Files
seaweedfs/seaweed-volume
bd953b0f84 volume server: group-commit fsync writes in the write queue (#11543)
* volume: split the write path into reusable steps

do_write_request ran its pre-append checks, the append, the sync
rollback, the index publish and the post-write bookkeeping inline, so a
batched write could only reuse it one needle at a time. Pull the steps
out (check_writable, prepare_write, undo_unsynced_append, publish_write,
finish_write) and the store's volume lookup plus disk-space check
(writable_volume_mut). do_write_request composes them in the same order
with the same early returns; no behaviour change.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

* volume server: group-commit fsync writes in the write queue

The write queue holds one store lock for a batch of up to 128 needles
but wrote them one at a time, so every fsync needle paid its own .dat
sync and its own .idx sync: 2N syncs per batch.

Add Volume::write_needles_grouped, after Go's processBatch. A volume's
entries are split into runs of distinct needle ids (a repeated id starts
a new run, so its dedup and cookie checks see the earlier write). A run
with a durable entry appends everything with append_at_ns chained
through a local, syncs the .dat once, and only then publishes the
entries and syncs the .idx once. A failed .dat sync truncates the .dat
back to the run start (marking the volume unavailable if that fails),
leaves last_append_at_ns and last_modified untouched, and fails every
entry of the run. Runs with no durable entry go through the unchanged
per-needle path. Store::write_volume_needles is the queue's entry point;
the handlers' non-queue path is unchanged.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

* volume server: fail closed when a failed append's rollback fails

append_needle discarded the truncate-back result, so a partial write
that could not be rolled back left unindexed bytes on the .dat while
the volume stayed writable; the next append would bury them mid-file,
past the load-time tail check. Route the rollback through
undo_unsynced_append, which marks the volume unavailable when the
truncate fails, so nothing more is appended over an unverified tail.

* volume server: keep a grouped run's I/O error streak from later appends

A synced run stages every append before any entry finishes, so the
success reset in finish_write ran after the failed appends queued
behind the last write to land and erased their media-error streak.
Sent one at a time, those errors would have counted and quarantined
the volume. Skip the reset when an append after the last landed write
added to the streak.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

* volume server: replay a grouped run's I/O error streak in queue order

Skipping the run's success reset whenever an append after the last
landed write failed kept the errors from before that write as well,
so a run like [EIO, EIO, landed, EIO] reached the quarantine count
that the same writes one at a time (one error) do not. Mark the
streak where each entry is staged and record the run's success at
the last landed write's mark: errors before it are cleared, the ones
after it still count.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

* volume server: replay a run's I/O error streak in one atomic step

Reads record their outcomes on the tracker without the volume's write
lock, so record_success_at's separate load and store could drop an
error a read counted in between, or restore errors a read had just
cleared. Keep the count and the clear counter in one atomic word and
apply the replay with a single fetch_update.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Co-authored-by: Chris Lu <chris.lu@gmail.com>
2026-10-01 20:59:40 +08:00
..

SeaweedFS Volume Server (Rust)

A drop-in replacement for the SeaweedFS Go volume server, rewritten in Rust. It uses binary-compatible storage formats (.dat, .idx, .vif) and speaks the same HTTP and gRPC protocols, so it works with an unmodified Go master server.

Building

Requires Rust 1.91.1+ (2024 edition), matching rust-version in Cargo.toml. The patch release matters: 1.91.0 does not build. The edition itself only needs 1.85; the higher floor comes from the dependency tree — chiefly the AWS SDK — so it moves with those crates. CI builds on the latest stable.

cd seaweed-volume
cargo build --release

The binary is produced at target/release/seaweed-volume.

Running

Start a Go master server first, then point the Rust volume server at it:

# Minimal
seaweed-volume --port 8080 --master localhost:9333 --dir /data/vol1 --max 7

# Multiple data directories
seaweed-volume --port 8080 --master localhost:9333 \
  --dir /mnt/ssd1,/mnt/ssd2 --max 100,100 --disk ssd

# With datacenter/rack topology
seaweed-volume --port 8080 --master localhost:9333 --dir /data/vol1 --max 7 \
  --dataCenter dc1 --rack rack1

# With JWT authentication
seaweed-volume --port 8080 --master localhost:9333 --dir /data/vol1 --max 7 \
  --securityFile /etc/seaweedfs/security.toml

# With TLS (configured in security.toml via [https.volume] and [grpc.volume] sections)
seaweed-volume --port 8080 --master localhost:9333 --dir /data/vol1 --max 7 \
  --securityFile /etc/seaweedfs/security.toml

