Post-merge review of #6658 (landed as 5b9e938b5) found two guidance defects
in the #6575 data-behind stop. The gate itself is correct; both are in the
remedy chain it prints.
1. The shared arm prescribed `bd migrate schema`, which cannot succeed in the
state that prints it. A data-behind refusal has a remote by construction —
behind-ness is read from `remotes/<name>/<branch>` — and the verb consent
(`sharedMigrateConsent`) has exactly one reader, inside `sharedNoRemoteGate`,
reachable only on the `!hasRemote` branch. So after the pull the retry routed
`smartAutoMigrate -> shared -> fallbackReasonSharedStore`, whose blunt body
prescribes the `bd migrate --force` the data-behind body explicitly says not
to use. The option, `dataBehindBody`, `AgentDirective`, `fallbackReasonNote`
and both `--global` retargets now name `bd migrate schema --force`, which
trips `forceOrEnvConsent` via `isForcedMigrate`.
Retargeting the command rather than widening the consent read is deliberate:
with a remote configured, #4259's cross-clone fork risk still demands the
designated-migrator confirmation, which is what `sharedNoRemoteGate`'s own
comment says the scoping is for.
2. The JSON payload's `docs` pointer stayed on the migrate-or-adopt anchor,
whose ordering rule says `bd dolt pull` is refused on every
pending-migration open — contradicting the `pull-first` option in the same
payload. The `IsDataBehind()` arm now points at a new "Clone behind the
remote" section, and the two living ordering rules (upgrading.md,
init-safety.md) record the exception.
Tests: the consent step is now checked by running it — the command is taken
from `Options()`, resolved through the same cobra lookup and predicates the
root pre-run uses, and fed to the real gate against a remote-backed shared
fixture, with the shipped bare verb as the counterfactual. Asserting the
option's wording is what let the dead command ship. `docs` is pinned in
`TestHandleRemoteMigrateGateJSON_DataBehind`.
Co-authored-by: Claude Opus 5 <noreply@anthropic.com>
25 KiB
| title | description |
|---|---|
| Upgrading | Upgrade the bd binary, refresh git hooks, run schema migrations, and handle remote-backed and cross-era databases |
How to upgrade bd and keep your projects in sync.
Checking for Updates
# Current version
bd version
# What changed since the version you were running — the delta, not the archive
bd upgrade review
bd upgrade review --json # Machine-readable
# The full release history (large; every version bd knows about)
bd info --whats-new
Short Version
- With your current
bd, sync remote-backed databases before installing the new binary:bd dolt pushbd dolt pull - Back up with that same current binary, before installing:
bd export --all -o .beads/backup/pre-migrate-$(date +%Y%m%d).jsonl - Upgrade using the command that matches your install method.
- After upgrading:
bd upgrade reviewbd hooks installbd version - If crossing a schema migration on a remote-backed database, only the
designated migrator runs:
bd migratebd dolt push
Other clones should install the new binary and run bd bootstrap, not
independently migrate. The full procedure is below.
Upgrading
Use the command that matches your install method.
| Install method | Platforms | Command |
|---|---|---|
| Quick install script | macOS, Linux, FreeBSD | curl -fsSL https://raw.githubusercontent.com/gastownhall/beads/main/scripts/install.sh | bash |
| PowerShell installer | Windows | irm https://raw.githubusercontent.com/gastownhall/beads/main/install.ps1 | iex |
| Homebrew | macOS, Linux | brew upgrade beads |
| go install (server-mode only) | macOS, Linux, FreeBSD, Windows | CGO_ENABLED=0 go install github.com/steveyegge/beads/cmd/bd@latest |
| go install (embedded-capable) | macOS, Linux, Windows | CGO_ENABLED=1 GOFLAGS=-tags=gms_pure_go go install github.com/steveyegge/beads/cmd/bd@latest |
| npm | macOS, Linux, Windows | npm update -g @beads/bd |
| bun | macOS, Linux, Windows | bun install -g --trust @beads/bd |
| From source (Unix shell) | macOS, Linux, FreeBSD | git pull && make build |
Quick install script (macOS/Linux/FreeBSD)
curl -fsSL https://raw.githubusercontent.com/gastownhall/beads/main/scripts/install.sh | bash
PowerShell installer (Windows)
irm https://raw.githubusercontent.com/gastownhall/beads/main/install.ps1 | iex
Homebrew
brew upgrade beads
{/* Canonical Homebrew tap-migration snippet. The installation page links
here; keep the two in sync. */}
If you still have the old tap formula installed as bd, switch to the
Homebrew core formula:
brew uninstall bd
brew untap gastownhall/beads 2>/dev/null || true
brew untap steveyegge/beads 2>/dev/null || true
brew install beads
go install
# Server-mode only
CGO_ENABLED=0 go install github.com/steveyegge/beads/cmd/bd@latest
# Embedded-capable
CGO_ENABLED=1 GOFLAGS=-tags=gms_pure_go go install github.com/steveyegge/beads/cmd/bd@latest
From Source
cd beads
git pull
make build
sudo mv bd /usr/local/bin/
After Upgrading
Important: After upgrading, update your hooks:
# 1. Check what changed
bd info --whats-new
# 2. Update git hooks to match new version
bd hooks install
# 3. Check for any outdated hooks
bd info # Shows warnings if hooks are outdated
# 4. If using Dolt backend, restart the server
bd dolt stop && bd dolt start
Why update hooks? Git hooks are versioned with bd. Outdated hooks may miss export refresh, legacy fallback, or safety fixes.
