Add Terraform support for VM-based SeaweedFS deployment (#9754)

* terraform: add cloud-agnostic core renderer module

Renders per-node weed argv, systemd units, config files, disk-mount and secret-fetch scripts, and cloud-init from an address map. Creates zero cloud resources. Flags verified against the weed binary: volume uses -mserver for the master list, gRPC is -port.grpc (auto http+10000), minFreeSpacePercent is a string, filer store via -defaultStoreDir.

* terraform: add mTLS and JWT security module

Generates the CA, per-component certs with distinct CNs, and JWT signing keys via the tls/random providers. Emits a core_security object plus PEMs for secret-store delivery.

* terraform: add AWS deployment module and examples

Reserves stable ENIs first, renders config via the core, then creates instances, prevent_destroy EBS data disks mounted at /data, and the cluster security group. With enable_security, generates certs/JWT, stores them in SSM SecureString, grants an instance role, and fetches them at boot so secrets stay out of user_data. Keyed for_each on every stateful tier.

* terraform: add local cluster test harnesses

run_local_cluster.sh and run_local_secure.sh render a cluster with the core and run real weed processes, asserting master quorum, volume registration, filer/s3 round-trips, mutual-TLS formation, and JWT enforcement. Use an isolated high port range with a guard so they never touch a cluster already running on the machine. The weed binary defaults to $(go env GOPATH)/bin/weed.

* terraform: add CI workflow and README

fmt/validate/tofu-test plus smoke jobs that build weed and run both harnesses.

* terraform: guard against empty filesystem UUID in mount script

An empty UUID made grep -q match any fstab line, skipping the fstab entry and breaking the mount. Fail fast when blkid returns no UUID.

* terraform: sanitize cluster name in WEED_CLUSTER env keys

Hyphens or spaces in cluster_name produced invalid systemd/bash env var names; map non-alphanumerics to underscores.

* terraform: omit empty jwt.signing block from security.toml

With enable_security and no JWT key, the template emitted [jwt.signing] key="". Gate the block on a non-empty key and cover it with a test.

* terraform: mark core security input as sensitive

The security object carries JWT signing keys; keep them out of plan output and known values.

* terraform: enforce jwt_length minimum of 32

* terraform: note region/AZ coupling in HA example

* terraform: guard WORKDIR before recursive delete in test harnesses

* terraform: fix README fence language and test count

* terraform: handle embedded s3 with no filer nodes

Indexing sort(keys(var.filers))[0] errored at plan time when embedded S3 was enabled but no filers were defined; fall back to an empty config source.

* terraform: scope kms:Decrypt to a configurable key arn

Replace the hardcoded Resource="*" with a kms_key_arn variable (default "*") so production can restrict decrypt to a specific CMK.

* terraform: encrypt EBS data volumes at rest

Set encrypted = true on the volume/filer data disks and the all-in-one example disk.

* terraform: protect filer instances from API termination

Filers hold the leveldb2 metadata store, so they are stateful and get the same disable_api_termination as masters and volumes.

