Batch feed ingestion, add a readiness check, fail open without Redis (C2, C8)

Ingestion did a get_or_create per entry. It now prefetches existing
articles for the poll in one query, bulk-inserts new ones and
bulk-updates changed ones (falling back to row-by-row only if a
concurrent fetch wins a race), and dispatches rule evaluation in
bounded chunks so a huge first fetch is several small tasks.

/health/ stays a DB-only liveness check (upgrade.sh gates on it right
after restart). New /health/ready/ also checks Redis and that
poll_due_feeds ticked within five minutes — proving beat and a worker
are both alive — and returns 503 with per-check detail otherwise.

The sidebar unread-count cache and login throttle now degrade instead
of erroring when Redis is unreachable: every authenticated page and
every login would otherwise 500 during a cache outage. Found while
verifying in a sandbox with no Redis.

Also includes the dashboard's oldest-article metric (core/views.py).

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
This commit is contained in:
ksmithandClaude Sonnet 5.5 committed 2026-11-01 08:51:20 -07:00
1 parent 8d2900424a
commit e5f1277014
14 files changed
+342 -36

No files matched your search

+1
View File
@@ -45,4 +45,5 @@ def _setup_exempt_paths():
reverse("pwa:service_worker"),
reverse("pwa:offline"),
reverse("core:health"),
reverse("core:health_ready"),
}
+18
View File
@@ -72,3 +72,21 @@ def test_too_many_failed_attempts_against_one_username_blocks_that_username(clie
response = client.post(reverse("accounts:login"), {"username": "bob", "password": "pw-bob-123"})
assert response.status_code == 302
assert response.wsgi_request.user == bob
@pytest.mark.django_db
def test_login_still_works_when_the_throttle_cache_is_down(client, monkeypatch):
from unittest.mock import Mock
from django.contrib.auth import get_user_model
get_user_model().objects.create_user("alice", password="pw-alice-123")
broken = Mock(side_effect=ConnectionError("redis down"))
monkeypatch.setattr("apps.accounts.views.cache.get", broken)
monkeypatch.setattr("apps.accounts.views.cache.incr", broken)
monkeypatch.setattr("apps.accounts.views.cache.set", broken)
bad = client.post(reverse("accounts:login"), {"username": "alice", "password": "wrong"})
assert bad.status_code == 200
good = client.post(reverse("accounts:login"), {"username": "alice", "password": "pw-alice-123"})
assert good.status_code == 302
+14
View File
@@ -1,3 +1,5 @@
import logging
from django.contrib import messages
from django.contrib.auth import login
from django.contrib.auth.decorators import user_passes_test
@@ -10,6 +12,8 @@ from .forms import CreateUserForm, SetupForm
from .middleware import SetupWizardMiddleware
from .models import SetupLock, User
logger = logging.getLogger(__name__)
is_admin = user_passes_test(lambda u: u.is_authenticated and u.is_staff)
# Two independent throttles, either of which can block an attempt: by
@@ -41,14 +45,24 @@ def _attempts_key(kind: str, value: str) -> str:
def _too_many_attempts(key: str, limit: int) -> bool:
# Fails open if the cache (Redis) is unreachable: logging in must keep
# working through a cache outage, at the cost of no throttling until
# it's back — an outage shouldn't lock every user out of the app.
try:
return cache.get(key, 0) >= limit
except Exception:
logger.warning("Login throttle cache unavailable; not rate limiting", exc_info=True)
return False
def _record_failed_attempt(key: str) -> None:
try:
try:
cache.incr(key)
except ValueError:
cache.set(key, 1, LOGIN_RATE_LIMIT_WINDOW_SECONDS)
except Exception:
logger.warning("Login throttle cache unavailable; attempt not recorded", exc_info=True)
class LoginView(DjangoLoginView):
+14
View File
@@ -1,3 +1,5 @@
import logging
from django.core.cache import cache
from django.db.models import Count
@@ -5,6 +7,8 @@ from apps.core.cache_utils import SIDEBAR_UNREAD_CACHE_TTL_SECONDS, sidebar_unre
from .models import SAVED_TAG_NAME, STARRED_TAG_NAME, Article, ArticleTag, UserArticleState
logger = logging.getLogger(__name__)
def with_user_state(articles, user):
"""Annotates an iterable of Article instances with is_read/is_starred/
@@ -56,7 +60,14 @@ def unread_counts_by_feed(user):
the current user's own read state, so this is rarely visibly stale.
