fix(subscriber): keep request status updates on the owning save's connection

[fallenbagel]

Description

The scanner loop could wedge the entire Postgres pool: every per-item Media save holds its connection inside an open transaction while the subscriber helpers awaited since #3223 acquire a second connection through the global getRepository(), so with bundles of 50 against the default pool of 10 and no acquisition timeout, every connection ends up held by a save waiting on a connection that can never be freed. This routes the subscriber DB work through event.manager so it rides the save's own connection, wrapped in event.manager.transaction() which nests as a savepoint on both the postgres and sqlite drivers, so a failure in related-request work rolls back to the savepoint and is logged as before instead of aborting the save. The detached addMovie/addSeries chains keep their own repositories since the transaction is gone by the time they run, and the per-season updates run sequentially because they now share one connection.

I also added a db timeout configuration which defaults to 30000ms when not set in env variables.

• Fixes Library scans (on some setups) deadlock the Postgres connection pool (app unresponsive, no errors, restart required) #3365

How Has This Been Tested?

• I only tested unit tests but this can be tested using the preview tag (preview-pgsql-starvation-fix). Awaiting testing by the reported users.
• Tested and confirmed working by the reported users.

Screenshots / Logs (if applicable)

Checklist:

• I have read and followed the contribution guidelines.
• Disclosed any use of AI (see our policy)
• I have updated the documentation accordingly.
• All new and existing tests passed.
• Successful build pnpm build
• Translation keys pnpm i18n:extract
• Database migration (if required)

Summary by CodeRabbit

• Bug Fixes

  • Improved reliability of media request and status updates by keeping related database operations within the same transaction.
  • Preserved correct ordering when processing TV season request updates.
  • Prevented unnecessary external service checks when requested movies are already available.
  • Added a configurable database connection wait timeout, defaulting to 30 seconds.

• Documentation

  • Documented the new database connection timeout setting, including how to allow indefinite waiting.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro Plus

Run ID: 1624c3ca-4792-49cc-87c8-952feb725bc2

📥 Commits

Reviewing files that changed from the base of the PR and between 7e79970 and 9c63bec.

📒 Files selected for processing (1)

• server/utils/nestedTransaction.ts

🚧 Files skipped from review as they are similar to previous changes (1)

• server/utils/nestedTransaction.ts

---

📝 Walkthrough

Walkthrough

PostgreSQL connection acquisition now has a configurable timeout. Media subscribers use event-bound EntityManager repositories and transactions. Radarr and Sonarr request processing receives the event manager, while detached callbacks retain independent repositories.

Changes

Media transaction handling

Layer / File(s)	Summary
Database connection timeout configuration server/datasource.ts, docs/extending-seerr/database-config.mdx	PostgreSQL datasources read DB_CONNECT_TIMEOUT_MS, with a 30-second default. Examples document TCP and Unix socket settings.
Media subscriber transactions server/subscriber/MediaSubscriber.ts	Repository access uses event managers. Season processing is sequential. Child and related request updates run inside event-manager transactions.
Media request subscriber transactions server/subscriber/MediaRequestSubscriber.ts	Radarr and Sonarr processing uses manager-bound repositories. Parent-status updates use nested transactions. Detached callbacks use independent repositories.
Nested transaction handling server/utils/nestedTransaction.ts	withNestedTransaction runs callbacks directly on SQLite and through a TypeORM transaction for other databases.

Estimated code review effort: 4 (Complex) | ~45 minutes

Merge Risk: ⚪ Minimal · up to 9c63b

The PR updates database transaction handling and timeout configuration to keep related request updates on the owning save connection; no actionable merge-blocking risk remains after normal checks and review.

Possibly related issues

• Radarr Scan permanently exhausts the Postgres connection pool on 3.4.x (regression from 3.3.0) #3363: Tracks PostgreSQL connection-pool deadlocks caused by subscriber work using separate repositories during media scans.

Possibly related PRs

• seerr-team/seerr#3223: Both changes update awaited media request status handling in MediaSubscriber.

Sequence Diagram(s)

sequenceDiagram
  participant MediaSubscriber
  participant EntityManager
  participant MediaRequestSubscriber
  participant PostgreSQL
  MediaSubscriber->>EntityManager: use event-bound repository
  EntityManager->>PostgreSQL: update media and related requests
  MediaRequestSubscriber->>EntityManager: dispatch Radarr or Sonarr work
  EntityManager->>PostgreSQL: run parent-status transaction

Loading

Poem

A rabbit checks each transaction line,
Media moves through the manager fine.
Radarr waits, then Sonarr sings,
PostgreSQL times waiting things.
Bound repositories keep records bright.

🚥 Pre-merge checks | ✅ 4

✅ Passed checks (4 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title clearly identifies the subscriber connection-handling fix that addresses the primary pull request change.
Linked Issues check	✅ Passed	The changes route subscriber work through the owning EntityManager and preserve rollback behavior for issue #3365.
Out of Scope Changes check	✅ Passed	The datasource timeout configuration and documentation support the connection-pool objective, and no unrelated changes are evident.

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[Copilot]

Pull request overview

Fixes a Postgres connection-pool deadlock risk during high-concurrency media scans by ensuring subscriber-side DB work uses the same connection/transaction context as the originating save, and adds a configurable DB connection timeout to avoid silent hangs when the pool is exhausted.

Changes:

• Route MediaSubscriber request/season status updates through event.manager and wrap them in nested transaction() calls (savepoints) to avoid acquiring a second pool connection.
• Update MediaRequestSubscriber helper methods to use the event-scoped EntityManager, and isolate parent-status updates behind a savepoint.
• Add DB_CONNECT_TIMEOUT_MS configuration (and documentation) to bound DB connection waits.

Reviewed changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated 2 comments.

File	Description
server/subscriber/MediaSubscriber.ts	Uses event.manager repositories + savepoint transactions to keep request status updates on the owning save connection (avoids pool wedge).
server/subscriber/MediaRequestSubscriber.ts	Passes an EntityManager into helper methods and wraps parent-status updates in a savepoint; keeps detached *arr follow-up using global repositories.
server/datasource.ts	Introduces DB_CONNECT_TIMEOUT_MS-driven timeout configuration for Postgres connections.
docs/extending-seerr/database-config.mdx	Documents the new DB_CONNECT_TIMEOUT_MS environment variable.

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[Copilot]

Pull request overview

Copilot reviewed 4 out of 4 changed files in this pull request and generated no new comments.

[Alexsaphir]

Tested on kubernetes, with postgres 18 (CNPG).

Before this fix, to not soft lock:

DB_POOL_SIZE: 100

With this fix, no soft lock:

DB_POOL_SIZE: 10
DB_CONNECT_TIMEOUT_MS: 10000

(Tested by using the image and validating that it work, with the default pool size)

[Copilot]

Pull request overview

Copilot reviewed 5 out of 5 changed files in this pull request and generated 1 comment.

[gauthier-th]

LGTM