Added /prune/ endpoint to removing "old" RPMs from a Repository.

[ggainey]

closes #2909.

[ggainey]

Draft until I figure out what is up w/ permissions - can't reproduce locally?!?

[ggainey]

Some notes on performance:

Some unscientific performance results follow. The repositories are a selection of "real" repos, of small (zoo), medium (baseos-9) and large (appstream-9, epel) amounts of content.

All runs used concurrency==1, oci-env, on my workstation (Intel® Core™ i7-6700K × 8), info-logging only.
Repositories are --no-autopublish.

Repository List:

name: zoo
size: 35
dups: 4

name: rhel_9_x86_64_appstream
size: 17524
dups: 11834

name: rhel_9_x86_64_baseos
size: 6185
dups: 4993

name: epel9
size: 21236
dups: 0

How "prune" was run

$ TGROUP=`http -b POST :5001/pulp/api/v3/rpm/prune/ repo_hrefs:='["*"]' repo_concurrency:=1 keep_days=0 dry_run=True | jq -r .task_group`

How timings were evaluated

for t in `http ":5001${TGROUP}" | jq -r .tasks[].pulp_href`
do 
  pulp task show --href "${t}" | jq '.started_at, .finished_at, .progress_reports[0].message, .progress_reports[0].total' 
done

Performance Runs

• dry-run, keep=0 (max deletes discovered only)

  • zoo : 46ms
  • epel : 850ms
  • base : 380ms
  • apps : 912ms

• prune, keep=1000 (no deletes discovered only)

  • zoo : 49ms
  • epel : 849ms
  • base : 392ms
  • apps : 929ms

• prune, keep=0 (max actual deletes)

  • zoo : 163ms
  • epel : 2214ms
  • base : 1823ms
  • apps : 4698ms

• prune, keep=0, post-max-prune (0 actual deletes)

  • zoo : 206ms
  • epel : 1884ms
  • base : 628ms
  • apps : 1365ms

[ggainey]

Drafting temporarily while I think on some suggestions.

[ipanova]

looks good!Thanks

[xsuchy]

Thank you.