Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
35 changes: 35 additions & 0 deletions pmoves/config/profiles/jetson-orin-1.yaml
Original file line number Diff line number Diff line change
@@ -0,0 +1,35 @@
id: jetson-orin-1
name: "Jetson Orin NX 16GB — Unit 1"
description: >
NVIDIA Jetson Orin NX 16GB edge inference node. 8-core Cortex-A78AE + Ampere GPU
(1024 CUDA cores, SM_87) + 16 GB unified LPDDR5. JetPack 7.0 platform.
Lightweight inference workloads — Gemma 3 4B/12B class models at INT4/INT8.
hardware:
cpu:
vendor: NVIDIA
model: "Jetson Orin NX (8-core Cortex-A78AE)"
cores: 8
threads: 8
arch: arm64
gpus:
- type: nvidia-ampere
model: "Jetson Orin NX Ampere (integrated)"
cuda_cores: 1024
compute_capability: "8.7"
memory_gb: 16 # unified with CPU
ram_gb: 16
unified_memory_gb: 16
storage: "NVMe (M.2)"
tags: [edge, jetson, orin, arm64, ampere, sm_87, unified-memory]
compose_overrides:
- docker-compose.arm64.override.yml
services:
- mesh-agent
- fleet-agent
tailscale:
role: edge-node
hostname_pattern: "pmoves-orin-1"
notes:
- "Unified 16 GB LPDDR5 — GPU and CPU share the pool. Target: Gemma 3 4B INT4, Gemma 3 12B INT8."
- "JetPack 7.0 (Ubuntu 22.04 base). Requires jetpack-specific container images."
- "Docker must use --runtime nvidia. Set in /etc/docker/daemon.json."
35 changes: 35 additions & 0 deletions pmoves/config/profiles/jetson-orin-2.yaml
Original file line number Diff line number Diff line change
@@ -0,0 +1,35 @@
id: jetson-orin-2
name: "Jetson Orin NX 16GB — Unit 2"
description: >
NVIDIA Jetson Orin NX 16GB edge inference node. 8-core Cortex-A78AE + Ampere GPU
(1024 CUDA cores, SM_87) + 16 GB unified LPDDR5. JetPack 7.0 platform.
Lightweight inference workloads — Gemma 3 4B/12B class models at INT4/INT8.
hardware:
cpu:
vendor: NVIDIA
model: "Jetson Orin NX (8-core Cortex-A78AE)"
cores: 8
threads: 8
arch: arm64
gpus:
- type: nvidia-ampere
model: "Jetson Orin NX Ampere (integrated)"
cuda_cores: 1024
compute_capability: "8.7"
memory_gb: 16 # unified with CPU
ram_gb: 16
unified_memory_gb: 16
storage: "NVMe (M.2)"
tags: [edge, jetson, orin, arm64, ampere, sm_87, unified-memory]
compose_overrides:
- docker-compose.arm64.override.yml
services:
- mesh-agent
- fleet-agent
tailscale:
role: edge-node
hostname_pattern: "pmoves-orin-2"
notes:
- "Unified 16 GB LPDDR5 — GPU and CPU share the pool. Target: Gemma 3 4B INT4, Gemma 3 12B INT8."
- "JetPack 7.0 (Ubuntu 22.04 base). Requires jetpack-specific container images."
- "Docker must use --runtime nvidia. Set in /etc/docker/daemon.json."
11 changes: 8 additions & 3 deletions pmoves/scripts/bootstrap-node.sh
Original file line number Diff line number Diff line change
Expand Up @@ -164,10 +164,15 @@ if ! docker info >/dev/null 2>&1; then
fi

if docker network inspect pmoves_bus >/dev/null 2>&1; then
ok "pmoves_bus network already exists"
actual_subnet=$(docker network inspect pmoves_bus --format '{{range .IPAM.Config}}{{.Subnet}}{{end}}' 2>/dev/null || echo "unknown")
if [[ "$actual_subnet" == "172.30.3.0/24" ]]; then
ok "pmoves_bus network already exists (subnet=$actual_subnet)"
else
warn "pmoves_bus exists but subnet mismatch: got=$actual_subnet expected=172.30.3.0/24"
fi
else
info "Creating pmoves_bus network (172.30.3.0/24)..."
if docker network create --driver bridge --subnet 172.30.3.0/24 pmoves_bus; then
info "Creating pmoves_bus network (172.30.3.0/24, internal)..."
if docker network create --driver bridge --subnet 172.30.3.0/24 --internal pmoves_bus; then
ok "pmoves_bus network created — subnet=172.30.3.0/24"
else
fail "docker network create pmoves_bus failed — check Docker daemon"
Expand Down
14 changes: 7 additions & 7 deletions pmoves/scripts/mesh-bind.sh
Original file line number Diff line number Diff line change
Expand Up @@ -3,7 +3,7 @@
#
# Verifies pmoves_bus exists with the correct subnet, runs a port-binding reality
# check, probes NATS reachability from inside the network, and publishes a
# mesh.node.online.v1 event. Safe to re-run at any time.
# mesh.node.announce.v1 event. Safe to re-run at any time.
#
# Usage:
# bash pmoves/scripts/mesh-bind.sh --node-id pmoves-rdna4 [--peer pmoves-spark]
Expand Down Expand Up @@ -169,7 +169,7 @@ else
dim "Ensure NATS is running on pmoves_bus and the container has service-name alias"
fi

# ── Step 4 — Publish mesh.node.online.v1 ─────────────────────────────────────
# ── Step 4 — Publish mesh.node.announce.v1 ─────────────────────────────────────

echo ""
echo "=== Step 4: Mesh Online Announce ==="
Expand All @@ -181,17 +181,17 @@ NATS_REACHABLE_FLAG="true"
PAYLOAD="{\"node_id\":\"$NODE_ID\",\"ts\":$(date +%s),\"nats_reachable\":$NATS_REACHABLE_FLAG}"

if command -v nats >/dev/null 2>&1; then
info "Publishing mesh.node.online.v1..."
info "Publishing mesh.node.announce.v1..."
dim "payload: $PAYLOAD"
if echo "$PAYLOAD" | nats pub --server "${NATS_URL:-nats://localhost:4222}" mesh.node.online.v1; then
ok "Published mesh.node.online.v1"
if echo "$PAYLOAD" | nats pub --server "${NATS_URL:-nats://localhost:4222}" mesh.node.announce.v1; then
ok "Published mesh.node.announce.v1"
ONLINE_PUBLISHED="yes"
else
warn "nats pub failed — NATS may not be running; mesh.node.online.v1 publish skipped"
warn "nats pub failed — NATS may not be running; mesh.node.announce.v1 publish skipped"
dim "Re-run mesh-bind once NATS is up to announce this node"
fi
else
warn "nats CLI not found — mesh.node.online.v1 publish skipped"
warn "nats CLI not found — mesh.node.announce.v1 publish skipped"
dim "Install nats CLI to enable automatic mesh online announce"
fi

Expand Down
Loading