As an open-source project, seaweedfs/seaweed-up has picked up 32 stars on GitHub (Go). bootstrap SeaweedFS over SSH
Snapshot summary built from the project's own GitHub metadata — there's no written TopGit review yet. The page will update automatically when a full review is published.
WHY NO REVIEW YET
TopGit writes full reviews for the most-starred, most-requested repositories. This page is a snapshot until then — see the READ ME tab for the original README in full.
git clone https://github.com/seaweedfs/seaweed-up.git
cd seaweed-up
go install
Cluster topology
A typical SeaweedFS deployment has three layers, plus an optional monitoring
stack. seaweed-up knows how to deploy and operate all of them from a single
YAML file.
Layer
Components
Role
Storage
master_servers
Raft quorum, volume topology (run 3)
volume_servers
Store data needles (scale for capacity)
File access
filer_servers
Metadata gateway (run 3, external DB)
s3_servers
S3 API gateway (scale for throughput)
sftp_servers, envoy_servers
Optional additional frontends
vector_servers
S3 Vectors engine for vector buckets
Backend ops
admin_servers
Coordinate balancing, EC, vacuum (run 1)
worker_servers
Execute admin-scheduled tasks (run a few)
Observability
monitoring
node_exporter + Prometheus + Grafana
The admin server, workers, and monitoring stack are not on the data path —
restarting them does not interrupt reads or writes. Monitoring is optional; see
Monitoring below.
Configuration examples
Two example specs live under examples/:
examples/minimum.yaml — single-host all-in-one,
useful for local smoke tests.
examples/typical.yaml — production-shaped
topology with 3 masters, 3 volume servers, 3 filers (PostgreSQL metadata),
2 S3 gateways, 1 admin, 2 workers, and a co-located Prometheus + Grafana
monitoring stack.
Cluster name
Set a top-level cluster_name in every spec that has filer_servers,
admin_servers, or worker_servers (both example specs above set one):
cluster_name: my-cluster
The filer needs a jwt.filer_signing key to register the IAM gRPC service
the Admin UI Users tab calls, and the admin server signs Bearer tokens with
that same key — this is required even with TLS off. The admin server also
requires it from every worker that connects, so each admin and worker
instance gets a copy in its own config dir (<dir.conf>/<instance>.d/security.toml).
seaweed-up generates
that key on first deploy and persists it locally under
~/.seaweed-up/clusters/<cluster_name>/ so later runs (redeploys, scale
out/in, upgrades) reuse it instead of rotating it and locking out existing
tokens. Without a cluster_name there's nowhere stable to persist it, so
cluster deploy fails fast with cluster name is required to persist jwt.filer_signing key rather than silently generating a throwaway one.
cluster_name also doubles as the <cluster-name> argument to cluster status / upgrade / scale / destroy, and can be supplied as a
positional argument to cluster deploy instead of in the YAML — e.g.
seaweed-up cluster deploy my-cluster -f cluster.yaml (the positional
argument, when given, overrides cluster_name in the file).
Shared config defaults
Cluster-wide values are written once and inherited by each entry, so you
don't repeat them per server:
global.s3_config — the s3.json (IAM identities) every s3_servers
entry uses. Define it once under global; the secret lives in one place.
s3_servers[].filer — defaults to the first filer_servers entry.
worker_servers[].admin — defaults to the first admin_servers entry.
worker_servers[].masters — the Lance maintenance worker's masters, whose
filers it keeps Lance tables through; defaults to every master_servers entry.
So a multi-gateway / multi-worker spec is just a list of IPs:
When the cluster nodes only have private addresses reachable through a
public jump host (the classic ssh bastion then ssh 10.0.0.x pattern),
declare the bastion under global.bastion. Every connection — deploy,
preflight, lifecycle, upgrade, prepare — is tunnelled through it, so you
can run seaweed-up from your laptop:
global:
bastion:
host: 203.0.113.10 # public jump host; "host" or "host:port"
user: chris # optional, defaults to the current OS user
identity: ~/.ssh/id_rsa # optional, falls back to the ssh agent
port: 22 # optional, defaults to 22
The master_servers / volume_servers IPs are then the private addresses
as seen from the bastion. The --user / --identity flags still apply
to the nodes themselves; the bastion: block holds the jump host's own
credentials. Omit the block for direct connections.
SSH host-key verification
By default seaweed-up does not verify SSH host keys. To enforce a real
trust boundary (against a man-in-the-middle on the path to the bastion or
the nodes), set global.ssh_host_key_check:
A volume server can run the standalone Rust weed-volume binary instead of
the Go weed volume. Both register with the same masters, read the same
options file, and share the on-disk format, so they coexist in one cluster —
set engine per volume server:
engine accepts go/weed (default) or rust/weed-volume. For the rust
engine, seaweed-up downloads weed-volume_large_disk_linux_<arch>.tar.gz (plus
.md5) from the release for --version — the asset published by SeaweedFS's
rust_binaries_release.yml. With --enterprise it pulls
weed-volume-enterprise_large_disk_linux_<arch>.tar.gz from the artifactory
release instead, so the rust engine works for both editions.
ignore (default) — no verification; preserves historical behavior.
accept-new — trust-on-first-use: unknown hosts are learned and
appended to ~/.ssh/known_hosts, but a host whose key has changed is
rejected. Good for first deploys.
strict — every host (bastion and nodes) must already be in
~/.ssh/known_hosts; anything else is refused.