Database Migrations
After major upgrades, check for database migrations:
# Inspect migration plan (AI agents)
bd migrate --inspect --json
# Preview migration changes
bd migrate --dry-run
# Apply migrations
bd migrate
# Migrate and clean up old files
bd migrate --yes
Upgrading to 1.3.0
A 1.2.2 or 1.1.x database sits at schema v53, and 1.3.0 knows v66. That is more than a routine upgrade: 1.2.2 was a recovery release that shipped the 1.1 code under a higher version number, so there is a release line's worth of schema between the two.
Back up first, with the binary you have now
Do this before you install 1.3.0. Under the new binary, bd export triggers
the migration before it exports, so a snapshot taken afterwards is a
post-migration snapshot and cannot protect you against the migration going
wrong. The same ordering rule as the section below applies: do all syncing with
your current binary, since once 1.3.0 is installed the pending-migration
gate refuses bd dolt push and bd dolt pull too.
# with your CURRENT bd:
bd dolt push # remote-backed stores only
bd export --all -o .beads/backup/pre-1.3.0-$(date +%Y%m%d).jsonl
For a Dolt-native snapshot that keeps history and config, configure a
destination and sync it. Bare bd backup takes no backup — it is a command
group that prints help and exits 0:
bd backup init <path-or-dolthub-url> # once, to configure a destination
bd backup sync # take the snapshot
What the migration looks like
On an embedded or local store, the first command you run after installing
applies the whole set, in place. (A shared dolt sql-server is never
auto-migrated — see Shared servers below.) The main series
runs 0054 → 0066 (13 migrations), and then the clone-local series runs through
the same printer with its own numbering, 0012 → 0026 — so it is about
28 migrations, and the counter visibly restarts partway through:
Applying migration 0065: widen_wisp_comments_text…
Applying migration 0066: add_events_journal_actor…
Applying migration 0012: create_leases… ← clone-local series, not a restart
A counter jumping backwards looks exactly like a loop. It is not one — let it finish. Expect the run to take noticeably longer than the commands after it (two passes rewrite rows rather than reshaping tables). It is crash-resumable and picks up where it left off.
Those lines go to stderr, and only when stderr is a terminal, so a piped or CI upgrade prints nothing at all. Silence there is not a stall either.
Upgrade every client that shares a store, together
A bd binary refuses a database migrated past the schema it knows, rather than
proceeding blind. One machine upgrading takes the shared store forward and every
client still on 1.2.2 stops working. Check for a second binary earlier in your
PATH with which -a bd, and restart any long-running bd serve — noting that
bd --readonly serve is now refused outright, so a server scripted with that
flag will not come back up until you drop it.
If the store is a shared dolt sql-server rather than a local one, follow
Shared servers below: 1.3.0 will not migrate it without
explicit consent, precisely so the fleet upgrade can come first.
If you need to go back, the rollback is a schema-cursor rollback rather than a downgrade of the data — see Accidental v1.2.1 Release, whose procedure is the same for any cursor rollback even though its worked example is that release.
Once you are on the new binary, bd upgrade review prints exactly the changes
between the version you were running and this one. Several commands changed
defaults, so read it before your first session.
Remote-backed databases and multiple clones
bd refuses to silently apply pending schema migrations to a database that has
a Dolt remote configured. Migrating more than one clone of a shared remote
independently forks the schema, after which bd dolt pull can no longer merge —
the break is silent and, across a primary-key-reshaping migration, unrecoverable
(#4259). The supported flow
is: one machine migrates and publishes; every other clone re-clones the migrated
database.
