Deployment Preparation

This chapter covers the practical preparation needed before deploying Curvine from a release package or from a locally built build/dist.

Supported Runtime BaselineCopy Anchor Link

The current main branch and release workflow are Linux-first. The officially documented tested distributions are:

macOS has partial support in the codebase and can be useful for development or limited local testing. Windows should be treated as limited support, not as a primary deployment target.

Resource RequirementsCopy Anchor Link

Curvine has no minimum resource requirements and can support extremely high concurrency and traffic with very small resources. Here is a reference configuration:

• CPU: 2 cores
• Memory: 4GB
• Network: 10Gbps
• SSD disk: 1 unit

Based on this reference, you can extrapolate other hardware resource requirements from one resource value. For example, if you have 2 SSD disks, you would need 4 CPU cores, 8GB memory, and 20Gbps bandwidth. 4GB memory is also acceptable, depending on business concurrency - if concurrency is not high, memory doesn't need to be increased.

Create Installation PackageCopy Anchor Link

Compile and package the software. For compilation steps, see Download and Compile Curvine.

From the project root, run:

make dist

This runs make all (if needed), then packages the contents of build/dist into a tar.gz file. The archive is created in the project root with a name like curvine-<platform>-<arch>-<timestamp>.tar.gz, or curvine-<version>-<platform>-<arch>.tar.gz if you set RELEASE_VERSION (e.g. RELEASE_VERSION=v1.0.0 make dist). This archive is the Curvine installation package for deployment or for building runtime images.

Environment Script and Main ConfigCopy Anchor Link

After unpacking the installation package, the key files are:

• conf/curvine-env.sh: shell-side environment setup
• conf/curvine-cluster.toml: main cluster configuration

The environment script exports these important variables:

• CURVINE_MASTER_HOSTNAME: defaults to localhost
• CURVINE_WORKER_HOSTNAME: defaults to localhost
• CURVINE_CLIENT_HOSTNAME: defaults to localhost
• CURVINE_CONF_FILE: points to conf/curvine-cluster.toml

The defaults come from etc/curvine-env.sh and are only suitable for single-node local use. In multi-NIC, container, or multi-node deployments, set these explicitly instead of relying on the defaults. Alternatively, set net_interface in curvine-cluster.toml to let each node resolve its own IPv4; see the Configuration Reference.

export CURVINE_MASTER_HOSTNAME=master1
export CURVINE_WORKER_HOSTNAME=10.0.0.21
export CURVINE_CLIENT_HOSTNAME=client1

conf/curvine-cluster.toml is the main TOML configuration file. The most common changes are:

1. Master / journal addresses
2. Worker storage directories
3. Client-side master_addrs
4. Log destinations and levels

Example:

format_master = false
format_worker = false

[master]
# Configure metadata storage directory
meta_dir = "data/meta"

# Configure master log directory
log = { level = "info", log_dir = "logs", file_name = "master.log" }

[journal]
# Raft peer list. Each hostname must match the effective master hostname
# on the corresponding node.
journal_addrs = [
    {id = 1, hostname = "master1", port = 8996},
    {id = 2, hostname = "master2", port = 8996},
    {id = 3, hostname = "master3", port = 8996}
]

# Configure raft log storage directory
journal_dir = "data/journal"

[worker]
# Reserved space, default is 0
dir_reserved = "0"

# Configure worker storage directories
data_dir = [
    "[SSD]/data/data1",
    "[SSD]/data/data2"
]

# Configure worker logs
log = { level = "info", log_dir = "logs", file_name = "worker.log" }

[client]
# Master RPC addresses
master_addrs = [
    { hostname = "master1", port = 8995 },
    { hostname = "master2", port = 8995 },
    { hostname = "master3", port = 8995 },
]

# Client log configuration
[log]
level = "info"
log_dir = "logs"
file_name = "curvine.log"

If you need the Java Hadoop client path, update conf/curvine-site.xml as well:

<property>
    <name>fs.cv.default.master_addrs</name>
    <value>master1:8995,master2:8995,master3:8995</value>
</property>