feat(reports): Cache per-project weekly report metrics by amy-chen23 · Pull Request #116739 · getsentry/sentry

amy-chen23 · 2026-06-02T22:40:01Z

Goal: Cache per-project total errors and total transactions during weekly report generation so the frontend can display week-over-week percentage change without re-querying Snuba.

Redis Cache Layer

Caches total errors and total transactions by Org Id, Project Id (no timestamp)
10 day TTL

…ort is run

linear-code · 2026-06-02T22:40:05Z

ID-1589

github-actions · 2026-06-02T22:42:24Z

🚨 Warning: This pull request contains Frontend and Backend changes!

It's discouraged to make changes to Sentry's Frontend and Backend in a single pull request. The Frontend and Backend are not atomically deployed. If the changes are interdependent of each other, they must be separated into two pull requests and be made forward or backwards compatible, such that the Backend or Frontend can be safely deployed independently.

Have questions? Please ask in the #discuss-dev-infra channel.

…s (typing)

amy-chen23 · 2026-06-03T17:55:48Z

only frontend change -- auto-added by santry bot

JoshFerge · 2026-06-08T21:39:04Z

just out of curiosity, what's the napkin math on how much cache space this will use?

amy-chen23 · 2026-06-08T21:50:55Z

just out of curiosity, what's the napkin math on how much cache space this will use?

good question! there are around 140-150 bytes per entry. Not sure if the number of projects is public 👀 , but since TTL is 14 days, the amount of sets could be 2 * num_of_projects at any given time.

so cache space would be 140 * 2 * num_of_projects. It's not trivial, so the alternative is storing in a DB, but that's more overhead/setup

shashjar · 2026-06-08T21:57:41Z

we currently have ~3M projects in US, so that works out to <1 gig of cache usage

amy-chen23 · 2026-06-08T22:20:14Z

+            logger.exception(
+                "weekly_reports.cache_project_metrics.failed",
+                extra={"organization_id": organization_id, "timestamp": timestamp},
+            )


we might not need this logger?

yeah think we can remove it

shashjar · 2026-06-08T23:09:09Z

@@ -218,15 +242,17 @@ def prepare_organization_report(
            lifecycle.record_halt(WeeklyReportHaltReason.EMPTY_REPORT)
            return

+    # Cache per-project metrics for reuse across report batches
+    _cache_project_metrics(ctx, organization_id, timestamp)


up to you, but maybe this should just call a top-level function in src/sentry/tasks/summaries/weekly_report_cache.py directly

agree -- moved this logic into src/sentry/tasks/summaries/weekly_report_cache.py

shashjar · 2026-06-08T23:09:36Z

+            logger.exception(
+                "weekly_reports.cache_project_metrics.failed",
+                extra={"organization_id": organization_id, "timestamp": timestamp},
+            )


yeah think we can remove it

shashjar · 2026-06-08T23:10:49Z

+from sentry.utils import json, metrics, redis
+from sentry.utils.dates import floor_to_utc_day, to_datetime
+
+CACHE_TTL = timedelta(days=14)


can we make this 10 days to prevent overlap of cache usage across 3 weeks?

also, you can define this as an int in number of seconds, e.g. 10 * 24 * 60 * 60, so that you don't have to do that conversion below

shashjar · 2026-06-08T23:12:39Z

+
+
+def _make_cache_key(org_id: int, project_id: int, timestamp: float) -> str:
+    return f"{KEY_PREFIX}:{org_id}:{project_id}:{timestamp}"


including a timestamp in the cache key doesn't seem necessary, conceptually we should just be caching the latest value we computed for the org ID/project ID

shashjar · 2026-06-08T23:15:10Z

@@ -218,15 +242,17 @@ def prepare_organization_report(
            lifecycle.record_halt(WeeklyReportHaltReason.EMPTY_REPORT)
            return

+    # Cache per-project metrics for reuse across report batches


can we put this behind an org feature flag to begin with?

cursor

Cursor Bugbot has reviewed your changes and found 2 potential issues.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

^{Reviewed by Cursor Bugbot for commit 59edfc4. Configure here.}

cursor · 2026-06-09T16:34:07Z

+            project_metrics[project_id] = {
+                "e": project_ctx.accepted_error_count,
+                "t": project_ctx.accepted_transaction_count,
+            }


Stale metrics after empty week

Medium Severity

cache_project_metrics skips Redis writes when check_if_project_is_empty() is true, so a zero-activity week leaves the prior week’s totals in cache. Week-over-week reads then treat old counts as the previous period instead of zero, and a cache hit avoids the intended Snuba fallback on miss.

