Support on-demand content in repair_metadata

closes #849

Author: jobselko

CHANGES/849.feature

@@ -0,0 +1 @@
+Added support for on-demand content to `repair_metadata` endpoint.

pulp_python/app/tasks/repair.py

@@ -1,12 +1,19 @@
 import logging
+from collections import defaultdict
 from gettext import gettext as _
+from itertools import groupby
 from uuid import UUID
 
+from django.db.models import Prefetch
 from django.db.models.query import QuerySet
 from pulp_python.app.models import PythonPackageContent, PythonRepository
-from pulp_python.app.utils import artifact_to_python_content_data
-from pulpcore.plugin.models import ProgressReport
-from pulpcore.plugin.util import get_domain
+from pulp_python.app.utils import (
+    artifact_to_python_content_data,
+    fetch_json_release_metadata,
+    parse_metadata,
+)
+from pulpcore.plugin.models import ContentArtifact, ProgressReport
+from pulpcore.plugin.util import get_domain, get_prn
 
 log = logging.getLogger(__name__)
 
@@ -34,11 +41,16 @@ def repair(repository_pk: UUID) -> None:
     content_set = repository.latest_version().content.values_list("pk", flat=True)
     content = PythonPackageContent.objects.filter(pk__in=content_set)
 
-    num_repaired = repair_metadata(content)
-    log.info(_("{} packages' metadata repaired.").format(num_repaired))
+    num_repaired, pkgs_not_repaired = repair_metadata(content)
+    log.info(
+        _(
+            "{} packages' metadata repaired. Not repaired packages due to either "
+            "inaccessible URL or mismatched sha256: {}."
+        ).format(num_repaired, pkgs_not_repaired)
+    )
 
 
-def repair_metadata(content: QuerySet[PythonPackageContent]) -> int:
+def repair_metadata(content: QuerySet[PythonPackageContent]) -> tuple[int, set[str]]:
     """
     Repairs metadata for a queryset of PythonPackageContent objects
     and updates the progress report.
@@ -47,24 +59,38 @@ def repair_metadata(content: QuerySet[PythonPackageContent]) -> int:
         content (QuerySet[PythonPackageContent]): The queryset of items to repair.
 
     Returns:
-        int: The number of packages that were repaired.
+        tuple[int, set[str]]: A tuple containing:
+            - The number of packages that were repaired.
+            - A set of packages' PRNs that were not repaired.
     """
-    # TODO: Add on_demand content repair
     immediate_content = (
         content.filter(contentartifact__artifact__isnull=False)
         .distinct()
         .prefetch_related("_artifacts")
     )
+    on_demand_content = (
+        content.filter(contentartifact__artifact__isnull=True)
+        .distinct()
+        .prefetch_related(
+            Prefetch(
+                "contentartifact_set",
+                queryset=ContentArtifact.objects.prefetch_related("remoteartifact_set"),
+            )
+        )
+        .order_by("name", "version")
+    )
     domain = get_domain()
 
     batch = []
     set_of_update_fields = set()
     total_repaired = 0
+    # Keep track of on-demand packages that were not repaired
+    pkgs_not_repaired = set()
 
     progress_report = ProgressReport(
         message="Repairing packages' metadata",
         code="repair.metadata",
-        total=immediate_content.count(),
+        total=content.count(),
     )
     progress_report.save()
     with progress_report:
@@ -78,11 +104,60 @@ def repair_metadata(content: QuerySet[PythonPackageContent]) -> int:
                 package, new_data, batch, set_of_update_fields
             )
 
+        # For on-demand content, we expect that:
+        # 1. PythonPackageContent always has correct name and version
+        # 2. RemoteArtifact always has correct sha256
+        for (name, version), group in groupby(
+            on_demand_content.iterator(chunk_size=BULK_SIZE),
+            key=lambda x: (x.name, x.version),
+        ):
+            group_set = set(group)
+            grouped_by_url = defaultdict(list)
+
+            for package in group_set:
+                for ra in package.contentartifact_set.get().remoteartifact_set.all():
+                    grouped_by_url[ra.remote.url].append((package, ra))
+
+            # Prioritize the URL that can serve the most packages
+            for url, pkg_ra_pairs in sorted(
+                grouped_by_url.items(), key=lambda x: len(x[1]), reverse=True
+            ):
+                if not group_set:
+                    break  # No packages left to repair, move onto the next group
+                remotes = set([pkg_ra[1].remote for pkg_ra in pkg_ra_pairs])
+                try:
+                    json_data = fetch_json_release_metadata(name, version, remotes)
+                except Exception:
+                    continue
+
+                for package, ra in pkg_ra_pairs:
+                    if package not in group_set:
+                        continue  # Package was already repaired
+                    # Extract data only for the specific distribution being checked
+                    dist_data = None
+                    for dist in json_data["urls"]:
+                        if ra.sha256 == dist["digests"]["sha256"]:
+                            dist_data = dist
+                            break
+                    if not dist_data:
+                        continue
+
+                    new_data = parse_metadata(json_data["info"], version, dist_data)
+                    new_data.pop("url")  # url belongs to RemoteArtifact
+                    total_repaired += update_package_if_needed(
+                        package, new_data, batch, set_of_update_fields
+                    )
+                    group_set.remove(package)
+                    progress_report.increment()
+            # Store and track the unrepaired packages after all URLs are processed
+            pkgs_not_repaired.update([get_prn(p) for p in group_set])
+            progress_report.increase_by(len(group_set))
+
     if batch:
         total_repaired += len(batch)
         PythonPackageContent.objects.bulk_update(batch, set_of_update_fields)
 
-    return total_repaired
+    return total_repaired, pkgs_not_repaired
 
 
 def update_package_if_needed(

pulp_python/app/utils.py

@@ -9,6 +9,7 @@
 from packaging.utils import canonicalize_name
 from packaging.requirements import Requirement
 from packaging.version import parse, InvalidVersion
+from pulpcore.plugin.models import Remote
 
 
 PYPI_LAST_SERIAL = "X-PYPI-LAST-SERIAL"
@@ -189,6 +190,37 @@ def artifact_to_python_content_data(filename, artifact, domain=None):
     return data
 
 
+def fetch_json_release_metadata(name: str, version: str, remotes: set[Remote]) -> dict:
+    """
+    Fetches metadata for a specific release from PyPI's JSON API. A release can contain
+    multiple distributions. See https://docs.pypi.org/api/json/#get-a-release for more details.
+    All remotes should have the same URL.
+
+    Returns:
+        Dict containing "info", "last_serial", "urls", and "vulnerabilities" keys.
+    Raises:
+        Exception if fetching from all remote URLs fails.
+    """
+    remote = next(iter(remotes))
+    url = remote.get_remote_artifact_url(f"pypi/{name}/{version}/json")
+
+    result = None
+    for remote in remotes:
+        downloader = remote.get_downloader(url=url, max_retries=1)
+        try:
+            result = downloader.fetch()
+            break
+        except Exception:
+            continue
+
+    if result:
+        with open(result.path, "r") as file:
+            json_data = json.load(file)
+        return json_data
+    else:
+        raise Exception(f"Failed to fetch {url} from any remote.")
+
+
 def python_content_to_json(base_path, content_query, version=None, domain=None):
     """
     Converts a QuerySet of PythonPackageContent into the PyPi JSON format