d]

Looks good to me. But an extra pair of eyes might be required before approving.

Nice work!

[Venefilyn]

Minor comments. Though one thing I'd like changed is the title. Try something that addresses the implementation and then how it fixes the bug in the description. Currently the title reads like a bug report

Implementation looks great!

[Venefilyn]

Needs a rebase

[hosekadam]

Rebased and applied the review

[hosekadam]

Rebased to have changes from #1095

[r0x0d]

Few comments, but in general, it looks pretty good!

[r0x0d]

/packit test --labels tier0