* terraform: stop instance before detaching in all-in-one example

* terraform: drop stale references to the removed plan doc

* terraform: correct stale mount-step comment in aws module

* terraform: mark Terraform support as experimental in README
This commit is contained in:
Chris Lu
2026-05-30 23:43:17 -07:00
committed by GitHub
parent 0e35235908
commit a10607f90a
29 changed files with 2993 additions and 0 deletions
+80
View File
@@ -0,0 +1,80 @@
name: "terraform: validate and test modules"
on:
push:
branches: [ master ]
paths: ['terraform/**', '.github/workflows/terraform_ci.yml']
pull_request:
branches: [ master ]
paths: ['terraform/**', '.github/workflows/terraform_ci.yml']
permissions:
contents: read
jobs:
validate:
name: fmt, validate, plan-level tests
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v5
- name: Set up OpenTofu
uses: opentofu/setup-opentofu@v1
with:
tofu_version: 1.12.1
- name: fmt check
working-directory: terraform
run: tofu fmt -recursive -check -diff
- name: validate core
working-directory: terraform/modules/core
run: |
tofu init -backend=false -input=false
tofu validate
- name: validate security
working-directory: terraform/modules/security
run: |
tofu init -backend=false -input=false
tofu validate
- name: plan-level tests (core)
working-directory: terraform/modules/core
run: tofu test
- name: validate examples
run: |
set -e
for ex in terraform/examples/*/; do
echo "== validate $ex =="
tofu -chdir="$ex" init -backend=false -input=false
tofu -chdir="$ex" validate
done
smoke:
name: local cluster smoke test (real weed)
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v5
- name: Set up Go
uses: actions/setup-go@v6
with:
go-version-file: go.mod
- name: Build weed
run: go build -o "$RUNNER_TEMP/weed" ./weed
- name: Set up OpenTofu
uses: opentofu/setup-opentofu@v1
with:
tofu_version: 1.12.1
- name: Run local cluster harness
working-directory: terraform/test/local
run: WEED="$RUNNER_TEMP/weed" ./run_local_cluster.sh
- name: Run local mTLS cluster harness
working-directory: terraform/test/local-secure
run: WEED="$RUNNER_TEMP/weed" ./run_local_secure.sh
+15
View File
@@ -0,0 +1,15 @@
# Terraform / OpenTofu working files
.terraform/
.terraform.lock.hcl
*.tfstate
*.tfstate.*
crash.log
crash.*.log
*.tfvars
!*.tfvars.example
override.tf
override.tf.json
*_override.tf
*_override.tf.json
.terraformrc
terraform.rc
+128
View File
@@ -0,0 +1,128 @@
# SeaweedFS on Terraform
> **Experimental.** This Terraform support is an early scaffold under active
> development. Interfaces (variables, outputs, module layout) may change without
> notice, and not every tier is implemented yet. Not recommended for production
> without your own review and testing.
Self-contained Terraform/OpenTofu modules to deploy SeaweedFS on cloud VMs
running the `weed` binary directly under systemd. No Helm and no Kubernetes
required.
What works today is verified end-to-end against a real `weed` cluster (see
"Test it locally" below).
## Layout
```text
terraform/
modules/
core/ cloud-agnostic renderer (ZERO cloud resources): turns an
address map + config into per-node weed argv, systemd units,
config files, disk-mount + secret-fetch scripts, and cloud-init.
Both the cloud wrappers and the local harnesses consume `nodes`.
security/ cloud-agnostic CA + per-component mTLS certs (distinct CNs) +
JWT signing keys (tls/random providers). Emits a `core_security`
object ready for the core, plus PEMs for secret-store delivery.
aws/ thin AWS wrapper: reserves stable ENIs (fixed private IPs)
first, feeds them to core, creates instances + protected EBS
data disks + SG; with enable_security it generates certs/JWT,
stores them in SSM SecureString, grants an instance role, and
renders a boot fetch-secrets.sh. Keyed for_each throughout.
examples/
aws-ha-distributed/ 3 masters + 3 volumes (1/AZ) + 2 filers + 1 S3,
secure-by-default (mTLS via SSM-delivered certs)
aws-all-in-one/ single `weed server` instance via core directly
test/
local/ render a cluster with core and run it as real weed processes
on 127.0.0.1 (no cloud, no docker), then assert it works
local-secure/ generate certs/JWT with the security module, run a real
mTLS cluster, and assert it forms + enforces JWT auth
```
## Design in one paragraph
The chart is the structural reference, not a dependency. A cloud-agnostic
**core** renders everything portable; thin per-cloud wrappers provision infra.
Addressing is an **input** to the core (wrapper reserves static IPs first), so
the wrapper -> core dependency is one-way with no apply-time cycle. Stateful
tiers (master/volume/filer) are keyed `for_each` maps, never `count`, so a
middle node can be replaced without reindexing its peers or reattaching the
wrong disk. Flag names are verified against the real `weed` binary
(notably: volume uses `-mserver` for the master list; gRPC is `-port.grpc`,
auto = http+10000; `minFreeSpacePercent` is a string).
## Test it locally
Requires `tofu` (or `terraform`), `jq`, `curl`, and a `weed` binary. Renders a
3-master + volume + filer + S3 cluster from the core module and runs it as real
processes, asserting quorum, volume registration, and filer/S3 round-trips:
```bash
cd terraform/test/local
WEED=/path/to/weed ./run_local_cluster.sh
# => 7 passed, 0 failed
```
The harness uses a high port range (29333/28080/28888/28333) so it does not
collide with a SeaweedFS cluster already running on the machine, and aborts if a
required port is taken. `KEEP=1 ./run_local_cluster.sh` leaves the cluster up.
### mTLS end-to-end
```bash
cd terraform/test/local-secure
WEED=/path/to/weed ./run_local_secure.sh
# => generates a CA + component certs + JWT, renders security.toml, runs a real
# mTLS cluster, asserts master/volume/filer form over mutual TLS and that the
# filer enforces JWT signing (unsigned writes get 401). 5 passed.
```
## Plan-level tests (no cloud)
```bash
cd terraform/modules/core && tofu test
# => 11 passed: peers list, -mserver vs -master, metrics gating,
# security.toml conditions, all-in-one inheritance, ...
```
## Validate the cloud wrappers
```bash
cd terraform/examples/aws-ha-distributed && tofu init && tofu validate
```
`apply` needs AWS credentials, a VPC, subnets, and an AMI with `weed` installed
(bake with Packer, or install at boot).
## Status
Implemented and verified:
- Tiers: master / volume / filer / s3 / all-in-one rendered by the core.
- Disk mount: cloud-init runs a `mount-disks.sh` that auto-discovers (or takes
explicit candidate devices), `blkid`-guards `mkfs`, mounts, and persists to
`/etc/fstab` by UUID. The AWS wrapper wires the protected EBS disk to `/data`.
- mTLS + JWT: the `security` module generates the CA + per-component certs
(distinct CNs) + JWT keys; the AWS wrapper stores them in SSM SecureString,
grants an instance role, and the core renders a boot fetch script so secrets
stay OUT of user_data. Proven end-to-end by `test/local-secure`.
Not yet implemented:
- sftp / admin / worker tiers (core does not render them yet).
- GCP/Azure wrappers, the data-plane provider, and the K8s-native module.
- CA in Vault PKI (currently TF-generated, so the CA key lives in state). KMS
decrypt in the IAM policy is scoped by
`ViaService` but uses `Resource="*"`; tighten to the SSM CMK in production.
## Gotchas
- This `weed` build starts an Iceberg REST catalog on `:8181` by default; set
`s3.iceberg_port = 0` to disable, or a port to relocate it.
- Disk auto-discovery assumes a single attached data disk per node. For multiple
data disks, pass explicit `devices` candidates in `disk_mounts`.
## Security note
`tofu output`/state can contain secrets. Generate secrets outside Terraform
(cloud secret manager / Vault) and fetch them at boot; never commit secret
material.
+133
View File
@@ -0,0 +1,133 @@
# SeaweedFS all-in-one on a single AWS instance.
#
# Demonstrates the layered design directly: call the cloud-agnostic core to
# render cloud-init for a `weed server` node, then attach it to one instance
# with a protected data disk. No wrapper module needed for the simple case.
#
# tofu init && tofu validate
# tofu apply # requires AWS credentials + a weed AMI
terraform {
required_version = ">= 1.3.0"
required_providers {
aws = {
source = "hashicorp/aws"
version = ">= 5.40"
}
}
}
provider "aws" {
region = var.region
}
variable "region" {
type = string
default = "us-east-1"
}
variable "vpc_id" {
type = string
}
variable "subnet_id" {
type = string
}
variable "availability_zone" {
type = string
default = "us-east-1a"
}
variable "ami_id" {
type = string
}
variable "private_ip" {
type = string
default = "10.0.1.50"
}
module "core" {
source = "../../modules/core"
weed_binary = "/usr/bin/weed"
# disable the distributed tiers; run a single all-in-one process
master = { enabled = false }
volume = { enabled = false }
filer = { enabled = false }
all_in_one = {
enabled = true
nodes = { a0 = { address = var.private_ip, data_dir = "/data" } }
s3 = { enabled = true }
}
s3_identities = [{
name = "anonymous"
access_key = ""
secret_key = ""
actions = ["Read", "List"]
}]
}
resource "aws_network_interface" "aio" {
subnet_id = var.subnet_id
private_ips = [var.private_ip]
security_groups = [aws_security_group.aio.id]
tags = { Name = "seaweedfs-all-in-one" }
}
resource "aws_security_group" "aio" {
name_prefix = "seaweedfs-aio-"
vpc_id = var.vpc_id
lifecycle {
create_before_destroy = true
}
}
resource "aws_vpc_security_group_ingress_rule" "s3" {
security_group_id = aws_security_group.aio.id
cidr_ipv4 = "10.0.0.0/8"
ip_protocol = "tcp"
from_port = 8333
to_port = 8333
}
resource "aws_vpc_security_group_egress_rule" "all" {
security_group_id = aws_security_group.aio.id
cidr_ipv4 = "0.0.0.0/0"
ip_protocol = "-1"
}
resource "aws_instance" "aio" {
ami = var.ami_id
instance_type = "m5.large"
user_data = module.core.cloud_init_by_node["all-in-one-a0"]
network_interface {
network_interface_id = aws_network_interface.aio.id
device_index = 0
}
metadata_options {
http_tokens = "required"
}
tags = { Name = "seaweedfs-all-in-one", Role = "all-in-one" }
}
resource "aws_ebs_volume" "data" {
availability_zone = var.availability_zone
size = 200
type = "gp3"
encrypted = true
tags = { Name = "seaweedfs-all-in-one-data" }
lifecycle {
prevent_destroy = true
}
}
resource "aws_volume_attachment" "data" {
device_name = "/dev/xvdf"
volume_id = aws_ebs_volume.data.id
instance_id = aws_instance.aio.id
stop_instance_before_detaching = true
}
output "instance_id" {
value = aws_instance.aio.id
}
@@ -0,0 +1,117 @@
# SeaweedFS HA on AWS: 3-master quorum + 3 volume servers (one per AZ) + 2
# filers (leveldb2-replicated HA) + 1 standalone S3 gateway.
#
# tofu init && tofu validate
# tofu apply # requires AWS credentials, a VPC, subnets, and a weed AMI
#
# This is a scaffold: it provisions instances, protected EBS data disks, and the
# security group. Mounting the EBS disk at /data and secret-store cert delivery
# are documented follow-ups (see terraform/README.md).
terraform {
required_version = ">= 1.3.0"
required_providers {
aws = {
source = "hashicorp/aws"
version = ">= 5.40"
}
}
}
provider "aws" {
region = var.region
}
variable "region" {
type = string
default = "us-east-1"
}
variable "vpc_id" {
type = string
}
variable "ami_id" {
description = "AMI with the weed binary at /usr/bin/weed."
type = string
}
# subnet per AZ
variable "subnet_a" { type = string }
variable "subnet_b" { type = string }
variable "subnet_c" { type = string }
# AZ defaults assume region us-east-1. When you change `region`, override az_a/az_b/az_c
# (and the matching subnets) with AZs that belong to that region, or the volume/filer
# EBS volumes will fail to create in a mismatched AZ.
variable "az_a" {
type = string
default = "us-east-1a"
}
variable "az_b" {
type = string
default = "us-east-1b"
}
variable "az_c" {
type = string
default = "us-east-1c"
}
variable "client_ingress_cidrs" {
type = list(string)
default = ["10.0.0.0/8"]
}
variable "ssh_ingress_cidrs" {
type = list(string)
default = []
}
module "seaweedfs" {
source = "../../modules/aws"
name = "seaweedfs"
vpc_id = var.vpc_id
ami_id = var.ami_id
# secure-by-default flagship: mTLS (certs + JWT generated by the security
# submodule, delivered via SSM and fetched at boot) plus monitoring.
enable_security = true
monitoring_enabled = true
masters = {
m0 = { subnet_id = var.subnet_a, private_ip = "10.0.1.10" }
m1 = { subnet_id = var.subnet_b, private_ip = "10.0.2.10" }
m2 = { subnet_id = var.subnet_c, private_ip = "10.0.3.10" }
}
volumes = {
v0 = { subnet_id = var.subnet_a, availability_zone = var.az_a, private_ip = "10.0.1.20", rack = var.az_a, data_center = var.region, data_volume_size_gb = 500 }
v1 = { subnet_id = var.subnet_b, availability_zone = var.az_b, private_ip = "10.0.2.20", rack = var.az_b, data_center = var.region, data_volume_size_gb = 500 }
v2 = { subnet_id = var.subnet_c, availability_zone = var.az_c, private_ip = "10.0.3.20", rack = var.az_c, data_center = var.region, data_volume_size_gb = 500 }
}
filers = {
f0 = { subnet_id = var.subnet_a, availability_zone = var.az_a, private_ip = "10.0.1.30" }
f1 = { subnet_id = var.subnet_b, availability_zone = var.az_b, private_ip = "10.0.2.30" }
}
s3_nodes = {
s0 = { subnet_id = var.subnet_a, private_ip = "10.0.1.40" }
}
client_ingress_cidrs = var.client_ingress_cidrs
ssh_ingress_cidrs = var.ssh_ingress_cidrs
}
output "master_peers" {
value = module.seaweedfs.master_peers
}
output "instance_ids" {
value = module.seaweedfs.instance_ids
}
output "security_group_id" {
value = module.seaweedfs.security_group_id
}
+205
View File
@@ -0,0 +1,205 @@
# ---- step 1: reserve stable addressing (ENIs with fixed private IPs) -------
# These exist BEFORE the core renders config, so the address map is known and
# the core <- wrapper dependency is one-way (no apply-time cycle).
resource "aws_network_interface" "master" {
for_each = var.masters
subnet_id = each.value.subnet_id
private_ips = [each.value.private_ip]
security_groups = [aws_security_group.cluster.id]
tags = merge(var.tags, { Name = "${var.name}-master-${each.key}", Role = "master" })
}
resource "aws_network_interface" "volume" {
for_each = var.volumes
subnet_id = each.value.subnet_id
private_ips = [each.value.private_ip]
security_groups = [aws_security_group.cluster.id]
tags = merge(var.tags, { Name = "${var.name}-volume-${each.key}", Role = "volume" })
}