"""
cache_key = sidebar_unread_cache_key(user.pk)
# The cache is an optimization here, never a dependency: every
# authenticated page render goes through this, so a Redis outage must
# degrade to recomputing the count, not to a 500 on every page.
try:
counts = cache.get(cache_key)
except Exception:
logger.warning("Sidebar cache unavailable; computing unread counts directly")
counts = None
if counts is not None:
return counts
@@ -68,5 +79,8 @@ def unread_counts_by_feed(user):
.annotate(unread=Count("id"))
)
counts = {row["feed_id"]: row["unread"] for row in rows}
try:
cache.set(cache_key, counts, SIDEBAR_UNREAD_CACHE_TTL_SECONDS)
except Exception:
logger.warning("Sidebar cache unavailable; skipping cache write")
return counts
+16
View File
@@ -58,3 +58,19 @@ def test_unread_counts_by_feed_cache_is_per_user(alice, feed):
assert unread_counts_by_feed(alice) == {feed.pk: 1}
assert unread_counts_by_feed(bob) == {feed.pk: 1}
@pytest.mark.django_db
def test_unread_counts_still_work_when_the_cache_is_down(alice, feed, monkeypatch):
"""A Redis outage must degrade the sidebar to recomputing counts, not
500 every authenticated page."""
from unittest.mock import Mock
Subscription.objects.create(user=alice, feed=feed)
Article.objects.create(feed=feed, guid="a-1")
broken = Mock(side_effect=ConnectionError("redis down"))
monkeypatch.setattr("apps.articles.queries.cache.get", broken)
monkeypatch.setattr("apps.articles.queries.cache.set", broken)
assert unread_counts_by_feed(alice) == {feed.pk: 1}
invalidate_sidebar_unread_cache(alice.pk)
+7
View File
@@ -1,5 +1,9 @@
import logging
from django.core.cache import cache
logger = logging.getLogger(__name__)
# Short enough that a missed invalidation point self-heals within one
# page load's worth of staleness; long enough to actually absorb repeat
# authenticated-page renders (apps.core.context_processors.sidebar runs
@@ -19,4 +23,7 @@ def invalidate_sidebar_unread_cache(user_id: int) -> None:
out the TTL above. Rule-driven state changes during background feed
ingestion aren't invalidated this way — those self-heal within the
TTL, same as a missed call site here would."""
try:
cache.delete(sidebar_unread_cache_key(user_id))
except Exception:
logger.warning("Sidebar cache unavailable; nothing to invalidate")
+42
View File
@@ -22,6 +22,48 @@ def test_health_check_works_before_first_run_setup(client):
assert response.status_code == 200
@pytest.mark.django_db
def test_health_ready_is_degraded_until_a_poll_tick_has_happened(client):
response = client.get(reverse("core:health_ready"))
assert response.status_code == 503
body = response.json()
assert body["status"] == "degraded"
assert body["checks"]["database"] is True
assert body["checks"]["redis"] is True
assert body["checks"]["poll_heartbeat"] is False
@pytest.mark.django_db
def test_health_ready_is_ok_after_poll_due_feeds_runs(client):
from apps.feeds.tasks import poll_due_feeds
poll_due_feeds()
response = client.get(reverse("core:health_ready"))
assert response.status_code == 200
assert response.json()["checks"] == {
"database": True,
"redis": True,
"poll_heartbeat": True,
}
@pytest.mark.django_db
def test_health_ready_flags_a_stale_poll_heartbeat(client):
from datetime import timedelta
from django.core.cache import cache
from django.utils import timezone
from apps.feeds.tasks import POLL_HEARTBEAT_CACHE_KEY
cache.set(POLL_HEARTBEAT_CACHE_KEY, timezone.now() - timedelta(minutes=30))
response = client.get(reverse("core:health_ready"))
assert response.status_code == 503
assert response.json()["checks"]["poll_heartbeat"] is False
@pytest.mark.django_db
def test_dashboard_requires_staff(client, django_user_model):
django_user_model.objects.create_user("regular", password="pw-regular-123")
+1
View File
@@ -6,6 +6,7 @@ app_name = "core"
urlpatterns = [
path("health/", views.health, name="health"),
path("health/ready/", views.health_ready, name="health_ready"),
path("dashboard/", views.dashboard, name="dashboard"),
path("dashboard/feeds/<int:feed_id>/retry/", views.retry_feed, name="retry_feed"),
]
+50 -3
View File
@@ -1,20 +1,27 @@
from django.contrib import messages
from django.contrib.auth.decorators import user_passes_test
from django.core.cache import cache
from django.db import connection
from django.db.models import Count, Q
from django.db.models import Count, Min, Q
from django.http import JsonResponse
from django.shortcuts import get_object_or_404, redirect, render
from django.utils import timezone
from django.views.decorators.http import require_POST
from apps.articles.models import Article
from apps.feeds.models import Feed, Subscription
from apps.feeds.tasks import fetch_feed, reset_feed_backoff
from apps.feeds.tasks import POLL_HEARTBEAT_CACHE_KEY, fetch_feed, reset_feed_backoff
is_admin = user_passes_test(lambda u: u.is_authenticated and u.is_staff)
