5 Commits

Author SHA1 Message Date
Andrey Antukh
e23ca02aa2
✨ Reference the file data storage object with a foreign key (#12769)
* ✨ Reference the file data storage object with a foreign key

The file data pointer lived in `file_data.metadata.storage-ref-id`, a JSON
key. The storage GC parsed it to find live references, and nothing stopped it
from physically deleting an object a live row still pointed at: a failure in
that scan lost the blob.

Migration 0159 moves the pointer to a nullable `storage_object_id` column on
`file_data`, backfills it from the metadata key, removes the key from the
rows it backfills, indexes the column and rebuilds the foreign key to
`storage_object(id)` as `NO ACTION DEFERRABLE INITIALLY IMMEDIATE`.
`INITIALLY IMMEDIATE` is deliberate: the GC catches the violation inside the
chunk transaction, while a deferred check would fire at commit, outside it,
and the sweep would never move past the blocked chunk.

The column is the only source of truth. The `storage-ref-id` entry is gone
from the file metadata schema, and the single point that serializes the
metadata drops it, so no write can resurrect it. `upsert!` does not query
the row: the pointer of the row being replaced travels in its params, and
the object is touched before the row is written. Fragments and snapshots
pass nil because a modified pointer always gets a new id; the offload passes
its own row; `insert-file!` drops it, since an INSERT never replaces a
persisted object.

`update-file!` takes the file map read from the database plus the fields to
change and merges them, so the pointer comes from the read map and no caller
can lose it by building a partial map. A map without `:data` raises
`:missing-file-data`: it is not `assert`, which is compiled out in the
deployed binary. The params cannot override the pointer or retarget the
write with `:id`/`:file-id`.

`bfc/file->response` strips the pointer from every response (`get-file`,
`create-file`, `duplicate-file`, the view-only bundle, the team access
request, the comments read, the library summary, the snapshot read and the
debug download). A `file-data` object is authentication-only, so a leaked id
would let any logged-in profile fetch the blob from `/assets/by-id`.
`get-file-fragment` resolves through the file data backend, so a fragment on
cold storage no longer hands nil to `blob/decode`, and answers only the keys
of its result schema.

The storage GC reference scan, the restore union and `objects-gc` read the
column. The offload reads `SELECT fd.*`, so its rows go through a
`decode-row` that decodes jsonb metadata and drops a JSON null, which the row
schema rejected; a row without a readable object raises
`:missing-storage-object` with the file and row ids. A chunk blocked by the
foreign key (SQLState 23503) is a real anomaly, since the reference scan
should have frozen the object: it stays an error with its own hint and the
chunk is retried on the next run until the reference is fixed.

The release is a coordinated restart, not a rolling deploy, and not
rollback-safe for the rows it writes: the previous code writes and scans the
metadata pointer, so a key it writes misses the backfill and removing it
loses the reference. `operations/2.20.0-file-data-storage-object-id.md`
builds the column, the index and the foreign key online and runs the
backfill with every instance stopped, then marks the step as applied; a
fresh or small instance does the whole job in the migration. The runbook
keeps the validation scan out of the stopped window and adds a lock timeout
and retry to the online statements.

Tests cover the foreign key shape, the backfill and its corrupt and orphan
guards, the rebuild of a missing index and of a wrong, non-deferrable or
deferred constraint, the deletion refused by a live reference, the metadata
cleanup, the offload row normalizations, the touched objects of the unshare
and the restore, the response stripping on every call site and the chunk
blocked by the foreign key. The backend memories and the developer guide
document the pointer, the GC behavior and the two migration paths.

Closes #12768

AI-assisted-by: deepseek-v4.1-flash
AI-assisted-by: inkling-2-small-free

* 🐛 Hold the file row lock across the migration write

`get-file` applies the pending migrations in memory and, when it finds
any, persists the result. That write read the row without a lock and
`update-file!` is an unconditional UPDATE, so a save that committed in
between was overwritten: the saved content, `revn` and `modified-at`
went back, the next save from that client failed with `:revn-conflict`
and the object the save left became an orphan the storage GC cannot
revisit once it is frozen.