^{Reviewed by Cursor Bugbot for commit 59edfc4. Configure here.}

10 day TTL avoids this. Since the report runs weekly, any stale entry expires before the next read

amy-chen23 · 2026-06-09T16:34:08Z

+
+
+def cache_project_metrics(ctx: OrganizationReportContext, organization_id: int) -> None:
+    project_metrics: dict[int, dict[str, int]] = {}


TODO: add feature flag

NVM -- not adding feature flag. We'll be adding feature flag to the PR that actually uses the cached metrics

…hat weekly_reports_cache is a pure cache layer

shashjar

two non-blocking comments, nice job

shashjar · 2026-06-10T17:52:36Z

+        raw = results[i]
+        if raw is None:
+            metrics.incr("weekly_report.cache.miss")
+        else:


should we record cache hits as well? another way you could do this would be to have a single metric called weekly_report.cache_read and then a tag on the metric is set to either "miss" or "hit" based on the result, and then DD allows you to break down those timeseries

this can be part of your next PR btw, doesn't have to block this one

shashjar · 2026-06-10T17:53:50Z

+
+from sentry.utils import json, metrics, redis
+
+CACHE_TTL = timedelta(days=10)


can just set this as an int, e.g. 10 * 24 * 60 * 60

CACHE_TTL_SEC = 10 * 24 * 60 * 60 # 10 days

Adds cache hit metric to weekly report cache More details on caching found in this [merged PR](#116739)

Resolves [ID-1589](https://linear.app/getsentry/issue/ID-1589/cache-total-project-errors-and-total-project-transactions-for-weekly). Goal: Cache per-project total errors and total transactions during weekly report generation so the frontend can display week-over-week percentage change without re-querying Snuba. **Redis Cache Layer** - Caches total errors and total transactions by Org Id, Project Id (no timestamp) - 10 day TTL --------- Co-authored-by: getsantry[bot] <66042841+getsantry[bot]@users.noreply.github.com>

Adds cache hit metric to weekly report cache More details on caching found in this [merged PR](#116739)

amy-chen23 added 3 commits June 2, 2026 14:58

update weekly report dev environment

8984411

feature flag for cache api endpoint

e120c09

cache logic for weekly report. cache is populated when the weekly rep…

a29ab0d

…ort is run

github-actions Bot added the Scope: Backend Automatically applied to PRs that change backend components label Jun 2, 2026

🛠️ Sync API Urls to TypeScript

8f90afe

github-actions Bot added the Scope: Frontend Automatically applied to PRs that change frontend components label Jun 2, 2026

vercel Bot deployed to Preview June 2, 2026 22:43 View deployment

sentry-warden Bot reviewed Jun 2, 2026

View reviewed changes

Comment thread src/sentry/api/endpoints/organization_weekly_report_metrics.py Outdated

amy-chen23 and others added 3 commits June 2, 2026 15:55

gate cache layer behind feature flag

103b522

moving imports up

419adb7

Merge branch 'master' into amyc/cache-weekly-report-metrics

1bab33b

vercel Bot deployed to Preview June 2, 2026 23:14 View deployment

amy-chen23 added 2 commits June 3, 2026 09:40

removing frontend changes

b6f5ee3

removing frontend change (for real this time)

c076d81

vercel Bot deployed to Preview June 3, 2026 17:41 View deployment

🛠️ Sync API Urls to TypeScript

0e4314c

vercel Bot deployed to Preview June 3, 2026 17:44 View deployment

amy-chen23 added 2 commits June 3, 2026 10:46

fixing cache miss issues with SATURDAY send time

8edfe8b

imported correct redis cluster. also added -> None to all test method…

20fd594

…s (typing)

amy-chen23 commented Jun 3, 2026

View reviewed changes

amy-chen23 closed this Jun 8, 2026

amy-chen23 reopened this Jun 8, 2026

amy-chen23 changed the title ~~feat(reports): Cache per-project weekly report metrics and expose via API~~ feat(reports): Cache per-project weekly report metrics Jun 8, 2026

amy-chen23 added 4 commits June 8, 2026 14:15

removed API endpoint logic

fa2b08c

removing file diff

a49272d

adding cache miss metrics

cf72ce3

adding math.floor for saturday (so less cache misses)