# Beat ticks poll_due_feeds every minute; this many minutes without a tick
# means beat or the worker (or the broker between them) isn't working.
POLL_HEARTBEAT_MAX_AGE_SECONDS = 5 * 60
def health(request):
"""Unauthenticated liveness/readiness check — reachable even before
"""Unauthenticated liveness check — reachable even before
first-run setup (see SetupWizardMiddleware). Used by scripts/upgrade.sh
to confirm a freshly restarted deploy actually came up, and is a
reasonable target for any external uptime monitor too."""
@@ -26,6 +33,45 @@ def health(request):
return JsonResponse({"status": "ok"})
def health_ready(request):
"""Readiness: liveness plus everything background fetching depends on
(C8) — the database, the Redis cache/broker, and a recent
poll_due_feeds tick, which proves beat is scheduling AND a worker is
consuming. /health/ above stays DB-only on purpose: scripts/upgrade.sh
gates a deploy on it right after restart, before beat has had a
chance to tick. Point an uptime monitor at this one instead."""
checks = {"database": False, "redis": False, "poll_heartbeat": False}
detail = {}
try:
with connection.cursor() as cursor:
cursor.execute("SELECT 1")
checks["database"] = True
except Exception:
pass
try:
cache.set("health-ready-probe", "1", timeout=10)
checks["redis"] = cache.get("health-ready-probe") == "1"
except Exception:
pass
try:
last_tick = cache.get(POLL_HEARTBEAT_CACHE_KEY)
except Exception:
last_tick = None
if last_tick is not None:
age = (timezone.now() - last_tick).total_seconds()
detail["poll_heartbeat_age_seconds"] = int(age)
checks["poll_heartbeat"] = age <= POLL_HEARTBEAT_MAX_AGE_SECONDS
ok = all(checks.values())
return JsonResponse(
{"status": "ok" if ok else "degraded", "checks": checks, **detail},
status=200 if ok else 503,
)
@is_admin
def dashboard(request):
feeds = Feed.objects.annotate(subscriber_count=Count("subscriptions"))
@@ -39,6 +85,7 @@ def dashboard(request):
"error_feed_count": erroring_feeds.count(),
"subscription_count": Subscription.objects.count(),
"article_count": Article.objects.count(),
"oldest_article_at": Article.objects.aggregate(oldest=Min("fetched_at"))["oldest"],
}
return render(
+69 -31
View File
@@ -18,6 +18,7 @@ import feedparser
import nh3
import requests
from django.conf import settings
from django.db import IntegrityError, transaction
from django.utils import timezone
from apps.articles.models import Article
@@ -131,6 +132,26 @@ def _entry_content_html(entry) -> str:
return getattr(entry, "summary", "") or ""
UPDATABLE_ARTICLE_FIELDS = (
"title",
"link",
"author",
"summary",
"content_html",
"published_at",
"content_hash",
)
def _ensure_pks(feed: Feed, articles: list[Article]) -> list[int]:
"""bulk_create only populates primary keys on backends that can
RETURN them (Postgres, SQLite 3.35+); look the rest up by guid."""
if all(a.pk for a in articles):
return [a.pk for a in articles]
guids = [a.guid for a in articles]
return list(Article.objects.filter(feed=feed, guid__in=guids).values_list("pk", flat=True))
def effective_interval_minutes(feed: Feed, configured_interval_minutes: int) -> int:
"""The actual polling interval to use for `feed`, applying exponential
backoff on top of the site-wide `configured_interval_minutes` (see
@@ -217,52 +238,69 @@ def fetch_and_ingest(feed: Feed) -> FetchResult:
)
entries = entries[:MAX_ENTRIES_PER_POLL]
