ASIC Miner – DHCP Failure

Power-cycle the miner at the wall for a full 30 seconds (not a soft reboot), then power on and wait 90 seconds. Many DHCP failures are stale state — the miner's `udhcpc` client got wedged on a previous bad lease. A hard cold boot resets the `eth0` interface and any init state. 30 seconds is the minimum to let electrolytics fully discharge.

Reboot your router, waiting 60 seconds between power-off and power-on. Router DHCP daemons (`dnsmasq`, `udhcpd`, ISP custom builds) accumulate stale state across weeks of uptime. A reboot clears the lease table, resets the UDP packet queue, and forces fresh negotiation. Do this before touching settings — it fixes roughly 40% of `ERR_DHCP` tickets on its own.

Widen the router's DHCP pool to at least 2× your current device count. Log into the router admin UI (usually 192.168.1.1), find LAN → DHCP Settings, and expand the range (e.g. `192.168.1.50` to `192.168.1.254`). Save and reboot. Budget routers default to 32–100 addresses; racks need the full `/24` pool minus static-infrastructure reservations.

Disable MAC filter / access control on the router if enabled. Look under Security, Parental Controls, Access Control, or Device Approval. If you want to keep the allowlist, add the miner's MAC manually — it's printed on a label stuck to the control board. ISP routers (Bell HomeHub, Rogers Ignite, Comcast xFi) ship with this feature enabled by default.

Give the miner a static DHCP reservation. In the router's DHCP settings, bind the miner's MAC to a specific IP (e.g. `192.168.1.240`). This bypasses the pool-exhaustion race and makes the miner findable at the same address forever. Document the reservation in a text file — six months from now when you add miner 10, you'll thank yourself.

Swap the ethernet cable with a known-good Cat5e or Cat6 patch lead. The most under-diagnosed DHCP failure is a physically marginal cable — kinked, pinched by a rack door, crushed by a vibrating chassis, or with a popped RJ45 boot. Cables that link at gigabit for a laptop can fall back to 10 Mbps half-duplex for an ASIC, which kills DHCP. Test with a $20 cable tester or swap blind.

Force the switch port to 100 Mbps full-duplex. On managed switches (TP-Link Omada, UniFi, MikroTik), set the port to `100-full`, not auto. Older Antminer control boards (S9 BeagleBone, S17 early revisions) auto-negotiate badly with modern gigabit gear and end up in a half-duplex state where DHCP packets collide and drop. An unmanaged 100 Mbps switch between router and miner is the low-tech version.

Check the switch's MAC table and error counters. On a managed switch, `show mac address-table` (Cisco) or the equivalent UI page tells you if the miner's MAC is being learned at all. `show interface eth X` should report zero or near-zero `CRC errors`, `runts`, and `collisions`. Non-zero values = cable or port hardware issue, not DHCP.

Run a packet capture on the LAN side. From a Linux/macOS laptop on the same subnet: `sudo tcpdump -i eth0 -n -e -vv 'port 67 or port 68'` and power-cycle the miner. You should see `DHCPDISCOVER` from the miner's MAC followed by `DHCPOFFER` from the router. `DISCOVER` but no `OFFER` = router rejecting silently. No `DISCOVER` at all = miner's DHCP client is broken or L2 is dead.

Check for IP conflicts. Power off the miner, log into the router, and note any lease assigned to the IP range you expect. `arp -a` on a LAN laptop shows active hosts. If another device is squatting on what would be the miner's IP, evict it with a reservation or widen the pool. IP conflicts silently break DHCP `REQUEST` acknowledgement on some router firmwares.

Attach a serial console to the control board (Antminer: `115200 8N1` on the 4-pin debug header; Whatsminer: USB-UART on the console port). Power-cycle. Watch for `udhcpc` or `dhclient` messages. Healthy output: `sending discover` → `sending select for 192.168.x.x` → `lease obtained`. Failure: `no lease, forking to background` looping every 30 s. Serial tells you exactly which packet fails.

Write a static IP config to SD card (Antminer). Format a micro-SD ≤ 16 GB to FAT32. Create `static-ip-conf.txt` with `ipaddress=192.168.1.200`, `netmask=255.255.255.0`, `gateway=192.168.1.1`, `dnsservers=1.1.1.1`. Power off, insert SD, power on. Bitmain firmware applies this on boot. Bypasses DHCP entirely — clean proof the rest of the miner works, and a permanent workaround.

Use the manufacturer's static-config tool. Whatsminer: front-panel menu → Network → Static IP. Avalon: AUC controller web UI → Miner settings. Innosilicon: `setip.exe` Windows tool scans the subnet and assigns static IP via the Innosilicon discovery protocol. Bitaxe / NerdAxe / NerdQAxe: AxeOS web UI → Settings → Network → Static IP. Each ecosystem has a flavour; static IP is always possible.

Flash DCENT_OS on Antminer hardware for better network diagnostics. DCENT_OS is D-Central's own open-source Antminer firmware — per-chip HW%, tuning, autotuning, stratum v2, and a materially better network stack than stock Bitmain. Exposes `dhclient` logs in the web UI and supports mDNS (`miner.local`), making `ERR_DHCP` recovery faster. Alternatives: Braiins OS+, LuxOS, Vnish.

Run a local DHCP server on a laptop to isolate fault domain. Linux: `sudo dnsmasq --no-daemon --interface=eth0 --dhcp-range=192.168.99.100,192.168.99.150,12h --dhcp-option=3,192.168.99.1`. Windows: `Tftpd32` with built-in DHCP. Connect miner directly via ethernet to the laptop NIC. If it gets a lease from the laptop, miner is healthy, upstream is at fault. If it still fails, the miner's NIC or control board is the failure.

Stop DIY when the control board ethernet PHY is suspected dead: no `LINK` LED regardless of cable or port, serial console shows `eth0: PHY not found` or `eth0: link down` permanently, miner otherwise healthy (boots, hashes with static IP on some models). Control-board-level repair — Microchip/Realtek/Broadcom PHY chip dead, RJ45 magnetics fried, or crystal oscillator failed. Book a D-Central ASIC Repair slot.

D-Central bench process: oscilloscope probe on the PHY's TX/RX differential pairs, magnetics continuity test, crystal frequency check with a frequency counter. Dead PHY = desolder and replace with salvaged-grade or new-old-stock part (common PHYs: KSZ8081, LAN8720, RTL8201F). Toasted RJ45 magnetics = full jack replacement. Post-repair, verify DHCP handshake against our lab router with a packet capture as proof-of-fix.

Ship safely. Wrap the control board in anti-static foam, double-box with ≥ 5 cm foam on every side, include a note with observed symptoms (LED behaviour), firmware version, diagnostic steps already tried, and contact info. Every minute of diagnostic time saved = money off your repair bill. If shipping the full miner, remove hashboards and PSU to reduce weight unless we've asked for the full unit for fault reproduction.