2d0b585

amy-chen23 requested a review from a team June 8, 2026 21:37

amy-chen23 removed the Scope: Frontend Automatically applied to PRs that change frontend components label Jun 8, 2026

amy-chen23 marked this pull request as ready for review June 8, 2026 21:53

amy-chen23 requested review from a team as code owners June 8, 2026 21:54

cursor Bot reviewed Jun 8, 2026

View reviewed changes

Comment thread src/sentry/tasks/summaries/weekly_reports.py Outdated

amy-chen23 commented Jun 8, 2026

View reviewed changes

shashjar reviewed Jun 8, 2026

View reviewed changes

removed timestamp from cache key

59edfc4

cursor Bot reviewed Jun 9, 2026

View reviewed changes

amy-chen23 commented Jun 9, 2026

View reviewed changes

no caching during dry run

94689e7

sentry Bot reviewed Jun 9, 2026

View reviewed changes

Comment thread src/sentry/tasks/summaries/weekly_report_cache.py Outdated

added feature flag. moved metric extraction to weekly_reports.py so t…

bde2262

…hat weekly_reports_cache is a pure cache layer

sentry Bot reviewed Jun 9, 2026

View reviewed changes

Comment thread src/sentry/tasks/summaries/weekly_reports.py Outdated

removing feature flag and wrapping caching in a try-catch

732aa53

amy-chen23 requested a review from shashjar June 10, 2026 16:41

shashjar approved these changes Jun 10, 2026

View reviewed changes

remove extra conversion of cache ttl

8c2ece0

amy-chen23 merged commit b658c3d into master Jun 10, 2026
64 checks passed

amy-chen23 deleted the amyc/cache-weekly-report-metrics branch June 10, 2026 18:31

amy-chen23 mentioned this pull request Jun 10, 2026

chore(weekly-report): add cache hit metric #117346

Merged

amy-chen23 added a commit that referenced this pull request Jun 10, 2026

chore(weekly-report): add cache hit metric (#117346)

9270c0a

Adds cache hit metric to weekly report cache More details on caching found in this [merged PR](#116739)

amy-chen23 added a commit that referenced this pull request Jun 10, 2026

chore(weekly-report): add cache hit metric (#117346)

2c01c64

Adds cache hit metric to weekly report cache More details on caching found in this [merged PR](#116739)

sentry-release-bot Bot mentioned this pull request Jun 15, 2026

publish: getsentry/sentry@26.6.0 getsentry/publish#8571

Closed

3 tasks



		def _make_cache_key(org_id: int, project_id: int, timestamp: float) -> str:
		return f"{KEY_PREFIX}:{org_id}:{project_id}:{timestamp}"



		def cache_project_metrics(ctx: OrganizationReportContext, organization_id: int) -> None:
		project_metrics: dict[int, dict[str, int]] = {}


		from sentry.utils import json, metrics, redis

		CACHE_TTL = timedelta(days=10)

Uh oh!

Conversation

amy-chen23 commented Jun 2, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

linear-code Bot commented Jun 2, 2026

Uh oh!

github-actions Bot commented Jun 2, 2026

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

JoshFerge commented Jun 8, 2026

Uh oh!

amy-chen23 commented Jun 8, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Uh oh!

shashjar commented Jun 8, 2026

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

cursor Bot left a comment

Choose a reason for hiding this comment

Uh oh!

cursor Bot Jun 9, 2026

Choose a reason for hiding this comment

Stale metrics after empty week

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

shashjar left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

amy-chen23 commented Jun 2, 2026 •

edited

Loading

amy-chen23 commented Jun 8, 2026 •

edited

Loading