resource "aws_network_interface" "filer" {
for_each = var.filers
subnet_id = each.value.subnet_id
private_ips = [each.value.private_ip]
security_groups = [aws_security_group.cluster.id]
tags = merge(var.tags, { Name = "${var.name}-filer-${each.key}", Role = "filer" })
}
resource "aws_network_interface" "s3" {
for_each = var.s3_nodes
subnet_id = each.value.subnet_id
private_ips = [each.value.private_ip]
security_groups = [aws_security_group.cluster.id]
tags = merge(var.tags, { Name = "${var.name}-s3-${each.key}", Role = "s3" })
}
# ---- step 2: render config with the cloud-agnostic core --------------------
module "core" {
source = "../core"
weed_binary = var.weed_binary
monitoring_enabled = var.monitoring_enabled
enable_security = var.enable_security
security = local.core_security
# Keep secrets out of cloud-init user_data when security is on; deliver them
# from SSM via the boot fetch script instead.
render_secret_files = !var.enable_security
boot_fetch_script = local.fetch_script
master = {
nodes = { for k, v in var.masters : k => { address = v.private_ip } }
}
volume = {
enabled = length(var.volumes) > 0
nodes = { for k, v in var.volumes : k => {
address = v.private_ip
rack = v.rack
data_center = v.data_center
data_dirs = [{ path = "/data", max_volumes = 0 }]
# auto-discover the single attached data disk and mount it at /data
disk_mounts = [{ mountpoint = "/data", fstype = "xfs" }]
} }
}
filer = {
enabled = length(var.filers) > 0
nodes = { for k, v in var.filers : k => {
address = v.private_ip
data_dir = "/data/filerldb2"
disk_mounts = [{ mountpoint = "/data", fstype = "xfs" }]
} }
s3 = var.embedded_s3
}
s3 = {
enabled = length(var.s3_nodes) > 0
nodes = { for k, v in var.s3_nodes : k => { address = v.private_ip } }
}
s3_identities = var.s3_identities
}
# ---- step 3: instances (keyed for_each; user_data = rendered cloud-init) ----
resource "aws_instance" "master" {
for_each = var.masters
ami = var.ami_id
instance_type = each.value.instance_type
key_name = var.key_name
user_data = module.core.cloud_init_by_node["master-${each.key}"]
iam_instance_profile = var.enable_security ? aws_iam_instance_profile.node[0].name : null
disable_api_termination = var.termination_protection
network_interface {
network_interface_id = aws_network_interface.master[each.key].id
device_index = 0
}
metadata_options {
http_tokens = "required" # IMDSv2
}
tags = merge(var.tags, { Name = "${var.name}-master-${each.key}", Role = "master" })
}
resource "aws_instance" "volume" {
for_each = var.volumes
ami = var.ami_id
instance_type = each.value.instance_type
key_name = var.key_name
user_data = module.core.cloud_init_by_node["volume-${each.key}"]
iam_instance_profile = var.enable_security ? aws_iam_instance_profile.node[0].name : null
disable_api_termination = var.termination_protection
network_interface {
network_interface_id = aws_network_interface.volume[each.key].id
device_index = 0
}
metadata_options {
http_tokens = "required"
}
tags = merge(var.tags, { Name = "${var.name}-volume-${each.key}", Role = "volume" })
}
resource "aws_instance" "filer" {
for_each = var.filers
ami = var.ami_id
instance_type = each.value.instance_type
key_name = var.key_name
user_data = module.core.cloud_init_by_node["filer-${each.key}"]
iam_instance_profile = var.enable_security ? aws_iam_instance_profile.node[0].name : null
disable_api_termination = var.termination_protection
network_interface {
network_interface_id = aws_network_interface.filer[each.key].id
device_index = 0
}
metadata_options {
http_tokens = "required"
}
tags = merge(var.tags, { Name = "${var.name}-filer-${each.key}", Role = "filer" })
}
resource "aws_instance" "s3" {
for_each = var.s3_nodes
ami = var.ami_id
instance_type = each.value.instance_type
key_name = var.key_name
user_data = module.core.cloud_init_by_node["s3-${each.key}"]
iam_instance_profile = var.enable_security ? aws_iam_instance_profile.node[0].name : null
network_interface {
network_interface_id = aws_network_interface.s3[each.key].id
device_index = 0
}
metadata_options {
http_tokens = "required"
}
tags = merge(var.tags, { Name = "${var.name}-s3-${each.key}", Role = "s3" })
}
# ---- step 4: protected data disks (decoupled; survive instance replacement) -
# The core's mount-disks.sh (wired via disk_mounts above) mkfs+mounts these at
# /data before the weed unit starts.
resource "aws_ebs_volume" "volume_data" {
for_each = var.volumes
availability_zone = each.value.availability_zone
size = each.value.data_volume_size_gb
type = each.value.data_volume_type
iops = each.value.data_volume_iops
encrypted = true
tags = merge(var.tags, { Name = "${var.name}-volume-${each.key}-data", Role = "volume" })
lifecycle {
prevent_destroy = true # break-glass: remove to allow destroy
}
}
resource "aws_volume_attachment" "volume_data" {
for_each = var.volumes
device_name = "/dev/xvdf"
volume_id = aws_ebs_volume.volume_data[each.key].id
instance_id = aws_instance.volume[each.key].id
stop_instance_before_detaching = true
}
resource "aws_ebs_volume" "filer_data" {
for_each = var.filers
availability_zone = each.value.availability_zone
size = each.value.data_volume_size_gb
type = "gp3"
encrypted = true
tags = merge(var.tags, { Name = "${var.name}-filer-${each.key}-data", Role = "filer" })
lifecycle {
prevent_destroy = true
}
}
resource "aws_volume_attachment" "filer_data" {
for_each = var.filers
device_name = "/dev/xvdf"
volume_id = aws_ebs_volume.filer_data[each.key].id
instance_id = aws_instance.filer[each.key].id
stop_instance_before_detaching = true
}
+34
View File
@@ -0,0 +1,34 @@
output "security_group_id" {
description = "Cluster security group id."
value = aws_security_group.cluster.id
}
output "master_private_ips" {
description = "Master private IPs keyed by node id."
value = { for k, v in var.masters : k => v.private_ip }
}
output "master_peers" {
description = "Master peer list passed to the weed processes."
value = module.core.master_peers
}
output "filer_private_ips" {
description = "Filer private IPs keyed by node id."
value = { for k, v in var.filers : k => v.private_ip }
}
output "s3_private_ips" {
description = "Standalone S3 gateway private IPs keyed by node id."
value = { for k, v in var.s3_nodes : k => v.private_ip }
}
output "instance_ids" {
description = "All instance ids keyed by role-node."
value = merge(
{ for k, v in aws_instance.master : "master-${k}" => v.id },
{ for k, v in aws_instance.volume : "volume-${k}" => v.id },
{ for k, v in aws_instance.filer : "filer-${k}" => v.id },
{ for k, v in aws_instance.s3 : "s3-${k}" => v.id },
)
}
+136
View File
@@ -0,0 +1,136 @@
# =============================================================================
# mTLS + JWT generation and secret delivery (active when enable_security=true).
#
# Generates the CA + per-component certs + JWT keys with the security submodule,
# stores them (plus the rendered security.toml / S3 config) in SSM Parameter
# Store as SecureString, and renders a boot fetch-secrets.sh that pulls them via
# the instance role. This keeps secrets OUT of cloud-init user_data / IMDS.
# Secrets still live in Terraform state (TF-generated); prefer Vault PKI for the
# CA in production.
# =============================================================================
locals {
all_private_ips = concat(
[for k, v in var.masters : v.private_ip],
[for k, v in var.volumes : v.private_ip],
[for k, v in var.filers : v.private_ip],
[for k, v in var.s3_nodes : v.private_ip],
)
ssm_prefix = "/seaweedfs/${var.name}"
deliver_s3_config = length(var.s3_nodes) > 0 || var.embedded_s3.enabled
# security.toml is identical across nodes; take it from the first master.
first_master = var.enable_security ? sort(keys(var.masters))[0] : ""
security_toml_path = "/etc/seaweedfs/security.toml"
s3_config_path = "/etc/seaweedfs/s3_config.json"
# Boot fetch entries: cert files + security.toml + (optional) S3 config.
fetch_entries = var.enable_security ? concat(
[for m in module.security[0].secret_manifest : { param = "${local.ssm_prefix}/certs/${m.key}", path = m.path, mode = m.mode }],
[{ param = "${local.ssm_prefix}/security.toml", path = local.security_toml_path, mode = "0600" }],
local.deliver_s3_config ? [{ param = "${local.ssm_prefix}/s3_config.json", path = local.s3_config_path, mode = "0600" }] : [],
) : []
fetch_script = var.enable_security ? templatefile("${path.module}/templates/fetch-secrets.sh.tftpl", {
run_as_user = "seaweedfs"
entries = local.fetch_entries
}) : ""
core_security = var.enable_security ? module.security[0].core_security : var.security
}
module "security" {
count = var.enable_security ? 1 : 0
source = "../security"
internal_domain = var.internal_domain
cert_dir = var.cert_dir
ip_sans = local.all_private_ips
}
# ---- SSM SecureString parameters -------------------------------------------
resource "aws_ssm_parameter" "cert" {
for_each = var.enable_security ? { for m in module.security[0].secret_manifest : m.key => m } : {}
name = "${local.ssm_prefix}/certs/${each.key}"
type = "SecureString"
value = module.security[0].secret_contents[each.key]
tags = var.tags
}
resource "aws_ssm_parameter" "security_toml" {
count = var.enable_security ? 1 : 0
name = "${local.ssm_prefix}/security.toml"
type = "SecureString"
value = module.core.secret_files_by_node["master-${local.first_master}"][local.security_toml_path]
tags = var.tags
}
resource "aws_ssm_parameter" "s3_config" {
count = var.enable_security && local.deliver_s3_config ? 1 : 0
name = "${local.ssm_prefix}/s3_config.json"
type = "SecureString"
# identical content across s3/filer nodes; render once via the security module's
# JSON is not available here, so read it back from any node that carries it.
value = local.s3_config_source
tags = var.tags
}
locals {
# Pull the rendered S3 identity JSON from whichever node carries it.
s3_config_source = !local.deliver_s3_config ? "" : (
length(var.s3_nodes) > 0 ?
module.core.secret_files_by_node["s3-${sort(keys(var.s3_nodes))[0]}"][local.s3_config_path] :
(length(var.filers) > 0 ? module.core.secret_files_by_node["filer-${sort(keys(var.filers))[0]}"][local.s3_config_path] : "")
)
}
# ---- IAM: instance role allowed to read the SSM params ----------------------
data "aws_iam_policy_document" "assume" {
count = var.enable_security ? 1 : 0
statement {
actions = ["sts:AssumeRole"]
principals {
type = "Service"
identifiers = ["ec2.amazonaws.com"]
}
}
}
data "aws_iam_policy_document" "ssm_read" {
count = var.enable_security ? 1 : 0
statement {
sid = "ReadSeaweedfsParams"
actions = ["ssm:GetParameter", "ssm:GetParameters"]
resources = ["arn:aws:ssm:*:*:parameter${local.ssm_prefix}/*"]
}
statement {
sid = "DecryptSecureStrings"
actions = ["kms:Decrypt"]
resources = [var.kms_key_arn] # defaults to "*"; set a CMK ARN in production
condition {
test = "StringEquals"
variable = "kms:ViaService"
values = ["ssm.*.amazonaws.com"]
}
}
}
resource "aws_iam_role" "node" {
count = var.enable_security ? 1 : 0
name_prefix = "${var.name}-node-"
assume_role_policy = data.aws_iam_policy_document.assume[0].json
tags = var.tags
}
resource "aws_iam_role_policy" "ssm_read" {
count = var.enable_security ? 1 : 0
name = "ssm-read"
role = aws_iam_role.node[0].id
policy = data.aws_iam_policy_document.ssm_read[0].json
}
resource "aws_iam_instance_profile" "node" {
count = var.enable_security ? 1 : 0
name_prefix = "${var.name}-node-"
role = aws_iam_role.node[0].name
}
+62
View File
@@ -0,0 +1,62 @@
# Cluster security group + the SeaweedFS port matrix.
# Intra-cluster: all TCP allowed within the SG (master 9333/19333, volume
# 8080/18080, filer 8888/18888, admin 33646, ...). Client-facing: only S3 (8333)
# and filer (8888). Metrics ports (9327) are never opened to clients.
resource "aws_security_group" "cluster" {
name_prefix = "${var.name}-cluster-"
description = "SeaweedFS cluster"
vpc_id = var.vpc_id
tags = merge(var.tags, { Name = "${var.name}-cluster" })
lifecycle {
create_before_destroy = true
}
}
# All intra-cluster TCP via SG self-reference.
resource "aws_vpc_security_group_ingress_rule" "intra_cluster" {
security_group_id = aws_security_group.cluster.id
referenced_security_group_id = aws_security_group.cluster.id
ip_protocol = "tcp"
from_port = 0
to_port = 65535
description = "intra-cluster (http + gRPC for all roles)"
}
resource "aws_vpc_security_group_ingress_rule" "s3" {
for_each = toset(var.client_ingress_cidrs)
security_group_id = aws_security_group.cluster.id
cidr_ipv4 = each.value
ip_protocol = "tcp"
from_port = 8333
to_port = 8333
description = "S3 gateway"
}
resource "aws_vpc_security_group_ingress_rule" "filer" {
for_each = toset(var.client_ingress_cidrs)
security_group_id = aws_security_group.cluster.id
cidr_ipv4 = each.value
ip_protocol = "tcp"
from_port = 8888
to_port = 8888
description = "filer HTTP"
}
resource "aws_vpc_security_group_ingress_rule" "ssh" {
for_each = toset(var.ssh_ingress_cidrs)
security_group_id = aws_security_group.cluster.id
cidr_ipv4 = each.value
ip_protocol = "tcp"
from_port = 22
to_port = 22
description = "SSH"
}
resource "aws_vpc_security_group_egress_rule" "all" {
security_group_id = aws_security_group.cluster.id
cidr_ipv4 = "0.0.0.0/0"
ip_protocol = "-1"
description = "all egress"
}
@@ -0,0 +1,21 @@
#!/usr/bin/env bash
# Rendered by terraform-aws-seaweedfs. Pulls certs + secret config from SSM
# Parameter Store (SecureString) at boot using the instance role. Region is
# auto-resolved by the AWS CLI from IMDS, so no region is baked in. Requires
# the AWS CLI and a working instance profile (see security.tf).
set -euo pipefail
run_as_user="${run_as_user}"
fetch() {
local param="$1" dest="$2" mode="$3"
mkdir -p "$(dirname "$dest")"
aws ssm get-parameter --with-decryption --name "$param" \
--query 'Parameter.Value' --output text > "$dest"
chmod "$mode" "$dest"
chown "$run_as_user:$run_as_user" "$dest" 2>/dev/null || true
}
%{ for e in entries ~}
fetch "${e.param}" "${e.path}" "${e.mode}"
%{ endfor ~}
+171
View File
@@ -0,0 +1,171 @@
# =============================================================================
# terraform-aws-seaweedfs - thin AWS wrapper around the cloud-agnostic core.
#
# Reserves stable per-node addressing (ENIs with fixed private IPs) FIRST, feeds
# that address map to the core to render cloud-init, then creates instances and
# protected EBS data disks. All tiers are keyed for_each (never count), so a
# middle node can be replaced without reindexing its peers/disks.
# =============================================================================
variable "name" {
