Node Configuration

InfraMind nodes are designed to run autonomously and securely, with minimal manual configuration. However, advanced operators and infrastructure teams may need to customize runtime behavior, optimize region placement, enable hardware features, or integrate nodes into their internal monitoring systems. Node configuration is handled through a mix of environment variables, config files, and systemd options, all of which are respected by the node agent at boot.

Most configuration is stored and interpreted from ~/.inframind/config.yaml, but for ephemeral control and compatibility with orchestrated environments (e.g. Kubernetes, Nomad, systemd), all essential parameters can also be set using environment variables.

Core Environment Variables

These environment variables are read at runtime and override values in the YAML config.

INFRA_NODE_KEY=ed25519:MHcCAQEEIBvYhF6uFs+T1aALHt...
INFRA_LOCATION_HINT=asia-south
ALLOW_GPU=true
CONTAINER_CACHE_PATH=/mnt/inframind/cache
INFRA_LOG_LEVEL=info
INFRA_MAX_JOBS=3

INFRA_NODE_KEY Defines the node’s private identity key used for signing job receipts and staking transactions. If not present, a key is generated and stored in ~/.inframind/identity.key. Format is Ed25519 base64 encoded.

To generate a persistent key:

openssl genpkey -algorithm ed25519 -out ~/.inframind/identity.key

To load from environment:

export INFRA_NODE_KEY=$(cat ~/.inframind/identity.key | base64)

INFRA_LOCATION_HINT Overrides automatic region detection. Used by the scheduler to optimize routing decisions.

Examples:

INFRA_LOCATION_HINT=us-west
INFRA_LOCATION_HINT=europe-central
INFRA_LOCATION_HINT=africa-north

Recommended when operating in regions with asymmetric latency or mixed IP geolocation.

ALLOW_GPU=true Enables GPU-based job eligibility. Requires nvidia-container-runtime and correct driver installation.

Validate:

nvidia-smi

If ALLOW_GPU is set to true, the agent performs a CUDA test on boot. Jobs with GPU requirements are only routed to nodes with verified GPU capability.

Optional tuning:

CUDA_VISIBLE_DEVICES=0,1
INFRA_GPU_PROFILE=A100

CONTAINER_CACHE_PATH Sets a persistent path for container images and model bundles. Default is ~/.inframind/cache.

For large disk volumes:

export CONTAINER_CACHE_PATH=/var/lib/inframind/cache

Set via CLI:

infra config set container_cache=/data/cache

INFRA_LOG_LEVEL Controls log verbosity.

• error – fatal failures only

• warn – non-critical issues

• info – default runtime events

• debug – detailed trace, container logs, scheduler handshakes

Set persistently:

export INFRA_LOG_LEVEL=debug

Or update config:

infra config set log_level=debug

INFRA_MAX_JOBS Limits concurrent running containers. Prevents overload on lower-spec systems.

INFRA_MAX_JOBS=1

If system resources are exceeded, excess jobs are rejected with a 503 Retry.

Persistence and Resilience

Logs are stored at:

~/.inframind/logs/inframind.log

To persist logs across reboots, rotate them weekly:

logrotate /etc/logrotate.d/inframind

Example logrotate config:

~/.inframind/logs/inframind.log {
  weekly
  rotate 4
  compress
  delaycompress
  missingok
  notifempty
  copytruncate
}

Restart on Failure

For systemd-managed nodes, restart policies should be enforced for high availability:

sudo systemctl edit inframind-node

Insert:

[Service]
Restart=always
RestartSec=5

Then:

sudo systemctl daemon-reexec
sudo systemctl restart inframind-node

This guarantees automatic recovery on crash, out-of-memory kill, or machine reboot.

Custom Configuration File

Modify ~/.inframind/config.yaml for persistent configuration:

node_id: auto
region_hint: us-east
allow_gpu: true
max_jobs: 2
container_cache: /mnt/cache
log_level: info
heartbeat_interval: 5

Changes take effect on next restart.

Integration with Supervisors

For orchestration tools like Docker Compose or Kubernetes:

services:
  inframind:
    image: inframind/node:latest
    ports:
      - "9000:9000"
    volumes:
      - ~/.inframind:/node
      - /mnt/cache:/mnt/cache
    environment:
      - INFRA_NODE_KEY=${INFRA_NODE_KEY}
      - INFRA_LOCATION_HINT=us-east
      - ALLOW_GPU=true
      - INFRA_LOG_LEVEL=info

InfraMind node configuration is designed to be declarative, override-friendly, and compatible with local or large-scale deployments. Whether deployed on a single board computer or inside a multi-GPU cluster, nodes remain protocol-compliant and automatically reflect their configuration in all heartbeat payloads and registry metrics. This ensures the scheduler can route jobs intelligently, and that every operator has full control over how their node behaves inside the mesh.