Common flags

Flag Default Description
--port 8080 HTTP listen port
--port.grpc port+10000 gRPC listen port
--master localhost:9333 Comma-separated master server addresses
--dir /tmp Comma-separated data directories
--max 8 Max volumes per directory (comma-separated)
--ip auto-detect Server IP / identifier
--ip.bind same as --ip Bind address
--dataCenter Datacenter name
--rack Rack name
--disk Disk type tag: hdd, ssd, or custom
--index memory Needle map type: memory, leveldb, leveldbMedium, leveldbLarge
--readMode proxy Non-local read mode: local, proxy, redirect
--fileSizeLimitMB 256 Max upload file size
--minFreeSpace 1 (percent) Min free disk space before marking volumes read-only
--securityFile Path to security.toml for JWT keys and TLS certs
--metricsPort 0 (disabled) Prometheus metrics endpoint port
--whiteList Comma-separated IPs with write permission
--preStopSeconds 10 Graceful drain period before shutdown
--compactionMBps 0 (unlimited) Compaction I/O rate limit
--pprof false Enable pprof HTTP handlers

Set RUST_LOG=debug (or trace, info, warn) for log level control. Set SEAWEED_WRITE_QUEUE=1 to enable batched async write processing.

Features

  • Binary compatible -- reads and writes the same .dat/.idx/.vif files as the Go server; seamless migration with no data conversion.
  • HTTP + gRPC -- full implementation of the volume server HTTP API and all gRPC RPCs including streaming operations (copy, tail, incremental copy, vacuum).
  • Master heartbeat -- bidirectional streaming heartbeat with the Go master server; volume and EC shard registration, leader failover, graceful shutdown deregistration.
  • JWT authentication -- signing key configuration via security.toml with token source precedence (query > header > cookie), file_id claims validation, and separate read/write keys.
  • TLS -- HTTPS for the HTTP API and mTLS for gRPC, configured through security.toml.
  • Erasure coding -- Reed-Solomon EC shard management: mount/unmount, read, rebuild, copy, delete, and shard-to-volume reconstruction.
  • S3 remote storage -- FetchAndWriteNeedle reads from any S3-compatible backend (AWS, MinIO, Wasabi, Backblaze, etc.) and writes locally. Supports VolumeTierMoveDatToRemote/FromRemote for tiered storage.
  • Needle map backends -- in-memory HashMap, LevelDB (via rusty-leveldb), or redb (pure Rust disk-backed) needle maps.
  • Image processing -- on-the-fly resize/crop, JPEG EXIF orientation auto-fix, WebP support.
  • Streaming reads -- large files (>1MB) are streamed via spawn_blocking to avoid blocking the async runtime.
  • Auto-compression -- compressible file types (text, JSON, CSS, JS, SVG, etc.) are gzip-compressed on upload.
  • Prometheus metrics -- counters, histograms, and gauges exported at a dedicated metrics port; optional push gateway support.
  • Graceful shutdown -- SIGINT/SIGTERM handling with configurable preStopSeconds drain period.

Testing

Rust unit tests

cd seaweed-volume
cargo test

Go integration tests

The Go test suite can target either the Go or Rust volume server via the VOLUME_SERVER_IMPL environment variable:

# Run all HTTP + gRPC integration tests against the Rust server
VOLUME_SERVER_IMPL=rust go test -v -count=1 -timeout 1200s \
  ./test/volume_server/grpc/... ./test/volume_server/http/...

# Run a single test
VOLUME_SERVER_IMPL=rust go test -v -count=1 -timeout 60s \
  -run "TestName" ./test/volume_server/http/...

# Run S3 remote storage tests
VOLUME_SERVER_IMPL=rust go test -v -count=1 -timeout 180s \
  -run "TestFetchAndWriteNeedle" ./test/volume_server/grpc/...

Load testing

A load test harness is available at test/volume_server/loadtest/. See that directory for usage instructions and scenarios.

Architecture

The server runs three listeners concurrently:

  • HTTP (Axum 0.7) -- admin and public routers for file upload/download, status, and stats endpoints.
  • gRPC (Tonic 0.12) -- all VolumeServer RPCs from the SeaweedFS protobuf definition.
  • Metrics (optional) -- Prometheus scrape endpoint on a separate port.

Key source modules:

Path Description
src/main.rs Entry point, server startup, signal handling
src/config.rs CLI parsing and configuration resolution
src/server/volume_server.rs HTTP router setup and middleware
src/server/handlers.rs HTTP request handlers (read, write, delete, status)
src/server/grpc_server.rs gRPC service implementation
src/server/heartbeat.rs Master heartbeat loop
src/storage/volume.rs Volume read/write/delete logic
src/storage/needle.rs Needle (file entry) serialization
src/storage/store.rs Multi-volume store management
src/security.rs JWT validation and IP whitelist guard
src/remote_storage/ S3 remote storage backend

See DEV_PLAN.md for the full development history and feature checklist.