description = "Name prefix for all resources."
type = string
default = "seaweedfs"
}
variable "vpc_id" {
description = "VPC to deploy into."
type = string
}
variable "ami_id" {
description = "AMI with the weed binary installed (e.g. baked with Packer)."
type = string
}
variable "weed_binary" {
description = "Path to weed on the AMI."
type = string
default = "/usr/bin/weed"
}
variable "key_name" {
description = "EC2 key pair for SSH (optional)."
type = string
default = null
}
variable "monitoring_enabled" {
description = "Bind metrics ports and pass them through to the core."
type = bool
default = false
}
variable "enable_security" {
description = "Enable mTLS (security.toml). Cert material must be supplied via the security variable / secret store."
type = bool
default = false
}
variable "security" {
description = "Passed through to the core security variable when enable_security=false. When enable_security=true this is ignored and certs/JWT are generated by the security submodule."
type = any
default = {}
}
variable "internal_domain" {
description = "Internal domain for generated component cert CNs / peer-auth wildcard (enable_security=true)."
type = string
default = "seaweedfs.internal"
}
variable "cert_dir" {
description = "On-host directory where certs are fetched/placed."
type = string
default = "/usr/local/share/ca-certificates"
}
variable "kms_key_arn" {
description = "KMS key ARN the instance role may use to decrypt SSM SecureStrings. Defaults to \"*\" (the SSM default key); set to a specific CMK ARN in production."
type = string
default = "*"
}
variable "termination_protection" {
description = "Set disable_api_termination on stateful instances (masters, volumes)."
type = bool
default = true
}
# ---- stateful tiers: keyed node maps ---------------------------------------
variable "masters" {
description = "Master nodes keyed by stable id (m0/m1/m2). Quorum size must be odd."
type = map(object({
subnet_id = string
private_ip = string
instance_type = optional(string, "t3.medium")
}))
validation {
condition = length(var.masters) % 2 == 1
error_message = "Master quorum must be an odd number of nodes (1, 3, 5)."
}
}
variable "volumes" {
description = "Volume nodes keyed by stable id. Each owns one protected EBS data disk."
type = map(object({
subnet_id = string
availability_zone = string
private_ip = string
instance_type = optional(string, "m5.large")
rack = optional(string)
data_center = optional(string)
data_volume_size_gb = optional(number, 100)
data_volume_type = optional(string, "gp3")
data_volume_iops = optional(number, null)
}))
default = {}
}
variable "filers" {
description = "Filer nodes keyed by stable id. Each owns a protected EBS disk for its leveldb2 store."
type = map(object({
subnet_id = string
availability_zone = string
private_ip = string
instance_type = optional(string, "t3.medium")
data_volume_size_gb = optional(number, 50)
}))
default = {}
}
variable "s3_nodes" {
description = "Standalone S3 gateway nodes keyed by stable id (stateless)."
type = map(object({
subnet_id = string
private_ip = string
instance_type = optional(string, "t3.medium")
}))
default = {}
}
variable "embedded_s3" {
description = "Embed the S3 gateway in the filer instead of standalone nodes."
type = object({
enabled = optional(bool, false)
port = optional(number, 8333)
domain_name = optional(string, "")
})
default = {}
}
variable "s3_identities" {
description = "Non-admin S3 identities for the gateway config (admin key goes via EnvironmentFile)."
type = list(object({
name = string
access_key = string
secret_key = string
actions = list(string)
}))
default = []
sensitive = true
}
# ---- networking ------------------------------------------------------------
variable "client_ingress_cidrs" {
description = "CIDRs allowed to reach the client-facing S3 (8333) and filer (8888) ports."
type = list(string)
default = []
}
variable "ssh_ingress_cidrs" {
description = "CIDRs allowed SSH (22)."
type = list(string)
default = []
}
variable "tags" {
description = "Extra tags applied to all resources."
type = map(string)
default = {}
}
+12
View File
@@ -0,0 +1,12 @@
terraform {
required_version = ">= 1.3.0"
required_providers {
aws = {
source = "hashicorp/aws"
# >= 5.40 for the modern aws_vpc_security_group_*_rule resources.
# Provider >= 6.0 is recommended for the EBS-attachment no-replace behavior;
# bump the floor here once you standardize on it.
version = ">= 5.40"
}
}
}
+373
View File
@@ -0,0 +1,373 @@
# =============================================================================
# SeaweedFS Terraform core - pure renderer (zero cloud resources).
#
# Computes, per node: the verified `weed` argv, the full systemd ExecStart,
# config files, environment, and rendered systemd unit + cloud-init.
# =============================================================================
locals {
# ---- shared security.toml (rendered when mTLS on OR any JWT key set) ----
jwt_any = anytrue([
var.security.jwt_signing_key != "",
var.security.jwt_signing_read_key != "",
var.security.jwt_filer_signing_key != "",
var.security.jwt_filer_signing_read_key != "",
])
render_security = var.enable_security || local.jwt_any
security_toml = local.render_security ? templatefile("${path.module}/templates/security.toml.tftpl", {
enable_security = var.enable_security
cert_dir = var.security.cert_dir
allowed_wildcard_domain = var.security.allowed_wildcard_domain
allowed_common_names = var.security.allowed_common_names
jwt_signing_key = var.security.jwt_signing_key
jwt_signing_read_key = var.security.jwt_signing_read_key
jwt_filer_signing_key = var.security.jwt_filer_signing_key
jwt_filer_signing_read_key = var.security.jwt_filer_signing_read_key
}) : ""
security_files = local.render_security ? { "/etc/seaweedfs/security.toml" = local.security_toml } : {}
# ---- S3 identity JSON (non-admin identities only) ----
# Empty access_key => anonymous identity (no credentials block), per SeaweedFS
# convention where the identity literally named "anonymous" grants anon access.
s3_config_json = jsonencode({
identities = [for id in var.s3_identities : {
name = id.name
credentials = id.access_key != "" ? [{ accessKey = id.access_key, secretKey = id.secret_key }] : []
actions = id.actions
}]
})
# ---- global flags (before the subcommand) ----
global_flags = compact([
format("-v=%d", var.log_level),
var.log_to_stderr ? "-logtostderr=true" : "",
])
# ---- master peers (sorted by key for determinism) ----
master_keys = sort(keys(var.master.nodes))
master_peer_list = [for k in local.master_keys :
format("%s:%d", var.master.nodes[k].address, coalesce(var.master.nodes[k].port, var.master.port))
]
master_peers = join(",", local.master_peer_list)
# ---- first filer (default s3 target + cluster env) ----
filer_keys = sort(keys(var.filer.nodes))
first_filer = length(local.filer_keys) > 0 ? format("%s:%d", var.filer.nodes[local.filer_keys[0]].address, coalesce(var.filer.nodes[local.filer_keys[0]].port, var.filer.port)) : ""
# ---- cluster discovery env (WEED_CLUSTER_*) ----
cluster_env = merge(
{ WEED_CLUSTER_DEFAULT = var.cluster_name },
local.master_peers != "" ? { "WEED_CLUSTER_${replace(upper(var.cluster_name), "/[^A-Z0-9_]/", "_")}_MASTER" = local.master_peers } : {},
local.first_filer != "" ? { "WEED_CLUSTER_${replace(upper(var.cluster_name), "/[^A-Z0-9_]/", "_")}_FILER" = local.first_filer } : {},
)
metrics_flags_master = var.monitoring_enabled ? compact([
format("-metricsPort=%d", var.master.metrics_port),
var.master.metrics_ip != "" ? format("-metricsIp=%s", var.master.metrics_ip) : "",
]) : []
metrics_flags_volume = var.monitoring_enabled ? compact([
format("-metricsPort=%d", var.volume.metrics_port),
var.volume.metrics_ip != "" ? format("-metricsIp=%s", var.volume.metrics_ip) : "",
]) : []
metrics_flags_filer = var.monitoring_enabled ? compact([
format("-metricsPort=%d", var.filer.metrics_port),
var.filer.metrics_ip != "" ? format("-metricsIp=%s", var.filer.metrics_ip) : "",
]) : []
metrics_flags_s3 = var.monitoring_enabled ? [format("-metricsPort=%d", var.s3.metrics_port)] : []
# ===========================================================================
# MASTER nodes
# ===========================================================================
master_nodes = var.master.enabled ? {
for k in local.master_keys : "master-${k}" => {
role = "master"
name = "master-${k}"
address = var.master.nodes[k].address
disk_mounts = var.master.nodes[k].disk_mounts
data_dir = coalesce(var.master.nodes[k].data_dir, var.master.data_dir)
data_dirs = [coalesce(var.master.nodes[k].data_dir, var.master.data_dir)]
ports = {
http = coalesce(var.master.nodes[k].port, var.master.port)
metrics = var.master.metrics_port
}
env = local.cluster_env
config_files = local.security_files
argv = concat(
local.global_flags,
["master"],
[
format("-port=%d", coalesce(var.master.nodes[k].port, var.master.port)),
format("-mdir=%s", coalesce(var.master.nodes[k].data_dir, var.master.data_dir)),
format("-ip=%s", var.master.nodes[k].address),
format("-ip.bind=%s", var.master.ip_bind),
format("-peers=%s", local.master_peers),
format("-defaultReplication=%s", var.master.default_replication),
format("-volumeSizeLimitMB=%d", var.master.volume_size_limit_mb),
format("-electionTimeout=%s", var.master.election_timeout),
format("-heartbeatInterval=%s", var.master.heartbeat_interval),
],
var.master.nodes[k].grpc_port != null ? [format("-port.grpc=%d", var.master.nodes[k].grpc_port)] : (
var.master.grpc_port != null ? [format("-port.grpc=%d", var.master.grpc_port)] : []),
var.master.volume_preallocate ? ["-volumePreallocate"] : [],
var.master.garbage_threshold != null ? [format("-garbageThreshold=%s", tostring(var.master.garbage_threshold))] : [],
var.master.raft_hashicorp ? ["-raftHashicorp"] : [],
var.master.resume_state ? ["-resumeState"] : [],
var.master.disable_http ? ["-disableHttp"] : [],
var.master.white_list != "" ? [format("-whiteList=%s", var.master.white_list)] : [],
local.metrics_flags_master,
var.master.extra_args,
)
}
} : {}
# ===========================================================================
# VOLUME nodes
# ===========================================================================
volume_keys = sort(keys(var.volume.nodes))
volume_nodes = var.volume.enabled ? {
for k in local.volume_keys : "volume-${k}" => {
role = "volume"
name = "volume-${k}"
address = var.volume.nodes[k].address
disk_mounts = var.volume.nodes[k].disk_mounts
data_dirs = [for d in(var.volume.nodes[k].data_dirs != null ? var.volume.nodes[k].data_dirs : var.volume.data_dirs) : d.path]
data_dir = (var.volume.nodes[k].data_dirs != null ? var.volume.nodes[k].data_dirs : var.volume.data_dirs)[0].path
ports = {
http = coalesce(var.volume.nodes[k].port, var.volume.port)
metrics = var.volume.metrics_port
}
env = local.cluster_env
config_files = local.security_files
argv = concat(
local.global_flags,
["volume"],
[
format("-port=%d", coalesce(var.volume.nodes[k].port, var.volume.port)),
format("-dir=%s", join(",", [for d in(var.volume.nodes[k].data_dirs != null ? var.volume.nodes[k].data_dirs : var.volume.data_dirs) : d.path])),
format("-max=%s", join(",", [for d in(var.volume.nodes[k].data_dirs != null ? var.volume.nodes[k].data_dirs : var.volume.data_dirs) : tostring(d.max_volumes)])),
format("-ip=%s", var.volume.nodes[k].address),
format("-ip.bind=%s", var.volume.ip_bind),
format("-mserver=%s", local.master_peers),
format("-readMode=%s", var.volume.read_mode),
format("-compactionMBps=%d", var.volume.compaction_mbps),
format("-minFreeSpacePercent=%s", var.volume.min_free_space_percent),
format("-index=%s", var.volume.index),
],
var.volume.nodes[k].grpc_port != null ? [format("-port.grpc=%d", var.volume.nodes[k].grpc_port)] : [],
(var.volume.nodes[k].idx_dir != null && var.volume.nodes[k].idx_dir != "") ? [format("-dir.idx=%s", var.volume.nodes[k].idx_dir)] : (var.volume.idx_dir != "" ? [format("-dir.idx=%s", var.volume.idx_dir)] : []),
var.volume.nodes[k].data_center != null ? [format("-dataCenter=%s", var.volume.nodes[k].data_center)] : [],
var.volume.nodes[k].rack != null ? [format("-rack=%s", var.volume.nodes[k].rack)] : [],
var.volume.nodes[k].public_url != null ? [format("-publicUrl=%s", var.volume.nodes[k].public_url)] : [],
var.volume.file_size_limit_mb != null ? [format("-fileSizeLimitMB=%d", var.volume.file_size_limit_mb)] : [],
var.volume.images_fix_orientation ? ["-images.fix.orientation"] : [],
var.volume.white_list != "" ? [format("-whiteList=%s", var.volume.white_list)] : [],
local.metrics_flags_volume,
var.volume.extra_args,
)
}
} : {}
# ===========================================================================
# FILER nodes (+ optional embedded S3)
# ===========================================================================
filer_s3_files = var.filer.s3.enabled ? { (var.filer.s3.config_path) = local.s3_config_json } : {}
filer_nodes = var.filer.enabled ? {
for k in local.filer_keys : "filer-${k}" => {
role = "filer"
name = "filer-${k}"
address = var.filer.nodes[k].address
disk_mounts = var.filer.nodes[k].disk_mounts
data_dir = coalesce(var.filer.nodes[k].data_dir, var.filer.data_dir)
data_dirs = [coalesce(var.filer.nodes[k].data_dir, var.filer.data_dir)]
ports = {
http = coalesce(var.filer.nodes[k].port, var.filer.port)
metrics = var.filer.metrics_port
}
env = merge(local.cluster_env, var.filer.extra_env)
config_files = merge(local.security_files, local.filer_s3_files)
argv = concat(
local.global_flags,
["filer"],
[
format("-port=%d", coalesce(var.filer.nodes[k].port, var.filer.port)),
format("-ip=%s", var.filer.nodes[k].address),
format("-ip.bind=%s", var.filer.ip_bind),
format("-master=%s", local.master_peers),
format("-defaultReplicaPlacement=%s", var.filer.default_replica_placement),
format("-dirListLimit=%d", var.filer.dir_list_limit),
format("-defaultStoreDir=%s", coalesce(var.filer.nodes[k].data_dir, var.filer.data_dir)),
],
var.filer.nodes[k].grpc_port != null ? [format("-port.grpc=%d", var.filer.nodes[k].grpc_port)] : [],
var.filer.max_mb != null ? [format("-maxMB=%d", var.filer.max_mb)] : [],
var.filer.disable_dir_listing ? ["-disableDirListing"] : [],
var.filer.disable_http ? ["-disableHttp"] : [],
var.filer.encrypt_volume_data ? ["-encryptVolumeData"] : [],