This applies to every upgrade that crosses a pending migration on a remote-backed database — the same procedure whether you are moving to a prerelease or to a stable release.
The gate is state-aware by default
(#4516): before blocking,
bd consults the remote's cached schema state and
- auto-migrates when the remote is at the same schema version as this
clone — no one has migrated yet, so this clone is a safe first-mover
(concurrent first-movers converge to identical tables). It reminds you to
bd dolt pushafterwards. This applies to embedded mode only: a shared server always stops for consent, because migrating it changes the schema every connected client sees (see Shared servers below). - stops and directs you to adopt (
bd bootstrap) when the remote has already been migrated by another clone. - stops for a human decision when this clone and the remote applied different content for the same migration (a genuine fork), or when the remote's schema state cannot be read from the cached ref.
Set BD_SMART_GATE=0 to opt out and make the gate block unconditionally.
The recipes below are the explicit path and work the same in either mode.
Important ordering: once the new binary is installed, a database with
pending migrations is gated on every open — bd dolt push and bd dolt pull are refused too, not just bd migrate. So do all syncing with your
current binary, before you install the new one.
There is one exception, and the gate names it when it applies: if the stop is
the data-behind one — this clone is level with the remote on schema but is
missing commits it has not pulled — then bd dolt pull is the remedy and is
allowed through, because migrating before that pull is what wedges the clone.
See Clone behind the remote. Every other
pending-migration refusal still blocks the pull, so the ordering rule above is
what to plan for.
Back up before you migrate. Schema migrations assume the database matches the shape the previous migrations left behind; real databases sometimes drift (interrupted writes, tooling bugs, very old bootstraps). A JSONL export is cheap, issue-complete, and importable by any bd version:
bd export --all -o .beads/backup/pre-migrate-$(date +%Y%m%d).jsonl
bd export captures issues, not Dolt history or config — for a full snapshot
also copy the .beads directory (or dolt backup in server mode) while no
bd command is running.
Single clone (including a solo user with a remote):
bd dolt push # 1. CURRENT binary: publish all local work
bd export --all -o .beads/backup/pre-migrate.jsonl # 2. backup (see above)
# 3. install the new binary (see Upgrading above)
bd migrate # 4. migrate as the designated migrator
bd dolt push # 5. publish the migrated schema
bd version # 6. confirm the new version is active
If bd's remote-migrate gate blocks the run, it prints the available options —
migrating here as the designated migrator, adopting the remote's already-migrated
database, or recovering a fork — and asks for an explicit operator decision.
Follow the guidance it prints.
For scripted or CI upgrades where nobody reads the output, run bd migrate as
an explicit step in exactly one job, never in all of them. If the gate blocks a
run it prints both the available options and the scripted override that fits
the situation.
Multiple clones sharing one remote:
# 1. With your CURRENT (old) binary, on EVERY clone: publish all work and get in
# sync, then stop editing until the upgrade is done.
bd dolt push
bd dolt pull
# 2. Designated migrator ONLY: back up, install the new binary, then migrate
# and publish.
bd export --all -o .beads/backup/pre-migrate.jsonl
bd migrate
bd dolt push
# 3. Every OTHER clone: install the new binary, then ADOPT the migrated database.
# (bd dolt pull is refused here — the clone still has pending migrations — so
# re-clone instead. Safe because step 1 already pushed all work.)
bd bootstrap
bd bootstrap replaces the local database, so any work not pushed in step 1 is
lost — that is why step 1 publishes everything first. If a clone was instead
migrated independently and bd dolt pull later fails with cannot merge because table dependencies has different primary keys in its common ancestor, the
schema has already forked — follow the recovery playbook:
the pk-fork-refused runbook.
Clone behind the remote
This is the one pending-migration stop whose remedy is a bd dolt pull, and
the one the gate links here rather than to the migrate-or-adopt recipe above.
It fires when the clone is level with the remote on schema but behind it in
data — the remote has commits this clone has not pulled. On schema alone the
clone looks like a safe first-mover, so before
#6575 the gate let it
migrate; the new schema commits then landed on a history missing those
commits, and once a migration moves a tracked table onto Dolt's ignore plane
every later bd dolt pull refuses to merge
(#6368). The clone can no
longer fetch what it was already behind on.
The remedy is the pull, and nothing else:
bd dolt pull # allowed through this stop, unlike other refusals
# then re-run the command that was refused
Nothing local is discarded: with no commits of its own the pull is a pure
fast-forward, and if the clone has diverged the pull merges (resolve any
conflicts, or decide them in bulk with bd dolt pull --strategy ours|theirs
on embedded storage / bd conflicts resolve on a server store). Once there is
nothing left to pull, the clone is the first-mover it looked like and the
migration proceeds on its own.
