ADI Docker

Docker Compose deployment for an ADI Chain mainnet external node.

This is adi-docker v0.1.0

What it runs

ADI is a zkSync-based zk-rollup L2. The external node is read-only — it syncs over P2P (devp2p) from peers and serves a standard JSON-RPC interface. Single container:

Service	Image	Role
adi	harbor.sde.adifoundation.ai/ghpc/adi-foundation-labs/server:v0.20.12-b1	ADI external node — JSON-RPC + WS on :3050, status on :3071, P2P on :3060 (TCP+UDP), metrics on :3312

As of v0.20.12, ENs sync via P2P instead of downloading HTTP block replays, and no longer use proof storage — the old proof-sync sidecar (azcopy against Azure Blob) is gone.

• Upstream setup script: https://github.com/ADI-Foundation-Labs/ADI-Stack-EN-Setup-script
• ADI docs: https://docs.adi.foundation/

Hardware

Component	Recommended
CPU	16 cores
RAM	32 GB
Storage	500+ GB NVMe

Initial sync takes roughly 1 day.

Required configuration

GENERAL_L1_RPC_URL must point at an archive Ethereum L1 RPC. Pruned L1 endpoints panic on startup with state at block is pruned. In production this is set per host in the Ansible inventory; for local runs, export it before ./adid up.

EXTERNAL_NETWORK_SECRET_KEY (P2P node identity, openssl rand -hex 32) and BOOT_NODE_URLS (peer discovery, provided by ADI) are required for the node to sync at all as of v0.20.12. The secret key is unique per node and must be set explicitly in production — this compose does not run ADI's external-node.sh, which normally auto-generates one on first start.

Quick start (local / non-Ansible)

cp default.env .env
# edit .env: at minimum set GENERAL_L1_RPC_URL to an archive Ethereum RPC
./adid up -d
./adid logs -f adi
# wait for the healthcheck to flip from starting to healthy
./adid check-sync

Operational commands

Command	Action
./adid up [-d]	Start the stack
./adid down	Stop and remove containers (volume preserved)
./adid logs [-f] [service]	Follow logs (service: adi)
./adid version	Print container image versions
./adid check-sync	Compare local block height against https://rpc.adifoundation.ai, also asserts eth_syncing=false
./adid update	Rebuild env from default.env and pull updated images
./adid terminate	Stop and destroy all data volumes (irreversible)

Compose overlays

• default.env ships with COMPOSE_FILE=adi.yml:rpc-shared.yml so the local dev workflow can hit http://127.0.0.1:3050.
• Production (via cmf-ansible-inventory) overrides this to COMPOSE_FILE=adi.yml:ext-network.yml — no 127.0.0.1 binding; traffic comes in over Traefik vhosts.

Ports

Port	Purpose	Exposed by rpc-shared.yml	Notes
3050	JSON-RPC HTTP + WebSocket (ADI multiplexes both)	yes, 127.0.0.1	Public over Traefik in prod
3060	P2P (devp2p), TCP+UDP	no, published directly in adi.yml	Must be reachable from the internet — not proxied by Traefik
3071	Status / health server	no	Host-internal only
3312	Prometheus /metrics	no	Scraped by prom_cluster on the same host

Image pinning

All image tags are pinned in default.env. Bump deliberately; do not use latest. The external node image is currently pinned to v0.20.12-b1 for mainnet.

Troubleshooting

Symptom	Likely cause	Fix
Failed to load genesis upgrade transaction: ... state at block is pruned panic	GENERAL_L1_RPC_URL points at a non-archive Ethereum endpoint	Replace with an archive L1 RPC
Node never logs Connected to peer <enode-id>	Port 3060 not open (firewall/security group), or BOOT_NODE_URLS/EXTERNAL_NETWORK_SECRET_KEY unset	Check ufw/security group for 3060 TCP+UDP; verify both env vars are set
Healthcheck stays starting for several minutes	Slow disk, insufficient RAM, or L2 replay lag from cold genesis	./adid logs adi; if RocksDB is busy persisting blocks, just wait.
./adid logs adi shows Connection refused to general_l1_rpc_url	Inventory L1 endpoint down or wrong URL	Verify the L1 host is reachable; check ansible secret rendering