var.filer.rack != "" ? [format("-rack=%s", var.filer.rack)] : [],
var.filer.data_center != "" ? [format("-dataCenter=%s", var.filer.data_center)] : [],
var.filer.filer_group != "" ? [format("-filerGroup=%s", var.filer.filer_group)] : [],
var.filer.s3.enabled ? [
"-s3",
format("-s3.port=%d", var.filer.s3.port),
format("-s3.config=%s", var.filer.s3.config_path),
] : [],
(var.filer.s3.enabled && var.filer.s3.https_port > 0) ? [format("-s3.port.https=%d", var.filer.s3.https_port)] : [],
(var.filer.s3.enabled && var.filer.s3.domain_name != "") ? [format("-s3.domainName=%s", var.filer.s3.domain_name)] : [],
local.metrics_flags_filer,
var.filer.extra_args,
)
}
} : {}
# ===========================================================================
# S3 standalone nodes
# ===========================================================================
s3_keys = sort(keys(var.s3.nodes))
s3_target = var.s3.filer_address != "" ? var.s3.filer_address : local.first_filer
s3_std_files = merge(local.security_files, { (var.s3.config_path) = local.s3_config_json })
s3_nodes = var.s3.enabled ? {
for k in local.s3_keys : "s3-${k}" => {
role = "s3"
name = "s3-${k}"
address = var.s3.nodes[k].address
disk_mounts = []
data_dir = ""
data_dirs = []
ports = {
http = coalesce(var.s3.nodes[k].port, var.s3.port)
metrics = var.s3.metrics_port
}
env = local.cluster_env
config_files = local.s3_std_files
argv = concat(
local.global_flags,
["s3"],
[
format("-port=%d", coalesce(var.s3.nodes[k].port, var.s3.port)),
format("-ip.bind=%s", var.s3.ip_bind),
format("-filer=%s", local.s3_target),
format("-config=%s", var.s3.config_path),
],
var.s3.https_port > 0 ? [format("-port.https=%d", var.s3.https_port)] : [],
var.s3.iceberg_port != null ? [format("-port.iceberg=%d", var.s3.iceberg_port)] : [],
var.s3.domain_name != "" ? [format("-domainName=%s", var.s3.domain_name)] : [],
var.s3.audit_log_config_path != "" ? [format("-auditLogConfig=%s", var.s3.audit_log_config_path)] : [],
var.s3.cert_file != "" ? [format("-cert.file=%s", var.s3.cert_file)] : [],
var.s3.key_file != "" ? [format("-key.file=%s", var.s3.key_file)] : [],
var.s3.cacert_file != "" ? [format("-cacert.file=%s", var.s3.cacert_file)] : [],
var.s3.verify_client_cert ? ["-tlsVerifyClientCert"] : [],
local.metrics_flags_s3,
var.s3.extra_args,
)
}
} : {}
# ===========================================================================
# ALL-IN-ONE nodes (weed server)
# ===========================================================================
aio_keys = sort(keys(var.all_in_one.nodes))
aio_s3_files = var.all_in_one.s3.enabled ? { (var.all_in_one.s3.config_path) = local.s3_config_json } : {}
aio_nodes = var.all_in_one.enabled ? {
for k in local.aio_keys : "all-in-one-${k}" => {
role = "server"
name = "all-in-one-${k}"
address = var.all_in_one.nodes[k].address
disk_mounts = var.all_in_one.nodes[k].disk_mounts
data_dir = coalesce(var.all_in_one.nodes[k].data_dir, var.all_in_one.data_dir)
data_dirs = [coalesce(var.all_in_one.nodes[k].data_dir, var.all_in_one.data_dir)]
ports = {
http = var.all_in_one.master_port
metrics = var.all_in_one.metrics_port
}
env = local.cluster_env
config_files = merge(local.security_files, local.aio_s3_files)
argv = concat(
local.global_flags,
["server"],
[
format("-dir=%s", coalesce(var.all_in_one.nodes[k].data_dir, var.all_in_one.data_dir)),
format("-ip=%s", var.all_in_one.nodes[k].address),
format("-ip.bind=%s", var.all_in_one.ip_bind),
"-master",
format("-master.port=%d", var.all_in_one.master_port),
format("-master.peers=%s:%d", var.all_in_one.nodes[k].address, var.all_in_one.master_port),
format("-master.defaultReplication=%s", var.all_in_one.default_replication),
format("-master.volumeSizeLimitMB=%d", var.all_in_one.volume_size_limit_mb),
"-volume",
format("-volume.port=%d", var.all_in_one.volume_port),
"-filer",
format("-filer.port=%d", var.all_in_one.filer_port),
format("-idleTimeout=%d", var.all_in_one.idle_timeout),
],
var.all_in_one.disable_http ? [format("-disableHttp=%t", var.all_in_one.disable_http)] : [],
var.all_in_one.s3.enabled ? [
"-s3",
format("-s3.port=%d", var.all_in_one.s3.port),
format("-s3.config=%s", var.all_in_one.s3.config_path),
] : [],
(var.all_in_one.s3.enabled && var.all_in_one.s3.domain_name != "") ? [format("-s3.domainName=%s", var.all_in_one.s3.domain_name)] : [],
var.monitoring_enabled ? [format("-metricsPort=%d", var.all_in_one.metrics_port)] : [],
var.all_in_one.extra_args,
)
}
} : {}
# ===========================================================================
# Merge + render systemd unit and cloud-init per node
# ===========================================================================
nodes_base = merge(local.master_nodes, local.volume_nodes, local.filer_nodes, local.s3_nodes, local.aio_nodes)
# Disk-mount script per node (empty unless the node declares disk_mounts).
mount_scripts = {
for name, n in local.nodes_base : name => length(n.disk_mounts) > 0 ? templatefile("${path.module}/templates/mount-disks.sh.tftpl", {
run_as_user = var.hardening.run_as_user
mounts = n.disk_mounts
}) : ""
}
nodes = {
for name, n in local.nodes_base : name => merge(n, {
exec_start = "${var.weed_binary} ${join(" ", n.argv)}"
# secret_files: the secret-bearing config a wrapper should deliver from a
# secret store when render_secret_files=false (else they go in cloud-init).
secret_files = n.config_files
mount_script = local.mount_scripts[name]
file_modes = { for p in keys(n.config_files) : p => "0600" }
systemd_unit = templatefile("${path.module}/templates/weed.service.tftpl", {
role = n.role
name = n.name
run_as_user = var.hardening.run_as_user
exec_start = "${var.weed_binary} ${join(" ", n.argv)}"
no_new_privileges = var.hardening.no_new_privileges
protect_system = var.hardening.protect_system
cap_drop_all = var.hardening.cap_drop_all
read_write_paths = n.data_dirs
requires_mounts_for = n.data_dirs
env_file = var.env_file
environment = n.env
})
cloud_init = templatefile("${path.module}/templates/cloud-init.yaml.tftpl", {
role = n.role
name = n.name
run_as_user = var.hardening.run_as_user
config_files = var.render_secret_files ? n.config_files : {}
file_modes = { for p in keys(n.config_files) : p => "0600" }
pre_runcmd = [for p in n.data_dirs : "install -d -o ${var.hardening.run_as_user} -g ${var.hardening.run_as_user} ${p}"]
mount_script = local.mount_scripts[name]
fetch_script = var.boot_fetch_script
systemd_unit = templatefile("${path.module}/templates/weed.service.tftpl", {
role = n.role
name = n.name
run_as_user = var.hardening.run_as_user
exec_start = "${var.weed_binary} ${join(" ", n.argv)}"
no_new_privileges = var.hardening.no_new_privileges
protect_system = var.hardening.protect_system
cap_drop_all = var.hardening.cap_drop_all
read_write_paths = n.data_dirs
requires_mounts_for = n.data_dirs
env_file = var.env_file
environment = n.env
})
})
})
}
}
+32
View File
@@ -0,0 +1,32 @@
output "nodes" {
description = "Per-node rendered artifacts keyed by node name. Each value: role, name, address, ports, data_dir(s), argv (list, after the weed binary), exec_start, env, config_files, systemd_unit, cloud_init."
value = local.nodes
sensitive = true # config_files/env may carry JWT keys or S3 identities
}
output "master_peers" {
description = "Comma-separated master peer list (address:port), as passed to -peers / -mserver / -master."
value = local.master_peers
}
output "first_filer" {
description = "address:port of the lexically-first filer node (default S3 target)."
value = local.first_filer
}
output "node_names" {
description = "List of rendered node names (non-sensitive)."
value = sort(keys(local.nodes))
}
output "cloud_init_by_node" {
description = "Map of node name -> rendered cloud-init user_data (for the per-cloud wrappers)."
value = { for name, n in local.nodes : name => n.cloud_init }
sensitive = true
}
output "secret_files_by_node" {
description = "Map of node name -> {path => content} of secret-bearing files (security.toml, S3 identity JSON). Wrappers push these to a secret store and fetch them at boot when render_secret_files=false."
value = { for name, n in local.nodes : name => n.secret_files }
sensitive = true
}
@@ -0,0 +1,44 @@
#cloud-config
# Rendered by the SeaweedFS Terraform core module for node ${name} (role ${role}).
# When render_secret_files=false, secret-bearing files are NOT here; they are
# fetched at boot by fetch-secrets.sh from the cloud secret store.
users:
- name: ${run_as_user}
system: true
shell: /usr/sbin/nologin
write_files:
%{ for path, content in config_files ~}
- path: ${path}
owner: "${run_as_user}:${run_as_user}"
permissions: '${lookup(file_modes, path, "0644")}'
content: |
${indent(6, content)}
%{ endfor ~}
%{ if mount_script != "" ~}
- path: /opt/seaweedfs/mount-disks.sh
permissions: '0755'
content: |
${indent(6, mount_script)}
%{ endif ~}
%{ if fetch_script != "" ~}
- path: /opt/seaweedfs/fetch-secrets.sh
permissions: '0755'
content: |
${indent(6, fetch_script)}
%{ endif ~}
- path: /etc/systemd/system/weed-${role}.service
permissions: '0644'
content: |
${indent(6, systemd_unit)}
runcmd:
%{ if mount_script != "" ~}
- /opt/seaweedfs/mount-disks.sh
%{ endif ~}
%{ if fetch_script != "" ~}
- /opt/seaweedfs/fetch-secrets.sh
%{ endif ~}
%{ for cmd in pre_runcmd ~}
- ${cmd}
%{ endfor ~}
- systemctl daemon-reload
- systemctl enable --now weed-${role}.service
@@ -0,0 +1,73 @@
#!/usr/bin/env bash
# Rendered by the SeaweedFS Terraform core module. Formats (only if blank) and
# mounts each data disk, persisting by UUID to /etc/fstab. Safe to re-run.
set -euo pipefail
run_as_user="${run_as_user}"
# Base block device backing "/" so auto-discovery never touches the root disk.
root_src="$(findmnt -no SOURCE / 2>/dev/null || true)"
root_base="$(lsblk -no pkname "$root_src" 2>/dev/null || true)"
[ -n "$root_base" ] && root_base="/dev/$root_base"
is_mounted() { findmnt -rno TARGET "$1" >/dev/null 2>&1; }
# Pick the first unmounted, non-root whole disk (single-data-disk node).
discover_dev() {
local d base
for d in /dev/nvme*n1 /dev/xvd[b-z] /dev/sd[b-z] /dev/vd[b-z]; do
[ -b "$d" ] || continue
[ "$d" = "$root_src" ] && continue
[ "$d" = "$root_base" ] && continue
base="/dev/$(lsblk -no pkname "$d" 2>/dev/null || true)"
[ "$base" = "$root_base" ] && continue
is_mounted "$d" && continue
echo "$d"; return 0
done
return 1
}
# Wait up to ~120s for one of the explicit candidate devices to appear.
wait_dev() {
local i d
for i in $(seq 1 60); do
for d in "$@"; do [ -b "$d" ] && { echo "$d"; return 0; }; done
sleep 2
done
return 1
}
mount_one() {
local mountpoint="$1" fstype="$2"; shift 2
local dev=""
if [ "$#" -gt 0 ]; then
dev="$(wait_dev "$@" || true)"
fi
if [ -z "$dev" ]; then
dev="$(discover_dev || true)"
fi
if [ -z "$dev" ]; then
echo "mount-disks: no device found for $mountpoint" >&2
return 1
fi
if ! blkid "$dev" >/dev/null 2>&1; then
echo "mount-disks: formatting $dev as $fstype"
"mkfs.$fstype" -q "$dev"
fi
mkdir -p "$mountpoint"
local uuid; uuid="$(blkid -s UUID -o value "$dev")"
if [ -z "$uuid" ]; then
echo "mount-disks: failed to read a filesystem UUID for $dev" >&2
return 1
fi
if ! grep -q "$uuid" /etc/fstab; then
printf 'UUID=%s %s %s defaults,nofail 0 2\n' "$uuid" "$mountpoint" "$fstype" >> /etc/fstab
fi
is_mounted "$mountpoint" || mount "$mountpoint"
chown -R "$run_as_user:$run_as_user" "$mountpoint"
echo "mount-disks: $dev mounted at $mountpoint"
}
%{ for m in mounts ~}
mount_one "${m.mountpoint}" "${m.fstype}"${length(m.devices) > 0 ? " ${join(" ", m.devices)}" : ""}
%{ endfor ~}
@@ -0,0 +1,53 @@
# SeaweedFS security.toml rendered by the Terraform core module.
# Rendered when enable_security OR any non-default JWT signing option is set.
# Presence of this file does NOT imply mTLS: the [grpc] TLS block is emitted
# only under enable_security; a JWT-only config carries [jwt.*] and no TLS.
%{ if enable_security ~}
[grpc]
ca = "${cert_dir}/ca/tls.crt"
# Peer-identity authorization: without an allow-list, mTLS collapses to
# "trusted the CA" with no authZ. Set allowed_wildcard_domain / commonNames.
%{ if allowed_wildcard_domain != "" ~}
allowed_wildcard_domain = "${allowed_wildcard_domain}"
%{ endif ~}
[grpc.master]
cert = "${cert_dir}/master/tls.crt"
key = "${cert_dir}/master/tls.key"
%{ if allowed_common_names != "" ~}
allowed_commonNames = "${allowed_common_names}"
%{ endif ~}
[grpc.volume]
cert = "${cert_dir}/volume/tls.crt"
key = "${cert_dir}/volume/tls.key"
[grpc.filer]
cert = "${cert_dir}/filer/tls.crt"
key = "${cert_dir}/filer/tls.key"
[grpc.client]
cert = "${cert_dir}/client/tls.crt"
key = "${cert_dir}/client/tls.key"
%{ endif ~}
%{ if jwt_signing_key != "" ~}
[jwt.signing]
key = "${jwt_signing_key}"
%{ endif ~}
%{ if jwt_signing_read_key != "" ~}
[jwt.signing.read]
key = "${jwt_signing_read_key}"
%{ endif ~}
%{ if jwt_filer_signing_key != "" ~}
[jwt.filer_signing]
key = "${jwt_filer_signing_key}"
%{ endif ~}
%{ if jwt_filer_signing_read_key != "" ~}
[jwt.filer_signing.read]
key = "${jwt_filer_signing_read_key}"
%{ endif ~}
@@ -0,0 +1,39 @@
[Unit]
Description=SeaweedFS ${role} (${name})
Documentation=https://github.com/seaweedfs/seaweedfs/wiki
After=network-online.target
Wants=network-online.target
%{ for path in requires_mounts_for ~}
RequiresMountsFor=${path}
%{ endfor ~}
[Service]
Type=simple
User=${run_as_user}
Group=${run_as_user}
ExecStart=${exec_start}
Restart=always
RestartSec=5
LimitNOFILE=1048576
TimeoutStopSec=60
# Hardening (OpenShift SCC analogs; relax via the hardening variable)
NoNewPrivileges=${no_new_privileges}
%{ if protect_system ~}
ProtectSystem=strict
ProtectHome=true
PrivateTmp=true
%{ for path in read_write_paths ~}
ReadWritePaths=${path}
%{ endfor ~}
%{ endif ~}
%{ if cap_drop_all ~}
CapabilityBoundingSet=
AmbientCapabilities=
%{ endif ~}
EnvironmentFile=-${env_file}
%{ for k, v in environment ~}
Environment="${k}=${v}"
%{ endfor ~}
[Install]
WantedBy=multi-user.target
@@ -0,0 +1,234 @@
# Plan-level tests for the core renderer. No cloud credentials required.
# cd terraform/modules/core && tofu test
variables {
weed_binary = "/usr/bin/weed"
}
# ---------------------------------------------------------------------------
run "master_peers_and_count" {
command = plan
variables {
master = {
nodes = {
m0 = { address = "10.0.0.10" }
m1 = { address = "10.0.0.11" }
m2 = { address = "10.0.0.12" }
}
}
volume = { enabled = false }
filer = { enabled = false }