Do not force past this stop. bd migrate --force applies the migration
while the clone is still behind — precisely the state the stop exists to
prevent — and the bd dolt push that would follow is rejected as a
non-fast-forward anyway.
On a shared server, the pull is still the first step, but the migration after it needs explicit consent, because it promotes the schema for every co-resident client at once (see Shared servers). Confirm those clients are upgraded, then:
bd dolt pull # first, as above
bd migrate schema --force # add --global for the shared global database
--force is the consent form here because this stop always has a remote
configured — the behind-ness is read from the remote-tracking ref — and the
bare bd migrate schema consent applies only to a shared database with no
remote. It is safe at this point and only at this point: the pull has already
landed the commits the clone was missing.
Shared servers
A server-mode database is served to every bd client connected to that
dolt sql-server, so a schema migration is not a local event: it promotes the
schema version for all of them at once, and clients still running an older
bd refuse the database until they are upgraded too. bd therefore never
auto-migrates a shared database on a version bump — with or without a remote
configured, though the two cases consent differently
(#5920).
Upgrade one server's clients like this:
# 1. Upgrade bd on every client of the server. Reads keep working throughout —
# an upgraded client reads the old schema, it just cannot write to it.
bd version # on each client, confirm the new version
# 2. Once, from a workspace already set up against this server: consent.
bd migrate schema # add --global for the shared global database
# 3. Confirm.
bd doctor
Between steps 1 and 2, an upgraded client reads normally and its writes are refused with the gate's guidance. Nothing is silently promoted, so there is no deadline — but the window is a degraded one, so keep it short.
If the shared server also has a Dolt remote, step 2 is not enough. Two
hazards now apply at once — the co-resident lockout above and the cross-clone
fork of Remote-backed databases
— so bd requires the stronger designated-migrator consent it describes:
bd migrate --force, from exactly one machine, followed by bd dolt push. If
another clone has already migrated and pushed, adopt its database with
bd bootstrap instead of migrating. On a shared server, adopting also promotes
the schema for every client of that server, so step 1 still comes first either
way.
A new client joining a server whose schema is behind is refused before its
workspace is written, so there is nothing to run bd migrate schema from
there: do step 2 from a client that is already set up, or join and migrate in
one step with BD_ALLOW_REMOTE_MIGRATE=1 bd init ….
The same rules apply in proxied-server mode, which is a shared server reached through a local proxy. Two differences worth knowing:
- read commands print a warning and keep serving the current schema, rather than the read simply not touching it;
bd serverefuses to start against a database with pending migrations, because a daemon has no operator to consent for it. Reconcile the schema first with step 2, then start the daemon — or, for an unattended service, putBD_ALLOW_REMOTE_MIGRATE=1in its environment as an explicit, auditable standing consent.
Cross-era Upgrades
If you're upgrading from a much older version of bd, inspect the storage layout
and metadata before running the current binary. A .beads/dolt/ directory alone
does not identify a legacy workspace: supported current server mode uses that
directory too. Current bd evaluates explicit server metadata, the presence of
that local root, and the bounded .local_version witness together. An explicit
server selection is not overridden by a stale .beads/embeddeddolt/ repository.
| Storage layout | Upgrade path |
|---|---|
Current embedded metadata with .beads/embeddeddolt/ and no explicit server selection |
Direct current-era upgrade |
Explicit server metadata plus .local_version from v0.55.4 through v0.62.0, whether or not .beads/dolt/ exists |
Explicit legacy Dolt export/import |
Explicit server metadata plus .beads/dolt/ and a witness whose major version is 1 or newer |
Normal current server-mode upgrade |
Explicit server metadata plus .beads/dolt/ and a missing or pre-v1 witness |
Explicit legacy Dolt export/import |
Explicit server metadata plus .beads/dolt/ and a witness that is present but unreadable |
Normal current server-mode upgrade, with a warning |
Explicit server metadata without .beads/dolt/, and a missing, malformed, or non-historical witness |
Normal current server-mode compatibility path |
.beads/dolt/ with missing metadata or persisted dolt_mode blank/embedded |
Explicit legacy Dolt export/import, except for the configured shared-server compatibility path described below |
One .beads/*.db file, such as beads.db or vc.db |
Sealed SQLite bridge |
The witness is whatever bd held in its own version string when it last touched
the workspace, so it may be a plain release, a release candidate, a build
carrying metadata, or a Go pseudo-version; all of those are read as the version
they name. A witness that is present but unreadable is not treated as a legacy
marker — no pre-v1 bd could have written one — so bd warns and continues
rather than refusing every command. A missing witness stays ambiguous and is
still refused.
