config.md

Ravel configuration

Ravel accept one unique TOML configuration file. This file is usually located at /etc/ravel/config.toml.

Common configuration

To run in a cluster configuration, you need to give an access to a NATS server both to the Agent and the Server.

[nats]
url = "nats://127.0.0.1:4222"
cred_file = "./nats.creds"

Daemon configuration

The Daemon is the process responsible to manage the Ravel Runtime and the Ravel Agent.The daemon holds its state in a bbolt.

[daemon]
database_path = "/var/lib/ravel/daemon.db"

Runtime configuration

The Ravel Runtime is responsible of all the VMs instances running on one host. It includes the Ravel Machines VMs and local VMs.

[daemon.runtime]
cloud_hypervisor_binary = "/opt/ravel/cloud-hypervisor" # Must be statically linked 
jailer_binary = "/opt/ravel/jailer" # Path to the Ravel jailer
init_binary = "./opt/ravel/initd"  # Path to Initd binary
linux_kernel = "/opt/ravel/vmlinux.bin" # A build of the cloud-hypervisor linux kernel

Agent configuration

The Ravel Agent is responsible of managing workloads assigned to one host in the Ravel cluster.

[daemon.agent]
resources = { cpus_mhz = 20000, memory_mb = 16_384 } # Resources allocatable by the agent
node_id = "ravel-1" # MUST be unique in a cluster because its uniqueness is not enforced.
region = "fr" # The region the agent will announce itself in
address = "127.0.0.1" # The cluster wide reachable address of the agent
port = 8080 # The HTTP port the agent will listen on for the internal API

In production environments, you definitly want to enable the mtls: Theses certificates can be generated with the ravel tls commands.

[daemon.agent.tls]
cert_file = "ravel-1-agent-cert.pem"
key_file = "ravel-1-agent-key.pem"
ca_file = "ravel-ca-cert.pem"

Server configuration

The Ravel server is responsible to accept API requests to schedule workloads on the cluster. The Ravel server store his state in a Postgres database and use HTTP and NATS to communicate with the agents.

[server]
postgres_url = "postgres://user:pass@host:5432/ravel-db" 
[server.api]
address = ":3000" # The HTTP address the server will listen on

# [server.api.tls]
# cert_file = ""
# key_file = ""
# ca_file = ""
[server.tls] # The server tls configuration used to communicate with the agents
cert_file = "server-1-server-cert.pem"
key_file = "server-1-server-key.pem"
ca_file = "ravel-ca-cert.pem"

Once the server is configure you can start it with the ravel server command.

Machine templates

The ravel server must be configured with some machine templates to start accepting workloads.

The vcpu_frequency is the frequency allocated by virtual-cpus in MHz. This value is converted CPU quotas used by Ravel with cgroups to control the CPU usage of the VMs. For each count of vcpus, you can define a list of memory configurations in MB.

[server.machine_templates.std]
vcpu_frequency = 2500
combinations = [
    { vcpus = 1, memory_configs = [
        1_024,
        2_048,
        4_096,
    ] },
    { vcpus = 2, memory_configs = [
        2_048,
        4_096,
        8_192,
    ] },
    { vcpus = 4, memory_configs = [
        4_096,
        8_192,
        16_384,
    ] },
]