The read now runs in a transaction of its own (`get-file` calls
`get-file*` inside `db/tx-run!` and accepts a cfg map only), so the
`file` row lock covers the re-check and the write. `get-file*` reads
the row once (`sql:get-file` plus `resolve-applied-migrations`) and
completes it with `attach-file-data`; only when a migration is pending
it reads the file again under `FOR UPDATE`, re-checks `need-migration?`
on that fresh map and persists it (`migrate-file`). The common case
stays at one row read and one migrations resolution, the migration case
keeps a single decoded copy in the heap, and a row that changed in
between is not overwritten. A migration found on the unlocked read is
applied in memory only, for the next read to persist.

`migrate-file` returns the reattached map when it persists, with the
`:storage-object-id` of the row the write left, and the realized map
when it does not; `:realize?` is what returns the read fully plain.
`fdata/upsert!` answers `{:written ... :storage-object-id ...}` instead
of a boolean, `update-file!` propagates it and the migrating read puts
it back in the map it returns, so a caller that persists again touches
the current object instead of the replaced one. `insert-file!` keeps
nil: an INSERT never replaces a persisted object.

The write also compares the stored pointer with the caller's value
(`IS NOT DISTINCT FROM`) and raises `:stale-file-data-pointer` when it
affects no row. Every writer holds the `file` row lock, so a pointer
that no longer matches is a protocol violation, not a race to retry.
The expected pointer is a separate argument because the storage branch
replaces `:storage-object-id` before the upsert, and the check lives in
`upsert-in-database`: the `legacy-db` branches do not use that SQL and
their count can legitimately be zero.

The save, snapshot and restore-backup reads take `FOR UPDATE` instead
of `FOR SHARE`: they inherit the lock, and the migration branch asks
for the exclusive one in the same transaction, so a shared lock that
upgrades deadlocks two concurrent readers. A data read rejects
`:lock-for-share?` for the same reason; only row-only reads accept it.
`sql:get-file-without-data` is a plain select now, not a CTE, which
makes locking legal without loading the data. Realization has a single
owner: `attach-file-data` only resolves and decodes, and `get-file*`
realizes once per path. The applied migrations no longer travel in the
`get-file-with-summary` response, cache hit and cache miss alike.

`embed-assets` reads each library without `:realize?`, so its
`:components` is an unloaded pointer map; `absorb-assets` walks it with
`vals`, `PointerMap.seq` calls `load!` and that raised with no loader
bound, so every export with `:merge-libraries` of a library with
components failed. The absorb runs under the library's own loader now.

The in-place binfile import (`import-file-and-overwrite*`,
`::overwrite`, `::bfc/file-id`) had no product caller and cannot carry
the previous `storage_object_id`, so overwriting a storage-backed row
would leave its object unreclaimed. `import-files!` rejects a cfg with
`::bfc/file-id` (`:not-supported`) and the helpers only that path used
go with it (`invalidate-thumbnails`, `*reference-file*`, the
`::overwrite` id branch of `index-object` and of `save-file!`, the
`on-conflict-do-nothing?` flag of the media insert).

Tests: a new `file-migration-read-test` pins the single read, the
locked re-read, the concurrent save interleaving, the double migration
and the read-only path; the upsert tests cover the return shape and the
stale pointer; the SQL recorder they share lives in
`backend-tests.helpers`, and the save, snapshot, restore and summary
sequences follow the new reads. The backend memories document the read
contract and the pointer.

AI-assisted-by: deepseek-v4.1-flash

* 🐛 Harden the get-file guard and skip its read-only transaction

The cfg guard of `get-file` was an `assert`, which `app.config` elides
when `:backend-asserts` is off, so a deployed binary accepted a bare
pool or connection while the docstring promised it raises. It throws
`IllegalArgumentException` now, the class the option guards of the same
namespace already use, and the existing test expects it instead of
`AssertionError`.

