chore(dolt): converge bd dolt clean-databases prefix list with firewall (#3618)
Converge the cleanup-side `staleDatabasePrefixes` list in `bd dolt clean-databases` with the firewall-side `testDatabasePrefixes` from be-c5p, so both prefix lists are textually identical and `benchdb_` / `beads_test` are no longer leaked by cleanup. Co-authored-by: Jim Wordelman <jim@wordelman.name> Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com> Agent-Signature: claude-fable-5-high on behalf of matt wilkie
This commit is contained in:
parent
a3e6deb0e8
commit
64d67e0976
2 changed files with 118 additions and 11 deletions
|
|
@ -903,16 +903,35 @@ servers are preserved.`,
|
|||
},
|
||||
}
|
||||
|
||||
// staleDatabasePrefixes identifies test/agent databases that should not persist
|
||||
// on the production Dolt server. These accumulate from interrupted test runs and
|
||||
// terminated agents, wasting server memory.
|
||||
// - testdb_*: BEADS_TEST_MODE=1 FNV hash of temp paths
|
||||
// - doctest_*: doctor test helpers
|
||||
// - doctortest_*: doctor test helpers
|
||||
// - beads_pt*: orchestrator patrol_helpers_test.go random prefixes
|
||||
// - beads_vr*: orchestrator mail/router_test.go random prefixes
|
||||
// - beads_t[0-9a-f]*: protocol test random prefixes (t + 8 hex chars)
|
||||
var staleDatabasePrefixes = []string{"testdb_", "doctest_", "doctortest_", "beads_pt", "beads_vr", "beads_t"}
|
||||
// staleDatabasePrefixes lists database name prefixes that
|
||||
// `bd dolt clean-databases` will drop. This is the cleanup side of the
|
||||
// test/prod split. Two sibling lists must converge with it (be-avn):
|
||||
// - internal/storage/dolt/store.go:testDatabasePrefixes (firewall side)
|
||||
// - .gc/system/packs/dolt/formulas/mol-dog-stale-db.toml (city formula)
|
||||
//
|
||||
// The firewall list, this cleanup list, and the formula list MUST
|
||||
// converge — operators rely on consistent semantics across `bd dolt
|
||||
// clean-databases`, the SQL-side firewall, and `gc dolt cleanup`.
|
||||
//
|
||||
// Origin of each prefix:
|
||||
// - testdb_ : applyConfigDefaults derives this for BEADS_TEST_MODE=1
|
||||
// without an explicit Database (FNV hash of cfg.Path).
|
||||
// - beads_test : convention for hand-written integration tests.
|
||||
// - beads_pt : property-test fixtures.
|
||||
// - beads_vr : version-roundtrip / migration fixtures.
|
||||
// - doctest_ : `bd doctor` self-check fixtures.
|
||||
// - doctortest_ : older `bd doctor` fixture name (kept for back-compat).
|
||||
// - benchdb_ : per-bench scratch DBs (uniqueBenchDBName below,
|
||||
// format `benchdb_<pid>_<8 hex>`).
|
||||
var staleDatabasePrefixes = []string{
|
||||
"testdb_",
|
||||
"beads_test",
|
||||
"beads_pt",
|
||||
"beads_vr",
|
||||
"doctest_",
|
||||
"doctortest_",
|
||||
"benchdb_",
|
||||
}
|
||||
|
||||
var doltCleanDatabasesCmd = &cobra.Command{
|
||||
Use: "clean-databases",
|
||||
|
|
@ -920,7 +939,7 @@ var doltCleanDatabasesCmd = &cobra.Command{
|
|||
Long: `Identify and drop leftover test and agent databases that accumulate
|
||||
on the shared Dolt server from interrupted test runs and terminated agents.
|
||||
|
||||
Stale database prefixes: testdb_*, doctest_*, doctortest_*, beads_pt*, beads_vr*, beads_t*
|
||||
Stale database prefixes: testdb_*, beads_test*, beads_pt*, beads_vr*, doctest_*, doctortest_*, benchdb_*
|
||||
|
||||
These waste server memory and can degrade performance under concurrent load.
|
||||
Use --dry-run to see what would be dropped without actually dropping.`,
|
||||
|
|
|
|||
88
release-gates/be-ripy-gate.md
Normal file
88
release-gates/be-ripy-gate.md
Normal file
|
|
@ -0,0 +1,88 @@
|
|||
# Release gate — be-ripy (be-avn `staleDatabasePrefixes` canonical sync)
|
||||
|
||||
**Date:** 2026-04-30
|
||||
**Deployer:** beads/deployer (deployer-1, fresh evaluation post-builder-rebase)
|
||||
**Bead (review):** be-ripy — Review: be-avn staleDatabasePrefixes canonical sync (AD-04 follow-up)
|
||||
**Feature bead:** be-avn (closed; source bead)
|
||||
**Reviewed commit:** `0f76453f` (rebased equivalent: `ee67d5f9` on `be-vzu-rebase-fix`)
|
||||
**Final branch:** `release/be-ripy` @ `1014c759` (cherry-pick of `ee67d5f9` onto `origin/main`)
|
||||
**Base:** `origin/main` @ `8694c535` ("doctor: detect AGENTS.md / CLAUDE.md user-authored divergence (#3600)")
|
||||
|
||||
## Verdict: PASS
|
||||
|
||||
## What this ships
|
||||
|
||||
`bd dolt clean-databases` cleanup-side prefix list converged with the
|
||||
firewall-side `testDatabasePrefixes`. After this change:
|
||||
|
||||
- `staleDatabasePrefixes` becomes `[testdb_, beads_test, beads_pt, beads_vr,
|
||||
doctest_, doctortest_, benchdb_]` — the architect-canonical 7-prefix list.