The policy applies to both the direct node connections and the bastion hop.
Vector buckets
The S3 gateway serves S3 Vectors on its own port; index and vector operations
need the engine, weed-vector, which vector_servers deploys. It registers
with the masters (where the gateway finds it) and keeps the indexes' files on
the filers and volume servers, so it needs no S3 credentials:
vector_servers:
- ip: 10.0.0.51 # gRPC on 10.0.0.51:8090; masters from master_servers
metrics_port: 9329 # opt in with an engine build that exposes metrics
config:
read-cache-mb: 1024 # further weed-vector flags by name
port defaults to 8090; the engine listens on ip:port, the address it
registers, so the gateway must reach it. masters defaults to every
master_servers entry.
seaweed-up downloads weed-vector_linux_<arch>.tar.gz (plus .md5) from the
release for --version; linux amd64/arm64 only. Use a SeaweedFS release after
4.52, whose engine and Lance worker keep tables on the filers. The enterprise
releases ship no weed-vector asset. With enable_tls, vector hosts receive
certificates and security.toml for mutual TLS to the gateway, masters,
filers, and volume servers. Without TLS
the engine serves any caller that reaches its port, bypassing the gateway's
IAM checks, and host prep opens 8090 to all sources: keep vector hosts
on the cluster's private network.
Vector indexes are built by the Lance maintenance worker that runs beside
every worker_servers entry: it gets the masters (worker_servers[].masters)
and the cluster's signing key, which the S3 gateway checks it by.
Vector metrics are opt-in for compatibility with older engine releases. When metrics_port is set, seaweed-up binds it to the vector host IP, adds a Prometheus scrape target, and checks /ready after upgrades. metrics_ip overrides the bind address. Monitoring also installs node_exporter on vector hosts.
Monitoring (Prometheus + Grafana)
Declare a monitoring: block and cluster deploy will stand up the full
observability stack as part of the cluster: node_exporter on every master/volume/filer host,
Prometheus and Grafana on the monitoring host, the SeaweedFS metrics ports
auto-enabled on master/volume/filer, and the bundled SeaweedFS dashboard
pre-loaded against a provisioned Prometheus datasource.
seaweed-up cluster deploy -f cluster.yaml # whole cluster + monitoring
seaweed-up cluster deploy -f cluster.yaml --component=monitoring # just the stack
With bind: 127.0.0.1 (the default) Grafana isn't exposed publicly — reach
it over a tunnel:
ssh -L 3000:localhost:3000 chris@<monitoring-host> # then open http://localhost:3000
(Grafana binds to 127.0.0.1 on the monitoring host, so the tunnel must
terminate there; add -J chris@<bastion> if the host is only reachable
through a jump host.)
Monitoring participates in the lifecycle commands too:
The lower-level building blocks remain available if you run your own
Prometheus/Grafana: cluster prometheus-config, cluster node-exporter install, and cluster dashboard install.
Metrics ports are assigned automatically when monitoring is enabled: each
master/volume/filer gets one (unique per host, starting at 9324), weed
is started with -metricsPort so it serves /metrics, and the scrape config
points at the same ports. To pin a specific port — e.g. for fixed firewall
rules — set metrics_port: on a master_servers / volume_servers /
filer_servers entry; explicit values are kept as-is.
Lifecycle
Start, stop, or restart every service in the cluster, or scope the operation
to a single component:
# Prepare hosts (ulimits, sysctls, firewall, time sync)
seaweed-up cluster prepare -f cluster.yaml
# Check cluster status
seaweed-up cluster status my-cluster
# Rolling upgrade
seaweed-up cluster upgrade my-cluster -f cluster.yaml --version=latest
# Scale out / in
seaweed-up cluster scale out my-cluster --add-volume=2
seaweed-up cluster scale in my-cluster --remove-node=10.0.0.23
# List and destroy
seaweed-up cluster list
seaweed-up cluster destroy my-cluster --remove-data
Erasure coding
With enough volume servers, enable erasure coding for better durability at
lower storage overhead. For example, with 8 volume servers you can run EC
5+3 (5 data + 3 parity shards) and tolerate up to 3 volume server failures
at 1.6x overhead, versus 3x for full replication.
No homepage URL was recorded for seaweedfs/seaweed-up in TopGit's last sync. The README tab above frequently contains screenshots and demo links, or check the repository description on GitHub.
Is seaweedfs/seaweed-up open source?
Yes — seaweedfs/seaweed-up ships under the Apache-2.0 license, which makes its source code freely readable (and, depending on license terms, forkable and reusable). Source: github.com/seaweedfs/seaweed-up.
What is seaweedfs/seaweed-up?
seaweedfs/seaweed-up (seaweedfs/seaweed-up) is a Go project on GitHub. From the project's own README: bootstrap SeaweedFS over SSH
What license does seaweedfs/seaweed-up use?
seaweedfs/seaweed-up is released under the Apache-2.0 license. Always verify the LICENSE file directly on GitHub for the authoritative terms — license strings can be edited out of sync with a project's actual stance.
Where do I read more about seaweedfs/seaweed-up?
This TopGit page is a snapshot — the READ ME tab shows the project's own README content (links stripped, images preserved). The GitHub repository at github.com/seaweedfs/seaweed-up is the definitive source.
Read full README in the tab above.
Still deciding about seaweed-up?
One click hands the question to an AI along with this page — see what it says about seaweed-up.