}
assert {
condition = output.master_peers == "10.0.0.10:9333,10.0.0.11:9333,10.0.0.12:9333"
error_message = "master -peers list is wrong"
}
assert {
condition = length(output.node_names) == 3
error_message = "expected exactly 3 master nodes"
}
}
# ---------------------------------------------------------------------------
run "volume_uses_mserver_not_master" {
command = plan
variables {
master = { nodes = { m0 = { address = "10.0.0.10" } } }
volume = { nodes = { v0 = { address = "10.0.0.20", rack = "rack-a", data_center = "dc1" } } }
filer = { enabled = false }
}
assert {
condition = contains(output.nodes["volume-v0"].argv, "-mserver=10.0.0.10:9333")
error_message = "volume must pass the master list via -mserver (verified flag), not -master"
}
assert {
condition = !contains(output.nodes["volume-v0"].argv, "-master=10.0.0.10:9333")
error_message = "volume should not use -master for the master list"
}
assert {
condition = contains(output.nodes["volume-v0"].argv, "-rack=rack-a") && contains(output.nodes["volume-v0"].argv, "-dataCenter=dc1")
error_message = "per-node rack/dataCenter must render"
}
}
# ---------------------------------------------------------------------------
run "filer_uses_master_and_store_dir" {
command = plan
variables {
master = { nodes = { m0 = { address = "10.0.0.10" } } }
volume = { enabled = false }
filer = { nodes = { f0 = { address = "10.0.0.30", data_dir = "/srv/filer" } } }
}
assert {
condition = contains(output.nodes["filer-f0"].argv, "-master=10.0.0.10:9333")
error_message = "filer must use -master for the master list"
}
assert {
condition = contains(output.nodes["filer-f0"].argv, "-defaultStoreDir=/srv/filer")
error_message = "filer leveldb2 store dir must render via -defaultStoreDir"
}
}
# ---------------------------------------------------------------------------
run "metrics_gated_off_by_default" {
command = plan
variables {
master = { nodes = { m0 = { address = "10.0.0.10" } } }
volume = { enabled = false }
filer = { enabled = false }
}
assert {
condition = length([for a in output.nodes["master-m0"].argv : a if startswith(a, "-metricsPort")]) == 0
error_message = "metrics port must NOT be bound when monitoring is disabled"
}
}
run "metrics_on_when_enabled" {
command = plan
variables {
monitoring_enabled = true
master = { nodes = { m0 = { address = "10.0.0.10" } }, metrics_port = 9327 }
volume = { enabled = false }
filer = { enabled = false }
}
assert {
condition = contains(output.nodes["master-m0"].argv, "-metricsPort=9327")
error_message = "metrics port must bind when monitoring is enabled"
}
}
# ---------------------------------------------------------------------------
run "security_toml_off_by_default" {
command = plan
variables {
master = { nodes = { m0 = { address = "10.0.0.10" } } }
volume = { enabled = false }
filer = { enabled = false }
}
assert {
condition = length(keys(output.nodes["master-m0"].config_files)) == 0
error_message = "security.toml must not render when security is off and no JWT keys set"
}
}
run "security_toml_on_with_mtls" {
command = plan
variables {
enable_security = true
security = {
allowed_wildcard_domain = ".seaweedfs.internal"
jwt_signing_key = "test-signing-key-not-a-real-secret-0123456789"
}
master = { nodes = { m0 = { address = "10.0.0.10" } } }
volume = { enabled = false }
filer = { enabled = false }
}
assert {
condition = contains(keys(output.nodes["master-m0"].config_files), "/etc/seaweedfs/security.toml")
error_message = "security.toml must render when enable_security is true"
}
}
# ---------------------------------------------------------------------------
run "all_in_one_renders_server" {
command = plan
variables {
master = { enabled = false }
volume = { enabled = false }
filer = { enabled = false }
all_in_one = {
enabled = true
nodes = { a0 = { address = "10.0.0.40" } }
s3 = { enabled = true }
}
}
assert {
condition = contains(output.nodes["all-in-one-a0"].argv, "server")
error_message = "all-in-one must invoke the weed server subcommand"
}
assert {
condition = contains(output.nodes["all-in-one-a0"].argv, "-s3") && contains(output.nodes["all-in-one-a0"].argv, "-s3.port=8333")
error_message = "all-in-one S3 must render when enabled"
}
}
# ---------------------------------------------------------------------------
run "security_toml_mtls_only_no_jwt" {
command = plan
variables {
enable_security = true
master = { nodes = { m0 = { address = "10.0.0.10" } } }
volume = { enabled = false }
filer = { enabled = false }
}
assert {
condition = contains(keys(output.nodes["master-m0"].config_files), "/etc/seaweedfs/security.toml")
error_message = "security.toml must render for mTLS even without JWT keys"
}
assert {
condition = !strcontains(output.nodes["master-m0"].config_files["/etc/seaweedfs/security.toml"], "[jwt.signing]")
error_message = "the jwt.signing block must be omitted when no signing key is set"
}
}
run "mount_fetch_and_secret_delivery" {
command = plan
variables {
enable_security = true
render_secret_files = false
boot_fetch_script = "#!/usr/bin/env bash\necho fetch\n"
security = {
jwt_signing_key = "test-signing-key-not-a-real-secret-0123456789"
}
master = { nodes = { m0 = { address = "10.0.0.10", disk_mounts = [{ mountpoint = "/data", fstype = "xfs" }] } } }
volume = { enabled = false }
filer = { enabled = false }
}
assert {
condition = strcontains(output.nodes["master-m0"].cloud_init, "/opt/seaweedfs/mount-disks.sh")
error_message = "cloud-init must write+run the mount script when disk_mounts is set"
}
assert {
condition = strcontains(output.nodes["master-m0"].cloud_init, "/opt/seaweedfs/fetch-secrets.sh")
error_message = "cloud-init must write+run the fetch script when boot_fetch_script is set"
}
assert {
condition = !strcontains(output.nodes["master-m0"].cloud_init, "[jwt.signing]")
error_message = "security.toml must NOT be inlined in cloud-init when render_secret_files=false"
}
assert {
condition = contains(keys(output.nodes["master-m0"].secret_files), "/etc/seaweedfs/security.toml")
error_message = "secret_files must expose security.toml for secret-store delivery"
}
}
run "mount_script_absent_without_disks" {
command = plan
variables {
master = { nodes = { m0 = { address = "10.0.0.10" } } }
volume = { enabled = false }
filer = { enabled = false }
}
assert {
condition = !strcontains(output.nodes["master-m0"].cloud_init, "mount-disks.sh")
error_message = "no mount script should render when the node declares no disk_mounts"
}
}
run "s3_standalone_targets_filer" {
command = plan
variables {
master = { nodes = { m0 = { address = "10.0.0.10" } } }
volume = { enabled = false }
filer = { nodes = { f0 = { address = "10.0.0.30" } } }
s3 = { enabled = true, nodes = { s0 = { address = "10.0.0.50" } } }
}
assert {
condition = contains(output.nodes["s3-s0"].argv, "-filer=10.0.0.30:8888")
error_message = "standalone S3 must point -filer at the first filer node"
}
assert {
condition = contains(output.nodes["s3-s0"].argv, "-config=/etc/seaweedfs/s3_config.json")
error_message = "standalone S3 must reference the rendered config path"
}
}
+318
View File
@@ -0,0 +1,318 @@
# =============================================================================
# SeaweedFS Terraform core module - input variables
#
# This module is a PURE RENDERER. It creates zero cloud resources. It takes a
# per-node address map plus configuration and emits, per node, the rendered
# `weed` argv, systemd unit, cloud-init, and config files. Both the per-cloud
# infra wrappers (consume cloud_init/systemd_unit) and the local test harness
# (consume argv/config_files) read its `nodes` output.
#
# Flag names below are verified against the real `weed` binary, not the chart.
# Notably: volume uses -mserver for the master list; gRPC is -port.grpc
# (auto = http+10000); minFreeSpacePercent is a string.
# =============================================================================
variable "weed_binary" {
description = "Path to the weed executable on the target host."
type = string
default = "/usr/bin/weed"
}
variable "cluster_name" {
description = "Logical cluster name (WEED_CLUSTER_DEFAULT)."
type = string
default = "sw"
}
variable "log_level" {
description = "Global glog verbosity (-v)."
type = number
default = 1
}
variable "log_to_stderr" {
description = "Log to stderr (journald) instead of -logdir. Recommended on systemd."
type = bool
default = true
}
variable "monitoring_enabled" {
description = "Gate the -metricsPort/-metrics.address flags. When false the metrics port is not bound, matching the chart."
type = bool
default = false
}
variable "enable_security" {
description = "Emit the security.toml [grpc] mTLS block and reference per-component certs. Independent of JWT."
type = bool
default = false
}
# ----------------------------------------------------------------------------
# Security / secrets (the core only references paths + renders security.toml;
# cert/secret material is generated by the wrapper, never defaulted here).
# ----------------------------------------------------------------------------
variable "security" {
description = "mTLS + JWT signing configuration. Secret values must be supplied (never defaulted in module code)."
type = object({
cert_dir = optional(string, "/usr/local/share/ca-certificates")
allowed_wildcard_domain = optional(string, "")
allowed_common_names = optional(string, "")
jwt_signing_key = optional(string, "")
jwt_signing_read_key = optional(string, "")
jwt_filer_signing_key = optional(string, "")
jwt_filer_signing_read_key = optional(string, "")
})
default = {}
sensitive = true
}
variable "hardening" {
description = "systemd hardening directives (OpenShift SCC analogs). Relax per field if needed."
type = object({
run_as_user = optional(string, "seaweedfs")
no_new_privileges = optional(bool, true)
protect_system = optional(bool, true)
cap_drop_all = optional(bool, true)
})
default = {}
}
variable "env_file" {
description = "Path to the systemd EnvironmentFile holding secret env (DB creds, S3 admin key) fetched at boot. Optional (-prefixed in the unit, so a missing file is tolerated)."
type = string
default = "/etc/seaweedfs/weed.env"
}
variable "render_secret_files" {
description = <<-EOT
When true (default), secret-bearing config (security.toml, S3 identity JSON)
is written directly into cloud-init user_data. Convenient for the local test
harness and simple deployments. Set to false to keep secrets OUT of user_data
(and out of the cloud metadata service): the files are then exposed only via
the secret_files_by_node output, and the wrapper is expected to deliver them
from a secret store via boot_fetch_script.
EOT
type = bool
default = true
}
variable "boot_fetch_script" {
description = "Optional shell script written to /opt/seaweedfs/fetch-secrets.sh (root, 0755) and run before the weed unit starts. Used by wrappers to pull certs/secrets from a cloud secret store at boot."
type = string
default = ""
}
# ----------------------------------------------------------------------------
# Master (Raft quorum). Keyed `nodes` map -> for_each, never count.
# ----------------------------------------------------------------------------
variable "master" {
description = "Master tier. nodes is a map keyed by stable node id (m0/m1/m2)."
type = object({
enabled = optional(bool, true)
nodes = optional(map(object({
address = string
port = optional(number)
grpc_port = optional(number)
data_dir = optional(string)
disk_mounts = optional(list(object({
mountpoint = string
fstype = optional(string, "ext4")
devices = optional(list(string), [])
})), [])
})), {})
port = optional(number, 9333)
grpc_port = optional(number, null) # null => weed auto = port+10000
metrics_port = optional(number, 9327)
metrics_ip = optional(string, "")
ip_bind = optional(string, "0.0.0.0")
data_dir = optional(string, "/var/lib/seaweedfs/master")
default_replication = optional(string, "000")
volume_size_limit_mb = optional(number, 1000)
volume_preallocate = optional(bool, false)
garbage_threshold = optional(number, null)
disable_http = optional(bool, false)
raft_hashicorp = optional(bool, false)
resume_state = optional(bool, true) # deliberate deviation: binary default false
election_timeout = optional(string, "10s")
heartbeat_interval = optional(string, "300ms")
white_list = optional(string, "")
extra_args = optional(list(string), [])
})
default = {}
}
# ----------------------------------------------------------------------------
# Volume (stateful disk owners). Keyed `nodes` map.
# ----------------------------------------------------------------------------
variable "volume" {
description = "Volume tier. Each node owns one or more data dirs (-> -dir / -max)."
type = object({
enabled = optional(bool, true)
nodes = optional(map(object({
address = string
port = optional(number)
grpc_port = optional(number)
public_url = optional(string)
rack = optional(string)
data_center = optional(string)
data_dirs = optional(list(object({
path = string
max_volumes = optional(number, 0) # 0 => auto on non-windows
})))
idx_dir = optional(string)
disk_mounts = optional(list(object({
mountpoint = string
fstype = optional(string, "ext4")
devices = optional(list(string), [])
})), [])
})), {})
port = optional(number, 8080)
metrics_port = optional(number, 9327)
metrics_ip = optional(string, "")
ip_bind = optional(string, "0.0.0.0")
data_dirs = optional(list(object({
path = string
max_volumes = optional(number, 0)
})), [{ path = "/data", max_volumes = 0 }])
idx_dir = optional(string, "")
read_mode = optional(string, "proxy")
compaction_mbps = optional(number, 50)
min_free_space_percent = optional(string, "1")
file_size_limit_mb = optional(number, null)
index = optional(string, "memory")
images_fix_orientation = optional(bool, false)
white_list = optional(string, "")
extra_args = optional(list(string), [])
})
default = {}
}
# ----------------------------------------------------------------------------
# Filer (+ optional embedded S3). Keyed `nodes` map.
# ----------------------------------------------------------------------------
variable "filer" {
description = "Filer tier. Default leveldb2-replicated HA (each filer has its own store)."
type = object({
enabled = optional(bool, true)
nodes = optional(map(object({
address = string
port = optional(number)
grpc_port = optional(number)
data_dir = optional(string)
disk_mounts = optional(list(object({