|
||||
- Net behavioral change vs. prior list:
|
||||
- **+** `benchdb_` now cleaned (closes the AD-04-documented bench-DB leaker)
|
||||
- **+** `beads_test` now cleaned (matches firewall scope when be-c5p lands)
|
||||
- **−** `beads_t<hex>` protocol-test prefix no longer matched by cleanup;
|
||||
those workspaces clean themselves up in tearDown per AD-04 problem statement
|
||||
- The doc-block on `staleDatabasePrefixes` is restructured to mirror
|
||||
`testDatabasePrefixes` (per-prefix origin notes plus a cross-reference
|
||||
to the two sibling lists). The `Long:` help string is updated.
|
||||
|
||||
## Criteria
|
||||
|
||||
| # | Criterion | Result | Evidence |
|
||||
|---|-----------|--------|----------|
|
||||
| 1 | Review PASS present | PASS | reviewer-gm-po6pox3 PASS verdict in be-ripy notes — 0 blockers, 0 high, 0 medium, 1 advisory low (out-of-scope doctor copy follow-up). |
|
||||
| 2 | Acceptance criteria met | PASS | All 4 ACs walked by reviewer (canonical list match, sibling list already matches, cross-reference doc-block, build/lint clean) — re-verified below. |
|
||||
| 3 | Tests pass | PASS | Build + vet clean on full repo; targeted tests PASS (see below). |
|
||||
| 4 | No HIGH-severity review findings open | PASS | 0 HIGH findings. The 1 LOW advisory (doctor's separate `staleDatabasePrefixes` is not converged) is explicitly out-of-scope per the bead's "Files in scope". |
|
||||
| 5 | Final branch is clean | PASS | `git status` clean on `release/be-ripy` (only untracked items are `.gc/` and `.gitkeep` from worktree harness, unrelated to this change). |
|
||||
| 6 | Branch diverges cleanly from main | PASS | `git cherry-pick ee67d5f9` onto `origin/main@8694c535` applied without conflict. Only one file changed (`cmd/bd/dolt.go`); no path overlap with the 32 commits on `origin/main` since the prior rebase base `f4c46d91`. |
|
||||
|
||||
## Test evidence (criterion 3)
|
||||
|
||||
Run from `release/be-ripy` @ `1014c759`:
|
||||
|
||||
- `go build -tags gms_pure_go ./...` — clean (exit 0).
|
||||
- `go vet -tags gms_pure_go ./...` — clean (exit 0).
|
||||
- `go test -tags gms_pure_go -count=1 -run 'TestStaleCommand|TestDoltClean' ./cmd/bd/` — PASS (0.089s). Includes `TestStaleCommandInit` (Docker-gated `TestStaleSuite` skipped — Docker not available in deployer harness; runs on CI).
|
||||
- `go test -tags gms_pure_go -count=1 -run 'TestStaleDatabasePrefixes' ./cmd/bd/doctor/` — PASS (0.073s, 12 sub-tests). The doctor package has its own `staleDatabasePrefixes` list (out-of-scope per AD-04 bead), confirmed unaffected.
|
||||
|
||||
The reviewer's targeted suite (`go build`, `golangci-lint run`) was independently re-verified at the source SHA before rebase. No new code paths in this commit — the change is a constant-list edit plus a doc-block rewrite, so existing coverage applies.
|
||||
|
||||
## Cherry-pick mechanics (criterion 6)
|
||||
|
||||
This deploy follows the precedent of PR #3562 / `release/be-l9q`: the bead
|
||||
covers a single commit, so the deployer cherry-picks just that commit onto
|
||||
fresh `origin/main` rather than shipping the full source branch. The
|
||||
reviewed SHA `0f76453f` was rebased by the builder to `ee67d5f9` after the
|
||||
gate-FAIL on the prior `be-l9q` deploy attempt; the builder confirmed
|
||||
content equivalence (`git diff ee67d5f9 0f76453f` shows only an unrelated
|
||||
`.gitattributes` upstream-noise hunk).
|
||||
|
||||
Cherry-pick on top of current `origin/main` (8694c535) was conflict-free.
|
||||
Touched file (`cmd/bd/dolt.go`) was not modified by any of the 32 commits
|
||||
on `origin/main` since the prior rebase base, so the rebased commit's tree
|
||||
applies textually.
|
||||
|
||||
## Cross-list convergence note (informational)
|
||||
|
||||
The new doc-block declares the cleanup list, the firewall list
|
||||
(`internal/storage/dolt/store.go:testDatabasePrefixes`), and the city
|
||||
formula list "must converge". On current `origin/main`, the firewall
|
||||
list does not yet contain `benchdb_` (that addition lands with be-c5p,
|
||||
not yet shipped). After this deploy:
|
||||
|
||||
- Cleanup list (this PR): `[testdb_, beads_test, beads_pt, beads_vr,
|
||||
doctest_, doctortest_, benchdb_]` — 7 prefixes
|
||||
- Firewall list (`origin/main`): `[testdb_, beads_test, beads_pt,
|
||||
beads_vr, doctest_, doctortest_]` — 6 prefixes (missing `benchdb_`)
|
||||
|
||||
The two lists drift on `benchdb_` until be-c5p ships. This is a
|
||||
defense-in-depth gap (firewall doesn't reject `benchdb_*` at creation
|
||||
time on the production server), not a runtime regression — the cleanup
|
||||
job still drops them. Not a gate blocker.
|
||||
|
||||
## Hand-off
|
||||
|
||||
- Final branch `release/be-ripy` @ `1014c759` ready to push.
|
||||
- PR target: `gastownhall/beads:main` (origin), pushed via `quad341/beads`
|
||||
(fork) per the rig's push-target protocol.
|
||||
Loading…
Add table
Add a link
Reference in a new issue