# One prefetch + bulk writes for the whole poll (C2), instead of a
# get_or_create (SELECT, then INSERT) per entry. Entries are keyed by
# guid so a feed that repeats one within a single response is only
# processed once.
parsed_entries: dict[str, dict] = {}
for entry in entries:
guid = _entry_guid(entry)
if guid in parsed_entries:
continue
title = (getattr(entry, "title", "") or "")[:1000]
link = (getattr(entry, "link", "") or "")[:2000]
author = (getattr(entry, "author", "") or "")[:500]
summary = (getattr(entry, "summary", "") or "")[:5000]
content_html = sanitize_html(_entry_content_html(entry))
published_at = _entry_published_at(entry)
content_hash = _content_hash(title, link, author, summary, content_html, published_at)
article, created = Article.objects.get_or_create(
feed=feed,
guid=guid,
defaults={
"link": link,
parsed_entries[guid] = {
"title": title,
"link": link,
"author": author,
"content_html": content_html,
"summary": summary,
"content_hash": content_hash,
"content_html": content_html,
"published_at": published_at,
},
)
"content_hash": _content_hash(title, link, author, summary, content_html, published_at),
}
if created:
new_ids.append(article.pk)
elif article.content_hash != content_hash:
article.title = title
article.link = link
article.author = author
article.summary = summary
article.content_html = content_html
article.published_at = published_at
article.content_hash = content_hash
article.save(
update_fields=[
"title",
"link",
"author",
"summary",
"content_html",
"published_at",
"content_hash",
existing = {a.guid: a for a in Article.objects.filter(feed=feed, guid__in=parsed_entries)}
to_create = [
Article(feed=feed, guid=guid, **fields)
for guid, fields in parsed_entries.items()
if guid not in existing
]
to_update = []
for guid, article in existing.items():
fields = parsed_entries[guid]
if article.content_hash != fields["content_hash"]:
for name, value in fields.items():
setattr(article, name, value)
to_update.append(article)
if to_update:
Article.objects.bulk_update(to_update, list(UPDATABLE_ARTICLE_FIELDS), batch_size=100)
updated_ids.extend(a.pk for a in to_update)
if to_create:
try:
with transaction.atomic():
Article.objects.bulk_create(to_create)
except IntegrityError:
# Another worker inserted some of these guids between the
# prefetch above and now (two fetches of the same feed can
# overlap) — fall back to row-by-row so nothing is lost or
# duplicated, at the cost of the old per-entry round trips
# for this rare case only.
for article in to_create:
fields = {name: getattr(article, name) for name in UPDATABLE_ARTICLE_FIELDS}
obj, created = Article.objects.get_or_create(
feed=feed, guid=article.guid, defaults=fields
)
updated_ids.append(article.pk)
if created:
new_ids.append(obj.pk)
else:
new_ids.extend(_ensure_pks(feed, to_create))
return FetchResult(
status=Feed.STATUS_OK, new_article_ids=new_ids, updated_article_ids=updated_ids
+15 -1
View File
@@ -4,6 +4,7 @@ from io import BytesIO
from celery import shared_task
from django.contrib.auth import get_user_model
from django.core.cache import cache
from django.db.models import Q
from django.utils import timezone
@@ -27,6 +28,12 @@ CLAIM_STALE_MINUTES = 10
# across several beat ticks instead of enqueuing everything — and every
# bounded Feed query below stays index-only work, not a full scan (C1).
POLL_BATCH_SIZE = 500
RULE_DISPATCH_CHUNK_SIZE = 50
# Written on every poll_due_feeds tick; apps.core.views.health_ready
# reads it to tell whether beat is scheduling and a worker is consuming.
POLL_HEARTBEAT_CACHE_KEY = "poll-due-feeds-heartbeat"
POLL_HEARTBEAT_TTL_SECONDS = 3600
@shared_task
@@ -39,6 +46,7 @@ def poll_due_feeds():
here, not via one Celery Beat entry per feed.
"""
now = timezone.now()
cache.set(POLL_HEARTBEAT_CACHE_KEY, now, timeout=POLL_HEARTBEAT_TTL_SECONDS)
claim_stale_before = now - timedelta(minutes=CLAIM_STALE_MINUTES)
due_feed_ids = list(
Feed.objects.filter(is_active=True)
@@ -125,7 +133,13 @@ def fetch_feed(feed_id: int):
len(result.new_article_ids),
len(result.updated_article_ids),
)
apply_filter_rules_for_feed.delay(feed_id, changed_article_ids)
