Relay review wave 2: in-process wake, signin and tokens, scale, streaming attachments, federation pins, report digest, metrics #252
Assignees
extant2000git
Clear assignees
extant2000git
No Assignees
Notifications
Due Date
No due date set.
Dependencies
No dependencies set.
Reference: extant2000/flamenet-relay#252
Reference in New Issue
Block a user
Wave 2 of the relay architecture review 2026-10-14, on top of
5a0331e. In-process wake-ups (Redis only with FN_RELAY_REPLICAS>1) and size-first polls; per-IP signin bucket, bcrypt off the event loop, SHA-256 app credentials with bcrypt rehash; account-service aud/iss/jti behind FN_ACCOUNT_TOKEN_STRICT (off); batched presence, roster_tags(digest) and prekeys partial indexes, cached key-log head with a per-IP limit, set-based retention, batched last-seen writes; streaming attachments with auth before the body; FN_FEDERATION_PEERS host=key pins; per-reporter limit and hourly report digest; /metrics on FN_METRICS_PORT (off by default); docs say self-hosting needs Postgres. Full suite 971/971 twice with Postgres, e2e conformance 266/0/2. Generated with Claude Code.WakeTests "a send that commits between the cursor query and the wait" failed in full runs (14 s, 36 s, 42 s against a 5 s bound) and passed alone. The wake was never lost: with timestamps, the send committed and signalled at +13.41 s and the poll returned at +14.12 s. The 13 s went before the send, in a process-wide stall. A `sample` taken during a stall showed all 8 cooperative threads busy: 7 in Bcrypt.hash at the default cost 12, called synchronously from fixtures (RosterTagTests, MobileRoutesTests, MobileExtrasTests, PasswordResetTests account helpers), and 1 in blockingShutdown's semaphore. One cost-12 hash is 0.31 s in a debug build; cost 4 is 1.3 ms. Every fixture now hashes at cost 4, as WaveZeroBSupport already did, and a source scan fails if one goes back to the default. The full run drops from about 60 s to 45 s. WakeTests now times the poll from the moment the send committed, so a stall before the send cannot read as a lost wake. With the product's pending flag disabled it still fails, at 31.8 s. DeliveryOrderPostgresTests counted every ungranted advisory lock on the server, including other suites' key-log appends in their own databases, so the second send could start before the first had inserted ("free" got id 1). The wait now counts only this database's lock on holdKey. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>Closing in favour of a PR from the branch itself. This one was opened over AGit against refs/pull/252/head, so it cannot pick up the CI fix in #253 that its image job needs: that job failed twice here on mac1 and passes on mac2. The branch release/relay-review-wave2-2026-10-14 now has main merged into it, including that fix, and carries the same 18 commits.
Pull request closed