Aleksey Shakhmatov
f2844e7894
Every reply came back in a single batch with `cursor.id = 0`, `getMore` was a
stub answering an empty `nextBatch` on the literal namespace `test.$cmd`, and
nothing read `batchSize`. That caps the useful collection size at what fits in
one 48 MiB message, which is the opposite of the tens-of-GB target and the
reason M0 made whole-index scans stream: the streaming candidate generator
existed with no consumer that could suspend.
## What a cursor is allowed to remember
A cursor holds no lock between requests, so everything it saves has to survive
arbitrary concurrent mutation. Nothing here is a pointer, and the two things
that look like stable addresses are not: `reset_tree` re-creates node ids 0 and
1 as different nodes, and `rebuild_collection` moves every document. Three
sources, chosen by query shape, each with a different memory contract:
- **stream** -- an index-ordered walk resumed from a `(key, off)` anchor plus a
`(leaf, slot)` hint. O(key) state, so this is what lets a cursor walk a
collection larger than memory. Survives a rebuild, because a repack changes
no key.
- **offsets** -- the matched slab offsets a narrowed plan already materialized,
8 bytes each. Killed by a rebuild with `QueryPlanKilled`, because those
offsets now name unrelated bytes.
- **buffered** -- canonical BSON copies, for a sort no index provides and for
aggregate/listing output. Depends on nothing, which is what lets a listing
hold a cursor over a `$cmd.*` namespace no collection backs.
`Collection.layout_epoch` and `Index.epoch` are the invalidation tokens, both
checked as error returns rather than assertions since a client reaches them by
keeping a cursor open across maintenance.
## Resume
`resume_forward`/`resume_reverse` are O(1) while the hint holds and fall back to
an exact-order band walk bounded by `resume_walk_max`. Without the hint, `seek`
lands at the *start* of an equal-key band, so `sort({status: 1})` over three
distinct values across 10M documents would cost ~5e10 comparisons to drain.
Two hazards found by draining a collection while writing to it, neither
predictable from reading the code:
- A deleted anchor must resume at its *band position*, or the rest of an
equal-key band is silently dropped -- most of the collection on a
low-cardinality index. Hence `band_index`.
- On a **unique** index a same-key entry can only be the anchor rewritten, so
resuming at it returned updated documents twice. Observed as duplicate `_id`s
while updating underneath a drain.
## Protocol
Measured against mongod 8.3.7 rather than recalled, which corrected three
assumptions: a bare `getMore` does *not* inherit the find's `batchSize` (4998 of
5000 documents come back), a namespace mismatch is `Unauthorized` (13) not
`CursorNotFound`, and `CursorInUse` is 143 not 12051.
`internalQueryFindCommandBatchSize` is 101, `cursorTimeoutMillis` 600000,
`clientCursorMonitorFrequencySecs` 4.
The rule everything follows is **never look ahead**: a batch that met its target
leaves the cursor open even when the source is in fact exhausted, so four
documents at `batchSize: 2` take three commands. `limit` acts as an EOF source,
which is what makes `batchSize == limit` close in one round trip. `skip` is
consumed once. `batchSize: 0` returns an empty batch with a live cursor.
Cursor ids are `(nonce << 20) | slot`, always positive. The nonce is not
decoration: without it a recycled slot serves one client another's documents.
Cursors are not connection-pinned, since the driver spec allows a `getMore` on
any connection to the same server; they end at exhaustion, `killCursors`, or the
idle sweep (a second monitor fiber, separate from the TTL one because the
cadences differ by an order of magnitude and a TTL failure must not stop
reclamation). The registry is fixed-capacity and evicts the least recently used
cursor, whose client sees the same 43 an idle timeout gives.
Fixed alongside, because cursors are what expose them:
- `listCollections` reported `"<db>."` with an *empty* collection part, which
makes the driver throw client-side -- so it would have broken the moment its
cursor stopped being id 0. Now `<db>.$cmd.listCollections`, as mongod uses.
- `count` ignored `skip` and `limit` entirely.
- `wire.end_message` now bounds a reply by the 48 MiB we advertise rather than
by `maxInt(u32)`; a reply past what we told the client to expect is not a large
reply, it is a desynchronized connection.
- Two `codeName` strings were wrong: 72 is `InvalidOptions` (MongoDB has no
`InvalidArgument`), and 40324 reports as `Location40324`.
## Verification
Unit 160/160 in ReleaseFast and ReleaseSafe; `tests/e2e/e2e7.js` adds 86 cursor
checks across five phases (batching/lifecycle/errors, streaming across churn,
aggregate+listings+count, expiry+capacity, restart) and is self-contained
because cursor behaviour is only observable with non-default flags. No
regressions: e2e 49, e2e3 16, e2e4 17, e2e2 2, e2e6 72. Spec 168 pass / 124
fail, +5 against the previous scorecard.
Mutation-checked, per the repo's second ground rule: `hint_slot + 1`, the
`band_index` off-by-one, both epoch bumps, the id nonce, the at-least-one-
document rule, and `stream_shape` returning null each turn the intended test
red. One claim was withdrawn rather than kept -- swapping `std.mem.order` for
`cmp_prefix` in the band walk changes nothing observable, so the comment now
says so instead of asserting a check that does not hold.
5.6 KiB
End-to-end tests with the official MongoDB Node.js driver
These exercise MultiforaDB from a real driver over TCP: full CRUD, query operators, aggregation, error codes, concurrent clients, crash recovery, and the whole lifecycle including server restarts.