# Chunked so one huge poll (a first fetch of a long archive feed)
# becomes several bounded rule-evaluation tasks, each well inside
# the worker's time limits, instead of one unbounded one (C2).
for start in range(0, len(changed_article_ids), RULE_DISPATCH_CHUNK_SIZE):
apply_filter_rules_for_feed.delay(
feed_id, changed_article_ids[start : start + RULE_DISPATCH_CHUNK_SIZE]
)
return {
"status": result.status,
+74
View File
@@ -264,3 +264,77 @@ def test_fetch_only_processes_the_first_max_entries_per_poll(feed):
assert len(result.new_article_ids) == MAX_ENTRIES_PER_POLL
assert Article.objects.filter(feed=feed).count() == MAX_ENTRIES_PER_POLL
assert not Article.objects.filter(feed=feed, guid=f"guid-{MAX_ENTRIES_PER_POLL}").exists()
def _feed_xml(count, title_suffix=""):
items = "".join(
f"<item><title>Item {i}{title_suffix}</title><guid>guid-{i}</guid></item>"
for i in range(count)
)
return (
f'<?xml version="1.0"?><rss version="2.0"><channel><title>T</title>{items}</channel></rss>'
)
@pytest.mark.django_db
@responses.activate
def test_ingest_uses_a_constant_number_of_article_queries(feed):
"""C2: one prefetch + bulk writes per poll, not a get_or_create per
entry — query count must not grow with the number of entries."""
from django.db import connection
from django.test.utils import CaptureQueriesContext
responses.add(responses.GET, FEED_URL, body=_feed_xml(40), status=200)
with CaptureQueriesContext(connection) as ctx:
result = fetch_and_ingest(feed)
assert len(result.new_article_ids) == 40
assert len(ctx.captured_queries) < 10
assert Article.objects.filter(feed=feed).count() == 40
@pytest.mark.django_db
@responses.activate
def test_ingest_dedups_a_guid_repeated_within_one_response(feed):
xml = (
'<?xml version="1.0"?><rss version="2.0"><channel><title>T</title>'
"<item><title>A</title><guid>same</guid></item>"
"<item><title>B</title><guid>same</guid></item>"
"</channel></rss>"
)
responses.add(responses.GET, FEED_URL, body=xml, status=200)
result = fetch_and_ingest(feed)
assert len(result.new_article_ids) == 1
assert Article.objects.filter(feed=feed).count() == 1
@pytest.mark.django_db
@responses.activate
def test_ingest_reports_updates_separately_from_new_articles(feed):
responses.add(responses.GET, FEED_URL, body=_feed_xml(3), status=200)
fetch_and_ingest(feed)
responses.replace(responses.GET, FEED_URL, body=_feed_xml(4, " edited"), status=200)
result = fetch_and_ingest(feed)
assert len(result.new_article_ids) == 1
assert len(result.updated_article_ids) == 3
assert Article.objects.get(feed=feed, guid="guid-0").title == "Item 0 edited"
@pytest.mark.django_db
@responses.activate
def test_ingest_falls_back_to_row_by_row_when_a_concurrent_insert_wins(feed):
from unittest.mock import patch
from django.db import IntegrityError
responses.add(responses.GET, FEED_URL, body=_feed_xml(3), status=200)
with patch.object(Article.objects, "bulk_create", side_effect=IntegrityError("race")):
result = fetch_and_ingest(feed)
assert len(result.new_article_ids) == 3
assert Article.objects.filter(feed=feed).count() == 3
+20
View File
@@ -208,3 +208,23 @@ def test_fetch_feed_schedules_the_next_fetch_using_the_site_wide_interval():
assert feed.next_fetch_at is not None
expected = before + timedelta(minutes=120)
assert abs((feed.next_fetch_at - expected).total_seconds()) < 5
@pytest.mark.django_db
@responses.activate
def test_fetch_feed_dispatches_rule_evaluation_in_bounded_chunks(monkeypatch):
monkeypatch.setattr("apps.feeds.tasks.RULE_DISPATCH_CHUNK_SIZE", 2)
feed = Feed.objects.create(feed_url="https://example.com/feed.xml")
items = "".join(f"<item><title>T{i}</title><guid>g{i}</guid></item>" for i in range(5))
responses.add(
responses.GET,
feed.feed_url,
body=f'<?xml version="1.0"?><rss version="2.0"><channel>{items}</channel></rss>',
status=200,
)
with patch("apps.feeds.tasks.apply_filter_rules_for_feed.delay") as mock_delay:
fetch_feed(feed.id)
chunk_sizes = [len(call.args[1]) for call in mock_delay.call_args_list]
assert chunk_sizes == [2, 2, 1]
+1 -1
View File
@@ -36,7 +36,7 @@ SECURE_PROXY_SSL_HEADER = ("HTTP_X_FORWARDED_PROTO", "https")
# over loopback (see ALLOWED_HOSTS above) it has no X-Forwarded-Proto
# header to prove it's "secure", and there's no TLS listener on gunicorn's
# own port to redirect to anyway.
SECURE_REDIRECT_EXEMPT = [r"^health/$"]
SECURE_REDIRECT_EXEMPT = [r"^health/$", r"^health/ready/$"]
# WhiteNoise serves /static/ directly from gunicorn instead of relying on
# the reverse proxy to alias the staticfiles directory on disk — the proxy