Current bd refuses recognized historical SQLite and legacy Dolt layouts before
opening storage or rewriting metadata. This is intentional: preserve the source
and complete the matching explicit migration below.
PostgreSQL and MySQL are removed backends, not supported cross-era upgrade
paths. Current bd refuses metadata that selects either backend, and the sealed
bridge below accepts SQLite sources only.
.beads/embeddeddolt/: direct upgrade
Upgrade the binary and run:
bd migrate
If the project was initialized before bd init automatically wired git origin
as the Dolt remote, verify the remote after upgrading:
bd dolt remote list
When the list is empty, fix it on the machine whose local database is authoritative:
bd export -o .beads/issues.pre-remote.jsonl # optional issue audit export
bd dolt remote add origin git+ssh://git@github.com/org/repo.git
bd dolt push
Commit the resulting .beads/config.yaml change so other clones can run
bd bootstrap or bd dolt pull.
Historical Dolt server mode: explicit migration
Do not run current bd init --force when .beads/dolt/ has missing metadata
or persisted dolt_mode is blank/embedded. Those old embedded layouts are
never current embedded storage. The same explicit path applies when metadata
selects backend: dolt, dolt_mode: server and .local_version records
v0.55.4 through v0.62.0.
First take a native snapshot while every writer is stopped. Restore that snapshot to a disposable workspace, export it with the verified historical binary, then import the export into a fresh current project. Keep the original and snapshot unchanged until the cutover has been reviewed. The SQLite sealed-copy helper below does not start or manage a Dolt SQL server.
Explicit server metadata with a v0.55.4–v0.62.0 witness is always refused,
including when there is no local .beads/dolt/ root. When that root does exist,
the guard admits explicit server mode only with a syntactically valid witness
whose major version is 1 or newer; a missing, malformed, or pre-v1 witness fails
closed. Without the local root, a missing, malformed, or non-historical witness
is admitted only as a compatibility layout.
The configured shared-server compatibility path applies only when persisted
metadata is missing or leaves dolt_mode blank/embedded; it does not override
an explicit server selection with a local root. Compatibility admission cannot
prove that a workspace is modern. If you know it was created by v0.55.4 through
v0.62.0, use this explicit bridge even when its witness was lost or damaged.
Otherwise, follow the normal bd migrate --dry-run and bd migrate flow for
an admitted server workspace.
One .beads/*.db file: sealed SQLite bridge
The old binary stored data in SQLite. The new binary uses Dolt.
Recommended: use the sealed-copy bridge (requires jq):
Stop every process that can write the old workspace before starting.
Run this from a source checkout at the exact commit you intend to run; installed
binaries do not include repository scripts. Record that commit with
git rev-parse HEAD before executing the script. Download the old bd asset
only from the official gastownhall/beads release and verify the asset with its
published SHA-256:
sha256sum -c checksums.txt --ignore-missing
On macOS or BSD, use shasum -a 256 on the downloaded archive and compare
the result with that archive's entry in checksums.txt.
scripts/migrate-legacy-to-current.sh \
--source /absolute/path/to/old-project \
--destination /absolute/path/to/old-project-cutover \
--source-version v0.50.3 \
--old-bd /absolute/path/to/verified-old-bd \
--new-bd /absolute/path/to/current-bd \
--prefix beads
For a source older than v0.49.6, also supply an authenticated v0.49.6 binary:
scripts/migrate-legacy-to-current.sh \
--source /absolute/path/to/old-project \
--destination /absolute/path/to/old-project-cutover \
--source-version v0.17.0 \
--old-bd /absolute/path/to/verified-old-bd \
--canonicalizer-bd /absolute/path/to/verified-v0.49.6-bd \
--new-bd /absolute/path/to/current-bd \
--prefix beads
Verify the historical binary against its official release checksum before running the bridge. The script verifies each binary's reported version, rejects Dolt and ambiguous layouts, retains a sealed source copy, and compares the candidate export with the canonical historical export. Activate the cutover manually only after reviewing those retained artifacts.
Troubleshooting Upgrades
Hooks out of date
bd hooks install
Database schema changed
bd migrate --dry-run
bd migrate
Recovery after upgrade
If you need to restore from a backup:
bd init
bd backup restore [path] --force
Or pull from a Dolt remote:
bd dolt pull