Hacktoberfest 2026: the issues maintainers tagged for October, open and beginner-friendly. Browse Hacktoberfest issues

cluster: registerEntity does not reopen the storage read, so messages read before registration wait for entityMessagePollInterval

Open
  • Is something wrong?
#9,717 0 comments 0 reactions 0 assignees View on GitHub

Maintainers usually reply within 1 day

Nobody has claimed this yet.

Assessment

Difficulty
3/5
Estimated time
1-2 days
Newbie friendliness
72/100
Issue type
Bug
Clarity
Clearly specified
Activity status
Active
Tech stack
typescript
Domain
distributed-systems

Research direction

Start in packages/effect/src/cluster/Sharding.ts at the registerEntity logic around lines 1928-1937, then trace how storageReadLatch is opened by the poll fiber and how reads reset addresses for unregistered entity types around lines 860-864. Reproduce with persisted messages and no new client message; done means registration triggers an immediate storage reread and the pending handler runs without waiting for entityMessagePollInterval.

Written by the indexing model from the issue text.

Description

What happens

At boot, Sharding starts reading persisted messages as soon as it is built. An entity can only register after that, because registerEntity needs Sharding. When a read finds a message for an entity type that is not registered yet, Sharding resets the address and drops the message from that read ("reset address in the case that the entity is slow to register").

registerEntity then sets the entity manager inside withStorageReadLock, but it does not open storageReadLatch. With no new client message, nothing opens the latch until the poll fiber does, after entityMessagePollInterval (10 s by default). So every persisted message for that entity waits up to 10 s after a restart.

Where (packages/effect/src/cluster/Sharding.ts on main; same shape in the published 4.0.5)
  • the read skips an unregistered type: lines 860-864 (MutableHashSet.add(resetAddresses, address));
  • with no client message and no full batch, only the poll fiber opens the latch: lines 1071-1076;
  • registerEntity sets the manager and opens only the per-type latch: lines 1928-1937.
Reproduction
  1. A SqlMessageStorage-backed SingleRunner with a persisted entity message whose handler has not run (for example, the process was killed mid-handler).
  2. Restart. Build the runner, then register the entity in a later layer.
  3. Send no new message. The handler runs about entityMessagePollInterval after boot, not at registration.

In our app (gent), a resumed turn took 10.26 s with the default interval and 0.24 s after we call sharding.pollStorage once after registerEntity. With the interval set to one hour, the message did not run within a 10 s test window.

Suggested fix

Open storageReadLatch after entityManagers.set(...) in registerEntity (when the shard set is already acquired), so messages reset during the registration window are read again at once.

Workaround

Call Sharding.pollStorage after each registerEntity at the composition root.

Dominant language
TypeScript
Stars
16.7k
Forks
808
Avg merge
10h 7m
Merged PRs (30d)
808

Getting set up

This project ships no dev container, Dockerfile or contributing guide, so setting up is up to you: start from its README, and see our first-contribution guide for the general steps.

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

More from Effect-TS/effect

All issues in Effect-TS/effect

Similar issues

More TypeScript issues

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.