Setup
cd tests/e2e
npm init -y >/dev/null
npm install mongodb
Run
Most suites expect a server running on port 27020:
zig build
zig-out/bin/multiforadb --port 27020 --db /tmp/mfdb-e2e.log --ttl-sweep-secs 1 &
node tests/e2e/e2e.js # CRUD + operators + aggregate + errors (29 checks)
node tests/e2e/e2e2.js concurrent # 8 clients: 4 writers + 4 readers (2 checks)
node tests/e2e/e2e2.js crash-a # write 50 docs, then kill -9 the server
node tests/e2e/e2e2.js crash-b # restart and verify all 50 survived
node tests/e2e/e2e3.js # secondary indexes: unique/sparse/compound (16 checks)
node tests/e2e/e2e4.js # TTL indexes: expiry + rejected specs (15 checks)
e2e4.js needs the server started with --ttl-sweep-secs 1 (the default is
60 seconds); the other suites do not care about the flag.
e2e6.js is the full-lifecycle suite and is self-contained: it spawns its
own server on port 27220 with a fresh log, runs the whole feature surface,
restarts the server twice (graceful SIGTERM, then kill -9 mid-write) and
verifies everything survived:
node tests/e2e/e2e6.js # 73 checks, ~15 s, needs no running server
E2E6_PORT=27300 node tests/e2e/e2e6.js # different port if 27220 is taken
e2e7.js is the cursor suite, self-contained for a different reason: cursor
behaviour is only observable with non-default flags. It spawns three servers in
turn -- default flags for batching/streaming/aggregate, then
--cursor-timeout-ms 800 --cursor-sweep-secs 1 --max-open-cursors 4 for idle
expiry and registry capacity, then a restart on the same database to confirm a
cursor does not survive one. Most of it uses raw runCommand, because the
driver hides cursor.id and that is the thing under test:
node tests/e2e/e2e7.js # 86 checks, needs no running server
E2E7_PORT=27310 node tests/e2e/e2e7.js # different port if 27230 is taken
Rebuild with zig build after any change under src/ before restarting the
server: zig build test compiles the test binary only and leaves
zig-out/bin/multiforadb stale, so the suites keep running against the old
rules and report failures that the source no longer explains.
e2e2.js concurrent is safe to repeat against a running server (it drops its
collection first); crash-a/crash-b are two halves of one scenario.
Multi-GB collections: big.js
big.js is a load harness, not a pass/fail suite: it spawns a server, bulk
loads up to ~5 GB, and reports insert throughput, the compaction behavior,
server RSS, per-operation latencies, reopen (replay) time, and kill -9
durability.
node tests/e2e/big.js --quick # 268 MB smoke run
node tests/e2e/big.js --size 5g --doc-size 128k --oid --batch 200 \
--compact-threshold 2g # ~5 GB, 40k docs
Options: --size/--doc-size/--batch (k/m/g suffixes), --oid
(ObjectId _ids — see below), --index <field> (secondary index before
loading), --compact-threshold <bytes> (passed to the server),
--port, --keep (keep the db file).
Measured behavior (all documented in the top-level README):
- Build in ReleaseFast —
zig builddefaults to it; a Debug server is 10-200x slower on every path. - Insert throughput collapses under the default 16 MiB compaction
threshold: every ~16 MB of writes rewrites the whole log with one fsync
per record (O(n²) total). With
--compact-threshold 2gthe rate stays flat (hundreds of MB/s at 128 KB docs in ReleaseFast). Raise the threshold for bulk loads. findOne({_id})is O(1) only for ObjectId_ids. Integer_ids are serialization-ambiguous (int32/int64/double compare equal but hash differently), so the docs-map fast path is skipped and every lookup is a full scan. Use the driver's default ObjectId ids on big collections.- The engine holds everything in RAM: ~1-1.2x the data size at 128 KB docs (more at 16 KB docs, where per-document arena overhead dominates). A 5 GB collection needs roughly 6-7 GB of RAM.
- Reopen of a 5 GB log replays in ~10 s (ReleaseFast); every committed write survives kill -9.
Comparing against real MongoDB: compare.js + compare-run.sh
bash tests/e2e/compare-run.sh [size] [doc-size] # e.g. 1g 16k
Starts mongod (brew install mongodb-community) on :27018 and MultiforaDB
on :27019, runs the same driver workload against each (durable writes:
MultiforaDB fsyncs per command, mongod runs with j: true), measures kill -9
reopen for both, and prints a side-by-side table. compare.js alone runs
one side (see its --help-style header comment).
Iteration-to-iteration comparison: bench-run.sh + concurrent.js
bash tests/e2e/bench-run.sh [size] [doc-size] ["clients..."] # e.g. 1g 16k "1 4 8 16 32"
Runs the main suite (compare-run.sh) plus a concurrent durable-write
comparison (concurrent.js, N clients each doing sequential insertOne
with {w:1, j:true} — the group-commit path under real contention), then
writes a machine-readable, versioned report to
tests/e2e/results/bench-<timestamp>.txt and prints a diff of the
MultiforaDB numbers against the previous run (results/bench-latest.txt).
The report has [main] / [concurrency] / [meta] sections with
name<TAB>value rows; bench-run.sh 1g 16k reproduces the phase8 gate
(see results/phase8.txt).