mountpoint = string
fstype = optional(string, "ext4")
devices = optional(list(string), [])
})), [])
})), {})
port = optional(number, 8888)
metrics_port = optional(number, 9327)
metrics_ip = optional(string, "")
ip_bind = optional(string, "0.0.0.0")
data_dir = optional(string, "/var/lib/seaweedfs/filer") # -defaultStoreDir for leveldb2
default_replica_placement = optional(string, "000")
dir_list_limit = optional(number, 100000)
max_mb = optional(number, null)
disable_dir_listing = optional(bool, false)
disable_http = optional(bool, false)
encrypt_volume_data = optional(bool, false)
rack = optional(string, "")
data_center = optional(string, "")
filer_group = optional(string, "")
extra_env = optional(map(string), {})
extra_args = optional(list(string), [])
s3 = optional(object({
enabled = optional(bool, false)
port = optional(number, 8333)
https_port = optional(number, 0)
domain_name = optional(string, "")
config_path = optional(string, "/etc/seaweedfs/s3_config.json")
}), {})
})
default = {}
}
# ----------------------------------------------------------------------------
# S3 gateway (standalone). Keyed `nodes` map (stateless, but kept keyed for
# parity; wrappers may instead drive an ASG).
# ----------------------------------------------------------------------------
variable "s3" {
description = "Standalone S3 gateway tier."
type = object({
enabled = optional(bool, false)
nodes = optional(map(object({
address = string
port = optional(number)
grpc_port = optional(number)
})), {})
port = optional(number, 8333)
https_port = optional(number, 0)
iceberg_port = optional(number, null)
metrics_port = optional(number, 9327)
ip_bind = optional(string, "0.0.0.0")
domain_name = optional(string, "")
filer_address = optional(string, "") # override; else first filer node
config_path = optional(string, "/etc/seaweedfs/s3_config.json")
audit_log_config_path = optional(string, "")
cert_file = optional(string, "")
key_file = optional(string, "")
cacert_file = optional(string, "")
verify_client_cert = optional(bool, false)
extra_args = optional(list(string), [])
})
default = {}
}
# ----------------------------------------------------------------------------
# S3 identities -> rendered into the s3 config JSON (non-admin only; the admin
# credential is delivered via EnvironmentFile, never the JSON).
# ----------------------------------------------------------------------------
variable "s3_identities" {
description = "Non-admin S3 identities rendered into the config JSON. Admin key goes in the EnvironmentFile, not here."
type = list(object({
name = string
access_key = string
secret_key = string
actions = list(string)
}))
default = []
sensitive = true
}
# ----------------------------------------------------------------------------
# All-in-one (single `weed server` process).
# ----------------------------------------------------------------------------
variable "all_in_one" {
description = "All-in-one tier (weed server). Mutually exclusive with the distributed tiers."
type = object({
enabled = optional(bool, false)
nodes = optional(map(object({
address = string
data_dir = optional(string)
disk_mounts = optional(list(object({
mountpoint = string
fstype = optional(string, "ext4")
devices = optional(list(string), [])
})), [])
})), {})
data_dir = optional(string, "/data")
ip_bind = optional(string, "0.0.0.0")
master_port = optional(number, 9333)
volume_port = optional(number, 8080)
filer_port = optional(number, 8888)
default_replication = optional(string, "000")
volume_size_limit_mb = optional(number, 1000)
idle_timeout = optional(number, 30)
metrics_port = optional(number, 9324)
disable_http = optional(bool, false)
s3 = optional(object({
enabled = optional(bool, false)
port = optional(number, 8333)
config_path = optional(string, "/etc/seaweedfs/s3/s3_config.json")
domain_name = optional(string, "")
}), {})
extra_args = optional(list(string), [])
})
default = {}
}
+6
View File
@@ -0,0 +1,6 @@
terraform {
# 1.3+ is required for optional() object attributes with defaults.
# The published wrappers pin a single honest floor;
# the core itself only needs language features available since 1.3.
required_version = ">= 1.3.0"
}
+68
View File
@@ -0,0 +1,68 @@
locals {
components = ["master", "volume", "filer", "client"]
ip_sans = distinct(concat(["127.0.0.1"], var.ip_sans))
}
# ---- CA ---------------------------------------------------------------------
resource "tls_private_key" "ca" {
algorithm = var.key_algorithm
ecdsa_curve = var.ecdsa_curve
rsa_bits = var.rsa_bits
}
resource "tls_self_signed_cert" "ca" {
private_key_pem = tls_private_key.ca.private_key_pem
is_ca_certificate = true
subject {
common_name = "SeaweedFS CA"
organization = "SeaweedFS"
}
validity_period_hours = var.ca_validity_hours
allowed_uses = ["cert_signing", "crl_signing", "digital_signature"]
}
# ---- per-component leaf certs (distinct CN per component) --------------------
resource "tls_private_key" "component" {
for_each = toset(local.components)
algorithm = var.key_algorithm
ecdsa_curve = var.ecdsa_curve
rsa_bits = var.rsa_bits
}
resource "tls_cert_request" "component" {
for_each = toset(local.components)
private_key_pem = tls_private_key.component[each.key].private_key_pem
subject {
# Distinct CN per component so peer-identity authZ (allowed_wildcard_domain
# / allowed_commonNames) is meaningful. Do NOT use one CN for all.
common_name = "${each.key}.${var.internal_domain}"
organization = "SeaweedFS"
}
dns_names = distinct(concat(
["${each.key}.${var.internal_domain}", "localhost"],
var.extra_dns_sans,
))
ip_addresses = local.ip_sans
}
resource "tls_locally_signed_cert" "component" {
for_each = toset(local.components)
cert_request_pem = tls_cert_request.component[each.key].cert_request_pem
ca_private_key_pem = tls_private_key.ca.private_key_pem
ca_cert_pem = tls_self_signed_cert.ca.cert_pem
validity_period_hours = var.cert_validity_hours
early_renewal_hours = var.cert_early_renewal_hours
allowed_uses = ["digital_signature", "key_agreement", "server_auth", "client_auth"]
}
# ---- JWT signing keys -------------------------------------------------------
resource "random_password" "jwt" {
for_each = var.generate_jwt ? toset(["signing", "signing_read", "filer_signing", "filer_signing_read"]) : toset([])
length = var.jwt_length
special = false # TOML-safe
}
+65
View File
@@ -0,0 +1,65 @@
output "ca_cert_pem" {
description = "CA certificate PEM."
value = tls_self_signed_cert.ca.cert_pem
}
output "certs" {
description = "Per-component { cert_pem, private_key_pem }."
value = {
for c in local.components : c => {
cert_pem = tls_locally_signed_cert.component[c].cert_pem
private_key_pem = tls_private_key.component[c].private_key_pem
}
}
sensitive = true
}
output "jwt" {
description = "Generated JWT keys (empty strings when generate_jwt = false)."
value = {
signing = var.generate_jwt ? random_password.jwt["signing"].result : ""
signing_read = var.generate_jwt ? random_password.jwt["signing_read"].result : ""
filer_signing = var.generate_jwt ? random_password.jwt["filer_signing"].result : ""
filer_signing_read = var.generate_jwt ? random_password.jwt["filer_signing_read"].result : ""
}
sensitive = true
}
# Ready to pass straight to the core module's `security` variable.
output "core_security" {
description = "Object shaped for the core module's `security` input (cert_dir, allowed_wildcard_domain, JWT keys)."
value = {
cert_dir = var.cert_dir
allowed_wildcard_domain = ".${var.internal_domain}"
allowed_common_names = ""
jwt_signing_key = var.generate_jwt ? random_password.jwt["signing"].result : ""
jwt_signing_read_key = var.generate_jwt ? random_password.jwt["signing_read"].result : ""
jwt_filer_signing_key = var.generate_jwt ? random_password.jwt["filer_signing"].result : ""
jwt_filer_signing_read_key = var.generate_jwt ? random_password.jwt["filer_signing_read"].result : ""
}
sensitive = true
}
# Non-sensitive manifest (key, on-host path, mode) so a wrapper can for_each
# over SSM/secret-store resources without tripping the sensitive-for_each rule.
output "secret_manifest" {
description = "List of { key, path, mode } for the CA + component cert/key files (no content)."
value = concat(
[{ key = "ca/tls.crt", path = "${var.cert_dir}/ca/tls.crt", mode = "0644" }],
flatten([for c in local.components : [
{ key = "${c}/tls.crt", path = "${var.cert_dir}/${c}/tls.crt", mode = "0644" },
{ key = "${c}/tls.key", path = "${var.cert_dir}/${c}/tls.key", mode = "0600" },
]]),
)
}
# Sensitive map keyed by the manifest key -> file content.
output "secret_contents" {
description = "Map of manifest key -> PEM content."
value = merge(
{ "ca/tls.crt" = tls_self_signed_cert.ca.cert_pem },
{ for c in local.components : "${c}/tls.crt" => tls_locally_signed_cert.component[c].cert_pem },
{ for c in local.components : "${c}/tls.key" => tls_private_key.component[c].private_key_pem },
)
sensitive = true
}
+89
View File
@@ -0,0 +1,89 @@
# =============================================================================
# SeaweedFS security material generator (cloud-agnostic).
#
# Generates the CA, per-component mTLS certs with DISTINCT CommonNames, and JWT
# signing keys. Outputs a `core_security` object ready to feed the core module's
# `security` variable, plus the raw PEMs for a wrapper to deliver via a secret
# store. NOTE: TF-generated secrets live in state; prefer Vault PKI for the CA
# in production.
# =============================================================================
variable "internal_domain" {
description = "Internal domain for component CNs (e.g. master.<domain>). Drives the allowed_wildcard_domain peer-auth check."
type = string
default = "seaweedfs.internal"
}
variable "ip_sans" {
description = "IP addresses to include as SANs on every component cert (all node private IPs + 127.0.0.1 is added automatically)."
type = list(string)
default = []
}
variable "extra_dns_sans" {
description = "Additional DNS SANs to include on every component cert."
type = list(string)
default = []
}
variable "key_algorithm" {
description = "CA/leaf key algorithm: ECDSA (default, modern) or RSA."
type = string
default = "ECDSA"
validation {
condition = contains(["ECDSA", "RSA"], var.key_algorithm)
error_message = "key_algorithm must be ECDSA or RSA."
}
}
variable "ecdsa_curve" {
description = "ECDSA curve when key_algorithm = ECDSA."
type = string
default = "P256"
}
variable "rsa_bits" {
description = "RSA key size when key_algorithm = RSA."
type = number
default = 3072
}
variable "ca_validity_hours" {
description = "CA certificate validity (default 10 years)."
type = number
default = 87600
}
variable "cert_validity_hours" {
description = "Component certificate validity. Short by default; certs hot-reload so reissue is cheap."
type = number
default = 72
}
variable "cert_early_renewal_hours" {
description = "Renew component certs this many hours before expiry."
type = number
default = 24
}
variable "cert_dir" {
description = "On-host directory where certs are placed; surfaced in core_security.cert_dir."
type = string
default = "/usr/local/share/ca-certificates"
}
variable "generate_jwt" {
description = "Generate JWT signing keys (signing, signing.read, filer_signing, filer_signing.read)."
type = bool
default = true
}
variable "jwt_length" {
description = "Length of generated JWT keys (>= 32 hardened)."
type = number
default = 40
validation {
condition = var.jwt_length >= 32
error_message = "jwt_length must be >= 32 (hardened minimum)."
}
}
+13
View File
@@ -0,0 +1,13 @@
terraform {
required_version = ">= 1.3.0"
required_providers {
tls = {
source = "hashicorp/tls"
version = ">= 4.0"
}
random = {
source = "hashicorp/random"
version = ">= 3.5"
}
}
}
+74
View File
@@ -0,0 +1,74 @@
# =============================================================================
# Local mTLS test harness: generate the CA + component certs + JWT keys with the
# security submodule, render security.toml with the core, then run a real
# master+volume+filer cluster with mTLS enabled on 127.0.0.1 and assert it works.
# Proves the generated security material is valid and accepted by weed.
# =============================================================================
terraform {
required_version = ">= 1.3.0"
}
variable "weed_binary" {
type = string
default = "/usr/bin/weed"
}
variable "workdir" {
type = string
default = "/tmp/seaweedfs-tftest-secure"
}
module "security" {
source = "../../modules/security"
internal_domain = "seaweedfs.internal"
ip_sans = ["127.0.0.1"]
cert_dir = "${var.workdir}/certs"
}
module "core" {
source = "../../modules/core"
weed_binary = var.weed_binary
enable_security = true
security = module.security.core_security
# We read security.toml back from the output and place it ourselves (under
# $HOME/.seaweedfs), so keep it in the rendered file set.
render_secret_files = true
# High, distinct ports so this never collides with a running SeaweedFS or the
# non-secure harness.
master = {
nodes = { m0 = { address = "127.0.0.1", port = 29336, data_dir = "${var.workdir}/master-m0" } }
election_timeout = "3s"
heartbeat_interval = "200ms"
}
volume = {
nodes = { v0 = { address = "127.0.0.1", port = 28081, data_dirs = [{ path = "${var.workdir}/volume-v0", max_volumes = 20 }] } }
}
filer = {
nodes = { f0 = { address = "127.0.0.1", port = 28889, data_dir = "${var.workdir}/filer-f0" } }
}
}
output "cluster" {
sensitive = true
value = {
for name, n in module.core.nodes : name => {
role = n.role
http_port = n.ports.http
data_dirs = n.data_dirs
argv = n.argv
env = n.env
}
}
}
output "security_toml" {
sensitive = true
value = module.core.secret_files_by_node["master-m0"]["/etc/seaweedfs/security.toml"]
}
output "certs" {
sensitive = true
value = [for m in module.security.secret_manifest : { path = m.path, mode = m.mode, content = module.security.secret_contents[m.key] }]
}
+128
View File
@@ -0,0 +1,128 @@
#!/usr/bin/env bash
# Generate mTLS material + security.toml with Terraform, then run a real
# master+volume+filer cluster with mTLS enabled and assert it works.
# ./run_local_secure.sh # render + run + assert + teardown
# KEEP=1 ./run_local_secure.sh
set -u
HERE="$(cd "$(dirname "$0")" && pwd)"
export PATH="/opt/homebrew/bin:$PATH"
TOFU="${TOFU:-tofu}"
WEED="${WEED:-$(go env GOPATH 2>/dev/null || echo "$HOME/go")/bin/weed}"
WORKDIR="${WORKDIR:-/tmp/seaweedfs-tftest-secure}"
LOGDIR="$WORKDIR/logs"
