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

Flaky e2e: gateway-fleet reads a worker's capacity before the worker has left starting

Closed
  • Is something wrong?
#414 1 comment 0 reactions 0 assignees View on GitHub

Maintainers usually reply within 1 day

A pull request for this has already been merged.

  • #421 by @V3RON — merged

Assessment

Difficulty
4/5
Estimated time
1-2 days
Newbie friendliness
48/100
Issue type
Bug
Clarity
Clearly specified
Activity status
Active
Tech stack
typescript
Domain
distributed-systems, testing

Research direction

Start in e2e/gateway-fleet.test.ts at waitForWorkers and iosLimit, then read src/gateway/worker-link.ts (#rebuildView, #refreshStarting) and the changesCapacityOrLeases filter to see why daemon.started never triggers a refresh. Triage must settle the product question (refresh on leaving starting) before the test fix, and an open PR (#421) already claims the work; done means waitForWorkers waits for capacity/catalog presence and CI passes under the CPU-load conditions described.

Written by the indexing model from the issue text.

Description

bug:ready flaky-test

What fails

e2e/gateway-fleet.test.ts > "gateway fleet > two workers join, status aggregates them, a kill disconnects one, drain flags the other" fails with AssertionError: expected 0 to be greater than 0 at expect(iosLimit(joined[0])).toBeGreaterThan(0). Seen once in CI on PR #409 (run 37477061691).

Cause: a race on main, not in a PR

The test waits only for both workers to be connected, then reads capacity from that first view. A worker that dials the gateway while its daemon is still starting answers status.get with health and host only (#398), so the gateway builds its view without capacity (worker-link.ts #rebuildView -> #refreshStarting). The view stays without capacity until the next refresh, and none is triggered when the worker becomes running: daemon.started is not a lease./device./install event (changesCapacityOrLeases), so the next read is a device or lease event or the 30s periodic tick. The test's iosLimit falls back to 0.

Evidence

  • On origin/main (37188b5), unmodified: 6 of 6 runs of the file pass under CPU load (16 busy loops on 12 cores), 5 of 5 unloaded.
  • On origin/main with await new Promise((r) => setTimeout(r, 1500)) as the first line of convergeStartup in src/daemon/main.ts (a slower startup): 3 of 3 runs fail with this exact assertion (2 tests failed each time).
  • The window is the worker's startup time, so anything that lengthens startup (PR #409 adds a serial read, doctor pass and lease reconcile) makes it likelier. It is not specific to that PR.

Two things to decide in triage

  1. Product: should the gateway refresh a worker's view as soon as it leaves starting? Today a freshly joined worker takes no routed work for up to 30s after its startup ends if no device or lease event happens first (check FleetCoordinator/routing on a view with no catalog).
  2. Test: waitForWorkers in the file should wait for capacity (and catalog) to be present before the test reads it.
Dominant language
TypeScript
Stars
15
Forks
1
Avg merge
9h 6m
Merged PRs (30d)
138

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 callstackincubator/simlock

All issues in callstackincubator/simlock

Similar issues

More TypeScript issues

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.