feat(outbox): archive old outbox events to cold storage

[nicolasburtey]

Port of GaloyMoney/lana-bank#7757 to the cala outbox (the cala side was explicitly out of scope there: "cala outbox archiving — needs cala on the new obix"). Wires the obix event-archive feature (GaloyMoney/obix#105, git branch dependency until it is released) into cala-ledger.

Summary

• cala-ledger/src/outbox/config.rs: OutboxArchiveConfig { storage, retention_days = 3, path_prefix = "outbox-archive/cala/" }. cala is a library without a cloud-storage crate, so the consumer supplies the backend via obix::EventArchiveStorage (GCS, S3, local fs; InMemoryArchiveStorage for tests). The boundary is obix's age-based DailyRetentionBoundary over CalaMailboxTables.
• CalaLedgerConfig: optional .outbox_archive(...) — when set, settled spans of history are swept from postgres to object storage as JSONL chunks and pre-watermark reads transparently fall back to the archive.
• CalaLedger::register_outbox_archiver(&mut job::Jobs): registers the cala.outbox.archiver job on the app's job executor (cala owns no executor of its own) — sweeps one settled day per run, rescheduling while catching up.
• Migration: cala_persistent_outbox_archive_chunks manifest folded into the existing cala obix setup migration (in-place edit, matching the recent pre-production edits of that file).
• .sqlx: offline cache regenerated (the 5 new derive-generated archive queries).

Not in scope / follow-ups

• Released obix: revert the git dependency to crates.io once feat!: archive old outbox events to object storage obix#105 is merged and released — a git dep blocks publishing cala-ledger.
• lana's dw-outbox-relay can then drop its obix-cala =0.5.0 pin and unify on one obix version.
• Settlement-marker-based boundary provider (cf. lana's intended EndOfDaySettled follow-up.

Test plan

• — incl. new end-to-end test : manual-clock day 0/1/3 events, archiver job sweeps days 0–1 to in-memory storage and prunes pg, day 3 stays, listener replay from crosses the archive→pg seam contiguously (runs against its own throwaway database — the shared test DB has past-dated events from other tests, which make shared-table span assertions meaningless)
• clean,
• and
• SQLX_OFFLINE=true cargo run -p cala-ledger --bin event-schemas --features json-schema
  Wrote AccountEvent schema to /Users/n/code/galoy/cala/cala-ledger/schemas/account_event_schema.json
  Wrote AccountSetEvent schema to /Users/n/code/galoy/cala/cala-ledger/schemas/account_set_event_schema.json
  Wrote JournalEvent schema to /Users/n/code/galoy/cala/cala-ledger/schemas/journal_event_schema.json
  Wrote TransactionEvent schema to /Users/n/code/galoy/cala/cala-ledger/schemas/transaction_event_schema.json
  Wrote EntryEvent schema to /Users/n/code/galoy/cala/cala-ledger/schemas/entry_event_schema.json
  Wrote TxTemplateEvent schema to /Users/n/code/galoy/cala/cala-ledger/schemas/tx_template_event_schema.json
  Wrote VelocityLimitEvent schema to /Users/n/code/galoy/cala/cala-ledger/schemas/velocity_limit_event_schema.json
  Wrote VelocityControlEvent schema to /Users/n/code/galoy/cala/cala-ledger/schemas/velocity_control_event_schema.json
  git diff --exit-code cala-ledger/schemas clean
• Verify against lana: cala outbox archiving over GCS, dw-outbox-relay on the unified obix
  EOF
  )

[github-actions]

📊 Performance Report

Commit: b265562
Updated: 2026-08-02 17:30:07 UTC

Cala Performance Benchmark Results (non-representative)

Criterion Benchmark Results (single-threaded)

Benchmark	Time per Run	Throughput	% vs Baseline
post_simple_transaction	4.428ms	225 tx/s	0 (baseline)
post_and_recalculate_ec_account_set	11.940ms	83 tx/s	-169.0%
post_and_batch_recalculate_ec_account_set	9.249ms	108 tx/s	-108.0%
post_multi_layer_transaction	8.039ms	124 tx/s	-81.0%
post_simple_transaction_with_effective_balances	5.898ms	169 tx/s	-33.0%
post_simple_transaction_with_skipped_velocity	4.089ms	244 tx/s	+7.0%
post_simple_transaction_with_velocity	5.357ms	186 tx/s	-20.0%
post_simple_transaction_with_hit_velocity	2.426ms	412 tx/s	+45.0%
post_simple_transaction_with_one_account_set	5.714ms	174 tx/s	-29.0%
post_simple_transaction_with_five_account_sets	8.205ms	121 tx/s	-85.0%
post_simple_transaction_with_ec_account_set	4.343ms	230 tx/s	+1.0%

Load Testing Results (parallel-execution)

Scenario	tx/s
1 parallel	149.82
2 parallel	209.20
5 parallel	227.45
10 parallel	273.48
20 parallel	284.00
2 contention	156.75
5 contention	195.02
2 acct_sets	153.87
5 acct_sets	163.78

---

Note: Performance results may vary based on system resources and database state.

Last updated by commit b265562