RUNDIR="$WORKDIR/run"
PASS=0
FAIL=0
ok() { echo " PASS: $1"; PASS=$((PASS + 1)); }
bad() { echo " FAIL: $1"; FAIL=$((FAIL + 1)); }
info() { echo "==> $1"; }
cleanup() {
info "tearing down"
if [ -d "$RUNDIR" ]; then
for pf in "$RUNDIR"/*.pid; do [ -f "$pf" ] && kill "$(cat "$pf")" 2>/dev/null; done
sleep 1
for pf in "$RUNDIR"/*.pid; do [ -f "$pf" ] && kill -9 "$(cat "$pf")" 2>/dev/null; done
fi
}
info "cleaning $WORKDIR"
case "$WORKDIR" in "" | "/" | "$HOME") echo "refusing to delete '$WORKDIR'" >&2; exit 2 ;; esac
rm -rf "$WORKDIR"
mkdir -p "$LOGDIR" "$RUNDIR" "$WORKDIR/.seaweedfs"
[ -x "$WEED" ] || { echo "weed not found at $WEED" >&2; exit 2; }
info "generating certs + rendering config with OpenTofu"
cd "$HERE"
"$TOFU" init -backend=false -input=false -no-color >/dev/null 2>&1 || { echo "tofu init failed"; exit 2; }
if ! "$TOFU" apply -auto-approve -input=false -no-color \
-var "weed_binary=$WEED" -var "workdir=$WORKDIR" >"$LOGDIR/tofu.log" 2>&1; then
echo "tofu apply failed"; tail -30 "$LOGDIR/tofu.log"; exit 2
fi
# write certs to their on-host paths
"$TOFU" output -json certs | jq -c '.[]' | while IFS= read -r c; do
p="$(echo "$c" | jq -r '.path')"; m="$(echo "$c" | jq -r '.mode')"
mkdir -p "$(dirname "$p")"
echo "$c" | jq -r '.content' > "$p"
chmod "$m" "$p"
done
# place security.toml where weed searches ($HOME/.seaweedfs)
"$TOFU" output -raw security_toml > "$WORKDIR/.seaweedfs/security.toml"
info "security.toml + $(ls "$WORKDIR"/certs | wc -l | tr -d ' ') cert dirs written"
OUT="$("$TOFU" output -json cluster)"
# derive ports from the rendered config (high range; never hardcoded)
port_of() { echo "$OUT" | jq -r --arg r "$1" '.[] | select(.role==$r) | .http_port' | head -1; }
MPORT="$(port_of master)"; VPORT="$(port_of volume)"; FPORT="$(port_of filer)"
busy=""
for p in $(echo "$OUT" | jq -r '.[].http_port'); do
lsof -nP -iTCP:"$p" -sTCP:LISTEN >/dev/null 2>&1 && busy="$busy $p"
done
[ -n "$busy" ] && { echo "Required port(s) already in use:$busy -- is another SeaweedFS running?" >&2; exit 3; }
[ "${KEEP:-0}" = "1" ] || trap cleanup EXIT INT TERM
# launch a node with HOME pointed at $WORKDIR so weed loads security.toml
launch() {
n="$1"
for d in $(echo "$OUT" | jq -r --arg n "$n" '.[$n].data_dirs[]?'); do mkdir -p "$d"; done
ENVS=("HOME=$WORKDIR")
while IFS= read -r e; do [ -n "$e" ] && ENVS+=("$e"); done \
< <(echo "$OUT" | jq -r --arg n "$n" '.[$n].env | to_entries[] | "\(.key)=\(.value)"')
ARGV=()
while IFS= read -r a; do ARGV+=("$a"); done \
< <(echo "$OUT" | jq -r --arg n "$n" '.[$n].argv[]')
info "launching $n (mTLS)"
env "${ENVS[@]}" "$WEED" "${ARGV[@]}" >"$LOGDIR/$n.log" 2>&1 &
echo "$!" > "$RUNDIR/$n.pid"
}
wait_http() {
url="$1"; t="${2:-30}"; i=0
while [ "$i" -lt "$t" ]; do
curl -fsS -o /dev/null --max-time 2 "$url" 2>/dev/null && return 0
i=$((i + 1)); sleep 1
done
return 1
}
launch master-m0
QOK=0; i=0
while [ "$i" -lt 40 ]; do
st="$(curl -fsS --max-time 2 "http://127.0.0.1:$MPORT/cluster/status" 2>/dev/null)" || { i=$((i+1)); sleep 1; continue; }
[ "$(echo "$st" | jq -r '.IsLeader // false')" = "true" ] && { QOK=1; break; }
i=$((i + 1)); sleep 1
done
[ "$QOK" -eq 1 ] && ok "master elected leader under mTLS" || bad "master leader (mTLS)"
launch volume-v0
# registration (master gRPC over mTLS) confirms the volume joined the cluster
AOK=0; i=0
while [ "$i" -lt 40 ]; do
fid="$(curl -fsS --max-time 2 "http://127.0.0.1:$MPORT/dir/assign" 2>/dev/null | jq -r '.fid // ""')"
[ -n "$fid" ] && { AOK=1; break; }
i=$((i + 1)); sleep 1
done
[ "$AOK" -eq 1 ] && ok "volume registered via mTLS gRPC (fid $fid)" || bad "volume registration over mTLS"
# /healthz flips to 200 after the first heartbeat completes (slower under mTLS)
wait_http "http://127.0.0.1:$VPORT/healthz" 60 && ok "volume /healthz up (mTLS)" || bad "volume /healthz"
launch filer-f0
wait_http "http://127.0.0.1:$FPORT/" 30 && ok "filer / up (mTLS)" || bad "filer /"
# [jwt.filer_signing] is active, so the filer requires a signed JWT for writes.
# An unsigned write MUST be rejected with 401 -- this proves the JWT signing key
# rendered into security.toml is enforced (positive security assertion).
echo "smoke" > "$WORKDIR/hello.txt"
code="$(curl -s -o /dev/null -w '%{http_code}' --max-time 5 -X POST \
-F "file=@$WORKDIR/hello.txt" "http://127.0.0.1:$FPORT/smoke/hello.txt" 2>/dev/null)"
[ "$code" = "401" ] && ok "filer rejects unsigned write (HTTP 401) => JWT signing enforced" \
|| bad "filer JWT enforcement (expected 401, got HTTP $code)"
echo
info "RESULTS: $PASS passed, $FAIL failed"
[ "$FAIL" -eq 0 ]
+100
View File
@@ -0,0 +1,100 @@
# =============================================================================
# Local test harness: render a small SeaweedFS cluster with the core module
# and run it as real `weed` processes on 127.0.0.1 (no cloud, no docker).
#
# 3 masters (quorum) + 1 volume + 1 filer + 1 standalone S3, each on a distinct
# port. run_local_cluster.sh consumes the `cluster` output, launches the weed
# processes from the rendered argv, and asserts the cluster actually works.
# =============================================================================
terraform {
required_version = ">= 1.3.0"
}
variable "weed_binary" {
description = "Path to the weed executable used for the local cluster."
type = string
default = "/usr/bin/weed"
}
variable "workdir" {
description = "Scratch directory for per-node data dirs and config files."
type = string
default = "/tmp/seaweedfs-tftest"
}
module "core" {
source = "../../modules/core"
weed_binary = var.weed_binary
monitoring_enabled = false
enable_security = false
# High port range so the harness does not collide with a SeaweedFS cluster
# that may already be running on this dev machine (default 9333/8080/8888/8333).
master = {
nodes = {
m0 = { address = "127.0.0.1", port = 29333, data_dir = "${var.workdir}/master-m0" }
m1 = { address = "127.0.0.1", port = 29334, data_dir = "${var.workdir}/master-m1" }
m2 = { address = "127.0.0.1", port = 29335, data_dir = "${var.workdir}/master-m2" }
}
# Tighten election so a 3-node local quorum converges quickly.
election_timeout = "3s"
heartbeat_interval = "200ms"
}
volume = {
nodes = {
v0 = {
address = "127.0.0.1"
port = 28080
rack = "rack-a"
data_dirs = [{ path = "${var.workdir}/volume-v0", max_volumes = 20 }]
}
}
}
filer = {
nodes = {
f0 = { address = "127.0.0.1", port = 28888, data_dir = "${var.workdir}/filer-f0" }
}
}
s3 = {
enabled = true
nodes = { s0 = { address = "127.0.0.1", port = 28333 } }
# this weed build starts an Iceberg REST catalog on 8181 by default; disable
# it so the test does not collide with a cluster already using that port.
iceberg_port = 0
config_path = "${var.workdir}/s3_config.json"
}
# Anonymous identity so the smoke test can PUT/GET without sigv4 signing.
s3_identities = [{
name = "anonymous"
access_key = ""
secret_key = ""
actions = ["Admin", "Read", "Write", "List", "Tagging"]
}]
}
output "cluster" {
description = "Node specs consumed by run_local_cluster.sh (read via `tofu output -json`, which emits values even when sensitive)."
sensitive = true
value = {
for name, n in module.core.nodes : name => {
role = n.role
address = n.address
http_port = n.ports.http
data_dirs = n.data_dirs
argv = n.argv
env = n.env
config_files = n.config_files
exec_start = n.exec_start
}
}
}
output "master_peers" {
value = module.core.master_peers
}
+170
View File
@@ -0,0 +1,170 @@
#!/usr/bin/env bash
# Render a SeaweedFS cluster with the Terraform core module and run it as real
# `weed` processes locally, then assert it actually works. No cloud, no docker.
#
# ./run_local_cluster.sh # render + run + assert + teardown
# KEEP=1 ./run_local_cluster.sh # leave the cluster running after asserts
# WEED=/path/to/weed ./run_local_cluster.sh
#
# Ports come from the rendered config (high range, to avoid colliding with a
# SeaweedFS cluster already running on this machine). Exits non-zero on any
# failed assertion or if a required port is already in use.
set -u
HERE="$(cd "$(dirname "$0")" && pwd)"
export PATH="/opt/homebrew/bin:$PATH"
TOFU="${TOFU:-tofu}"
WEED="${WEED:-$(go env GOPATH 2>/dev/null || echo "$HOME/go")/bin/weed}"
WORKDIR="${WORKDIR:-/tmp/seaweedfs-tftest}"
LOGDIR="$WORKDIR/logs"
RUNDIR="$WORKDIR/run"
PASS=0
FAIL=0
ok() { echo " PASS: $1"; PASS=$((PASS + 1)); }
bad() { echo " FAIL: $1"; FAIL=$((FAIL + 1)); }
info() { echo "==> $1"; }
cleanup() {
info "tearing down"
if [ -d "$RUNDIR" ]; then
for pf in "$RUNDIR"/*.pid; do [ -f "$pf" ] && kill "$(cat "$pf")" 2>/dev/null; done
sleep 1
for pf in "$RUNDIR"/*.pid; do [ -f "$pf" ] && kill -9 "$(cat "$pf")" 2>/dev/null; done
fi
}
info "cleaning $WORKDIR"
case "$WORKDIR" in "" | "/" | "$HOME") echo "refusing to delete '$WORKDIR'" >&2; exit 2 ;; esac
rm -rf "$WORKDIR"
mkdir -p "$LOGDIR" "$RUNDIR"
[ -x "$WEED" ] || { echo "weed binary not found/executable at $WEED" >&2; exit 2; }
info "rendering cluster config with OpenTofu"
cd "$HERE"
"$TOFU" init -backend=false -input=false -no-color >/dev/null 2>&1 || { echo "tofu init failed"; exit 2; }
if ! "$TOFU" apply -auto-approve -input=false -no-color \
-var "weed_binary=$WEED" -var "workdir=$WORKDIR" >"$LOGDIR/tofu.log" 2>&1; then
echo "tofu apply failed; see $LOGDIR/tofu.log"; tail -30 "$LOGDIR/tofu.log"; exit 2
fi
OUT="$("$TOFU" output -json cluster)"
[ -n "$OUT" ] || { echo "empty cluster output"; exit 2; }
# ---- derive ports by role (never hardcoded) ---------------------------------
port_of() { echo "$OUT" | jq -r --arg r "$1" '.[] | select(.role==$r) | .http_port' | head -1; }
mports() { echo "$OUT" | jq -r '.[] | select(.role=="master") | .http_port' | sort; }
MPORT1="$(mports | head -1)"
VPORT="$(port_of volume)"; FPORT="$(port_of filer)"; SPORT="$(port_of s3)"
NMASTERS="$(mports | wc -l | tr -d ' ')"
# ---- refuse to run if a required port is already taken (foreign cluster) -----
busy=""
for p in $(echo "$OUT" | jq -r '.[].http_port'); do
lsof -nP -iTCP:"$p" -sTCP:LISTEN >/dev/null 2>&1 && busy="$busy $p"
done
[ -n "$busy" ] && { echo "Required port(s) already in use:$busy -- is another SeaweedFS running?" >&2; exit 3; }
[ "${KEEP:-0}" = "1" ] || trap cleanup EXIT INT TERM
node_names() { echo "$OUT" | jq -r 'keys[]'; }
node_field() { echo "$OUT" | jq -r --arg n "$1" --arg f "$2" '.[$n][$f]'; }
launch_node() {
n="$1"
for p in $(echo "$OUT" | jq -r --arg n "$n" '.[$n].config_files | keys[]?'); do
mkdir -p "$(dirname "$p")"
echo "$OUT" | jq -r --arg n "$n" --arg p "$p" '.[$n].config_files[$p]' > "$p"
done
for d in $(echo "$OUT" | jq -r --arg n "$n" '.[$n].data_dirs[]?'); do mkdir -p "$d"; done
ENVS=()
while IFS= read -r e; do [ -n "$e" ] && ENVS+=("$e"); done \
< <(echo "$OUT" | jq -r --arg n "$n" '.[$n].env | to_entries[] | "\(.key)=\(.value)"')
ARGV=()
while IFS= read -r a; do ARGV+=("$a"); done \
< <(echo "$OUT" | jq -r --arg n "$n" '.[$n].argv[]')
info "launching $n"
if [ "${#ENVS[@]}" -gt 0 ]; then
env "${ENVS[@]}" "$WEED" "${ARGV[@]}" >"$LOGDIR/$n.log" 2>&1 &
else
"$WEED" "${ARGV[@]}" >"$LOGDIR/$n.log" 2>&1 &
fi
echo "$!" > "$RUNDIR/$n.pid"
}
wait_http() {
url="$1"; t="${2:-30}"; i=0
while [ "$i" -lt "$t" ]; do
curl -fsS -o /dev/null --max-time 2 "$url" 2>/dev/null && return 0
i=$((i + 1)); sleep 1
done
return 1
}
# ---- launch masters and wait for quorum -------------------------------------
for n in $(node_names); do
case "$(node_field "$n" role)" in master) launch_node "$n";; esac
done
info "waiting for master quorum across ports: $(mports | tr '\n' ' ')"
QUORUM_OK=0; i=0
while [ "$i" -lt 40 ]; do
LEADERS=0; AGREED_LEADER=""; mismatch=0; reachable=0
for port in $(mports); do
st="$(curl -fsS --max-time 2 "http://127.0.0.1:$port/cluster/status" 2>/dev/null)" || continue
reachable=$((reachable + 1))
[ "$(echo "$st" | jq -r '.IsLeader // false')" = "true" ] && LEADERS=$((LEADERS + 1))
ldr="$(echo "$st" | jq -r '.Leader // ""')"
if [ -n "$ldr" ]; then
if [ -z "$AGREED_LEADER" ]; then AGREED_LEADER="$ldr"; elif [ "$AGREED_LEADER" != "$ldr" ]; then mismatch=1; fi
fi
done
if [ "$reachable" -eq "$NMASTERS" ] && [ "$LEADERS" -eq 1 ] && [ "$mismatch" -eq 0 ] && [ -n "$AGREED_LEADER" ]; then
QUORUM_OK=1; break
fi
i=$((i + 1)); sleep 1
done
if [ "$QUORUM_OK" -eq 1 ]; then
ok "$NMASTERS-master quorum: exactly one leader ($AGREED_LEADER), all agree"
else
bad "master quorum (reachable=$reachable leaders=$LEADERS leader='$AGREED_LEADER')"
fi
# ---- volume -----------------------------------------------------------------
for n in $(node_names); do
case "$(node_field "$n" role)" in volume) launch_node "$n";; esac
done
wait_http "http://127.0.0.1:$VPORT/healthz" 40 && ok "volume /healthz up" || bad "volume /healthz"
ASSIGN_OK=0; i=0
while [ "$i" -lt 20 ]; do
fid="$(curl -fsS --max-time 2 "http://127.0.0.1:$MPORT1/dir/assign" 2>/dev/null | jq -r '.fid // ""')"
[ -n "$fid" ] && { ASSIGN_OK=1; break; }
i=$((i + 1)); sleep 1
done
[ "$ASSIGN_OK" -eq 1 ] && ok "master assigned fid ($fid) => volume registered" || bad "master /dir/assign (no volume?)"
# ---- filer ------------------------------------------------------------------
for n in $(node_names); do
case "$(node_field "$n" role)" in filer) launch_node "$n";; esac
done
wait_http "http://127.0.0.1:$FPORT/" 30 && ok "filer / up" || bad "filer /"
PAYLOAD="seaweedfs-terraform-smoke-$$"
echo "$PAYLOAD" > "$WORKDIR/hello.txt"
curl -fsS -F "file=@$WORKDIR/hello.txt" "http://127.0.0.1:$FPORT/smoke/hello.txt" >/dev/null 2>&1
got="$(curl -fsS --max-time 5 "http://127.0.0.1:$FPORT/smoke/hello.txt" 2>/dev/null)"
[ "$got" = "$PAYLOAD" ] && ok "filer PUT/GET round-trip" || bad "filer round-trip (got '$got')"
# ---- s3 ---------------------------------------------------------------------
for n in $(node_names); do
case "$(node_field "$n" role)" in s3) launch_node "$n";; esac
done
wait_http "http://127.0.0.1:$SPORT/status" 30 && ok "s3 /status up" || bad "s3 /status"
curl -fsS -X PUT "http://127.0.0.1:$SPORT/smoke-bucket" >/dev/null 2>&1
sleep 1
curl -fsS -X PUT --data-binary "$PAYLOAD" "http://127.0.0.1:$SPORT/smoke-bucket/obj.txt" >/dev/null 2>&1
s3got="$(curl -fsS --max-time 5 "http://127.0.0.1:$SPORT/smoke-bucket/obj.txt" 2>/dev/null)"
[ "$s3got" = "$PAYLOAD" ] && ok "s3 PUT/GET round-trip" || bad "s3 round-trip (got '$s3got')"
echo
info "RESULTS: $PASS passed, $FAIL failed"
[ "${KEEP:-0}" = "1" ] && info "KEEP=1: cluster left running. Logs in $LOGDIR."
[ "$FAIL" -eq 0 ]