prunerepo and DNF5

[xsuchy]

Prunerepo still uses DNF3
https://pagure.io/prunerepo/blob/6cafec4a70f93aa87778c3dc3b96d3de7cb0de02/f/prunerepo/helpers.py#_177-179

We done that because of
https://pagure.io/prunerepo/issue/18

DNF3 is slower than DNF5, we should revisit this.

[praiskup]

$ dnf repoquery --repofrompath=prunerepo_query,pwd --disablerepo='*' --enablerepo prunerepo_query --refresh --quiet --qf '%{location}\n'

dnf5-5.2.8.1-3.fc41.x86_64

Seems like %location / --location is still giving us an empty output.

[praiskup]

There's a work-around for rpm-software-management/dnf5#2048, it no longer blocks us.

But we need to update the pair_srpm_rpm.py to no longer use import dnf, but import libdnf5. Not sure, yet.

[praiskup]

https://pagure.io/prunerepo/pull-request/22

[xsuchy]

Needs to be build and deployed to production.

[praiskup]

Released: https://bodhi.fedoraproject.org/updates/?search=prunerepo

copr-backend still builds fine against the new package:
https://download.copr.fedorainfracloud.org/results/@copr/copr-pull-requests/fedora-rawhide-x86_64/08666885-copr-backend/builder-live.log.gz

[praiskup]

After a prior team ACK, deployed to prod. The repo pruner is running now. Going to close tomorrow.

[praiskup]

Hmm, the script I run manually ended after ~2hours, that is a bit too fast? Monitoring another run.

[praiskup]

[2025-02-16 03:46:01,481][  INFO][PID:1473868][copr_prune_results][frontend.py:_send_attempt:70] Sending GET request to frontend URL - https://copr.fedorainfracloud.org/backend/chroots-prunerepo-status/
[2025-02-17 03:25:01,234][  INFO][PID:2393753][copr_prune_results][frontend.py:_send_attempt:70] Sending GET request to frontend URL - https://copr.fedorainfracloud.org/backend/chroots-prunerepo-status/
[2025-02-18 03:23:02,317][  INFO][PID:1827606][copr_prune_results][frontend.py:_send_attempt:70] Sending GET request to frontend URL - https://copr.fedorainfracloud.org/backend/chroots-prunerepo-status/
[2025-02-18 11:03:44,800][  INFO][PID:748139][copr_prune_results][frontend.py:_send_attempt:70] Sending GET request to frontend URL - https://copr.fedorainfracloud.org/backend/chroots-prunerepo-status/
[2025-02-19 03:39:02,195][  INFO][PID:3097089][copr_prune_results][frontend.py:_send_attempt:70] Sending GET request to frontend URL - https://copr.fedorainfracloud.org/backend/chroots-prunerepo-status/
[2025-02-20 03:35:02,552][  INFO][PID:2864581][copr_prune_results][frontend.py:_send_attempt:70] Sending GET request to frontend URL - https://copr.fedorainfracloud.org/backend/chroots-prunerepo-status/
[root@copr-be ~][PROD]# grep PID:1473868 /var/log/copr-backend/pruner.log | tail -n1
[2025-02-16 12:30:15,501][  INFO][PID:1473868][copr_prune_results][copr_prune_results.py:run:191] --------------------------------------------
[root@copr-be ~][PROD]# grep PID:2393753 /var/log/copr-backend/pruner.log | tail -n1
[2025-02-17 12:54:27,259][  INFO][PID:2393753][copr_prune_results][copr_prune_results.py:run:191] --------------------------------------------
[root@copr-be ~][PROD]# grep PID:1827606 /var/log/copr-backend/pruner.log | tail -n1
[2025-02-18 03:49:17,136][  INFO][PID:1827606][copr_prune_results][copr_prune_results.py:run:172] Pruning tasks are delegated to background workers, waiting.
[root@copr-be ~][PROD]# grep PID:748139 /var/log/copr-backend/pruner.log | tail -n1
[2025-02-18 14:51:12,642][  INFO][PID:748139][copr_prune_results][copr_prune_results.py:run:191] --------------------------------------------
[root@copr-be ~][PROD]# grep PID:3097089 /var/log/copr-backend/pruner.log | tail -n1
[2025-02-19 07:05:09,684][  INFO][PID:3097089][copr_prune_results][copr_prune_results.py:run:191] --------------------------------------------
[root@copr-be ~][PROD]# grep PID:2864581 /var/log/copr-backend/pruner.log | tail -n1
[2025-02-20 07:51:37,868][  INFO][PID:2864581][copr_prune_results][copr_prune_results.py:run:191] --------------------------------------------

These data show the last prunerepo starts, and when particular PIDs finished. The first 2025-02-18 was canceled by me, and the following one was running the updated prunerepo (but benefited from kernel filesystem caches of the canceled run). Anyway, we are down from ~9h to ~4.5h. Nice!

Sorry for the edit. Firefox/GitHub ate my well-edited comment, and submitted previous version.