A read with `:read-only?` applies the pending migrations in memory and
writes nothing, so the transaction the migration write needs only costs
two round trips there. `get-file` runs it with `db/run!` instead: with
a cfg that carries a connection it still joins the caller's
transaction, and only the reads that would have opened their own skip
it. A lock option needs a read-write transaction from the caller,
because PostgreSQL rejects row locks in a read-only one; the docstring
states that.

Tests: the guard test expects the new class; a cfg marked
`::db/read-only` reads a pending migration without persisting it (the
config flag that fills that key is legacy, but the key is set by hand);
and a read-only read is pinned as not opening a transaction while the
same call without the option does. The backend memory documents the
split and the guard.

AI-assisted-by: deepseek-v4.1-flash
2026-11-14 19:27:18 +01:00
Andrey Antukh
679e70e48c 🐛 Keep file_change kind nullable so rollback keeps writing
Migration 0158 created file_change with kind NOT NULL, which blocks
any emergency rollback to the previous version: the old code inserts
without kind and every save would fail. kind is still a reserve
(reads use label IS NOT NULL), so it stays nullable until the
rollback window closes; the NOT NULL comes in a later migration.
The runbook keeps parity, the schema and migration tests cover the
nullable column and the rollback insert, and the storage memory
records the decision.
AI-assisted-by: muse-spark-1.3-contributor
2026-11-13 08:05:15 +00:00
Andrey Antukh
9bac1d290b 📚 Clarify when to run the file_change partition procedure
The procedure straddles the upgrade to 2.20.0 and the document said
nothing about the order. The preparation (phases 0 and 1) goes before
the version update, the bulk and the delta passes after it, and only
the cutover needs a scheduled maintenance window. The subtitle claimed
the whole thing was online, which is not true.

The copy marks no longer travel as a session GUC with a date typed by
hand. The delta mark is seeded at the bulk with `now()`, and a CHECK
keeps it from ending up empty; the instant of the cutover is recorded
in the same table after the commit, so the checks of phase 5 read it
instead of comparing against a date the operator retypes in every
session.

The procedure was run end to end on a scratch database, with the 23
SQL blocks extracted from the document and the application writes
simulated between the phases.

AI-assisted-by: deepseek-v4.1-flash
2026-11-13 06:58:46 +00:00
Andrey Antukh
34c702bc80
✨ Add table partitioning for file-change (#12417)
* ✨ Partition file_change by file_id

Partition `file_change` by `HASH (file_id)` in 16 partitions, with the
same bounds as `file_data`, so a join between them is resolved partition
by partition. The primary key becomes `(file_id, id)`, which a
partitioned table needs, and every access filters by `file_id`.

The new table carries a `kind` column (1 snapshot, 2 revision) as a
reserve: the queries still tell them apart by the label, and `kind`
takes over when they are rewritten. The payload of a snapshot stays in
`file_data`; the `data` column is kept for reading old `legacy-db`
payloads.

The four snapshot commands require `file-id`, because it prunes the
partition. The objects GC locks and deletes by the pair, in batches. The
old table is kept as `file_change_legacy` as the way back.

The automatic migration refuses to run above 10 GiB or 2M rows and
points to the runbook; big instances partition online with that runbook,
which copies by id, follows the `updated_at` watermark in delta passes
and checks that no row of the source is missing from the target.

Closes #12416

AI-assisted-by: deepseek-v4.1-flash

* 🐛 Address the review of the file_change partition

The production runbook promised a rollback it did not have: the section
was a sketch, and its statements could write to the table about to be
dropped. The rollback is removed, and so is every comment that described
it. `file_change_legacy` stays as a frozen copy of the previous layout.

The cutover check now aborts the transaction when a row of the old table
is missing in the new one, instead of leaving the decision to the
operator.

The five secondary indexes of the new table are built after the last
delta pass: the bulk and the delta passes only need the primary key, so
the batches no longer maintain them.

The copy of 0158 is now replayed in a test against the pre-0158 shape
with a labelled and an unlabelled row, plus an empty instance that must
leave no legacy table behind. The updated_at assertion of the snapshot
mutations is bounded by the test clock, and the repair revision stops
writing the `features` of the file, like the normal save.

AI-assisted-by: deepseek-v4.1-flash

* 📎 Fix linter issues

* 🐛 Fix stale helper call in rpc_file_snapshot_test

The suite does not load: `th/run-pending-tasks!` does not exist in
backend-tests.helpers. The pending jobs runner is `run-pending-jobs`
(no arguments), the name every other call site of the suite uses.

AI-assisted-by: inkling-2-small-free

* 📚 Align the partition change with the 2.20.0 milestone

2.19.0 is already shipped; this work targets 2.20.0. The operations
file of the partition procedure carries the 2.20.0 prefix, the
references that point at it (the migration, the runbook table and the
memory) follow, and the RPC doc/changes tags of the snapshot commands
record 2.20 as the release that makes `file-id` required.

AI-assisted-by: inkling-2-small-free

* 🐛 Fix review findings on the partition rebase

The 0157 legacy test was missed by the rebase and inserted file_change
rows without the new mandatory `kind` column, so the backend suite fails
against the migrated schema. The inserts now carry `:kind`
(snapshot-kind for the labelled rows, revision-kind for the rest).

The post-cutover verification block of the runbook terminates the
first query with a semicolon so the block pastes into psql without a
syntax error.

The guard message of the migration names the step
(0158-partition-file-change) instead of "0158", ambiguous since two
migrations share that number.

Closes #12416

AI-assisted-by: inkling-2-small-free

* 🐛 Fix the stale label in the snapshot restore test

The rename block of snapshots-crud runs before the restore block,
so the restored version is already named "label1-updated" when the
restore publishes its message. The command names the version by the
label the row has at that moment, not by the one it had when it was
created, so the test was asserting a stale value.

Assert the current label, and say why in the comment.

AI-assisted-by: deepseek-v4.1-flash

* ♻️ Split the version event tests into one test per command

The workspace version events and the plugin proxy send the file-id
of the open file on every snapshot command, because the command
needs it to prune the partition of file_change. Both tests checked
that with a doseq and t/testing blocks, so a failure named the whole
group and not the command that broke.

Give each command its own test. The workspace ones share a helper
that runs the event and returns the command and params it sends; the
plugin ones keep their own mock of rp/cmd!.

AI-assisted-by: deepseek-v4.1-flash
2026-11-10 23:27:17 +01:00
Andrey Antukh
35baa06606
✨ Migrate legacy file payloads to file_data and delete orphans (#12785)
* ✨ Migrate legacy file payloads to file_data and delete orphans

Copy file.data and labelled file_change.data to file_data on boot
(type main/snapshot, backend db, deleted_at preserved), then delete
file_change rows with data but no label, which no read path reaches.
Large instances pre-migrate with batched statements and fake this
step. Closes #12784.

AI-assisted-by: muse-spark-1.3-contributor

* 📚 Add 2.19.0 operations runbook for legacy file_data migration

Manual pre-migration procedure for big instances, same statements as
migration 0157 in batched transactions, following the convention of
the file-change partition runbook. Closes #12784.

AI-assisted-by: muse-spark-1.3-contributor

* 🐛 Address review feedback on the legacy file data migration

Fix the issues raised in the review of the migration that moves the
legacy file payloads to file_data.

The migration deletes the unlabelled file_change rows that carry data,
but restore-files still cleared their date, and the retention test
asserted they came back. A row without a date never reaches objects-gc,
which is the leak fixed in #12419. Restore now unmarks only snapshots,
the rows that carry a label, and the retention test checks that the
legacy payload keeps its date instead of coming back.

Cover the conflict path that had no test: a pre-existing snapshot that
collides with a labelled change row, and a second run that must leave
the legacy columns null and the orphan row absent.

The runbook analyses file_data after the inserts, so the planner does
not wait for autovacuum.

Drop app.srepl.procs.fdata-storage: the copy it performed is what the
migration replaces and nothing requires it.

Closes #12784

AI-assisted-by: deepseek-v4.1-flash
2026-11-04 12:25:01 +01:00