Goldshell – Dashboard Flipping Between 0 H/s and Pool Error

Power-cycle for 60 s at the wall. Not a soft reboot — pull the barrel/C13, wait a full minute so the bulk caps drain, then re-energize. This clears wedged supervisor state and any half-open stratum sockets. If the flap was triggered by a transient (router reboot, ISP blip), this alone fixes it maybe 20% of the time.

Move the miner to ethernet. Unplug WiFi in Network Config, plug a Cat5e/6 into the RJ45, reboot. Goldshell's own firmware-upgrade KB explicitly says "use ethernet" — that same rule applies to steady-state mining on the older BOX hardware. Cold hard truth: Goldshell's WiFi radio is a weak-signal tell, not a reliable long-term link.

Rewrite the worker string. Log in at `http://<miner-ip>/` (default `admin` / `123456789`), go to Miner Config, overwrite: URL `stratum+tcp://dxpool.net:9008` (substitute your coin's endpoint), Worker `<account>.<worker>` (e.g. `alice.kdbox01`), Password `x`. No trailing spaces, no smart-quote dashes from a Word-pasted doc, no raw wallet address when the pool expects a registered account. Save, reboot, observe 15 minutes.

Verify you're on a Goldshell-supported pool. Goldshell firmware speaks clean Stratum V1. Dxpool and F2pool are the two it's tested against. NiceHash (requires extranonce.subscribe and specific diff params), Prohashing (requires `n=` and `o=` parameters), and custom stratum proxies often fail silently. If you're on any of those, fall back to Dxpool or F2pool as the diagnostic baseline. Return to your preferred pool *after* you prove the miner hashes cleanly for an hour.

DHCP-reserve the MAC. On the router admin, find the Goldshell's MAC (visible on `find.goldshell.com` or the router client list) and bind it to a fixed IP. Also set DNS to `1.1.1.1` and `8.8.8.8` as primaries. Reboot the miner. This kills the "pool DNS fails intermittently" and "IP changed overnight" flap classes in one shot — 3 minutes of work that prevents hours of head-scratching.

Flash the latest stable firmware — over ethernet. Grab the current release for your exact model from `github.com/goldshellminer/firmware`. Log in, System → Update Firmware, upload the `.bin`, do not touch anything for 5–10 minutes. After reboot, confirm version in System → Overview. Known-bad: KD-BOX `2.1.1` / `2.1.3` (GitHub Issue #50 — cold-air-from-fans regression). Known-bad for algo switching: anything pre-`2.2.3`. If the current build is one you're suspicious of, check the repo commit log for a pool/stratum change and consider rolling back one point release.

Measure PSU rail under load. Multimeter on DC, probe the 12 V barrel or PCIe 6-pin output *while the miner is in a hashing window* (not during a `0 H/s` flap). Expect ≥11.8 V sustained. Below = PSU brownout — swap with a known-good 750 W+ unit for KD-BOX Pro, 1800 W+ for LT5/LT5 Pro, 2200 W+ for LT6/CK6/HS5. Cheap Amazon PSUs age fast under 24/7 load; if you've been running the same PSU 18+ months under continuous duty, it's likely the problem.

Measure wall voltage during peak evening hours. Voltage sag when the neighbourhood A/C kicks on is a real phenomenon in Canadian suburban 240 V split-phase service and in US 120 V residential. Clip a voltage logger across the outlet for 24 hours. Dropout below 228 V on 240 V systems or 112 V on 120 V systems during a flap window is your electrical environment, not the miner. Fix: dedicated 240 V circuit (strongly preferred) or upgrade the branch circuit, not more miner tinkering.

Run a ping test to the pool from a same-LAN laptop. `ping -t dxpool.net` on Windows or `ping dxpool.net` on macOS/Linux for 10 minutes. Loss >1% or RTT spikes >200 ms = ISP or upstream issue. Flap stops = blame the ISP, not the miner. Some ISPs (looking at you, last-mile Canadian cable MSOs) route mining-pool traffic poorly at peak. VPN to a better exit is a workaround; better upstream is the fix.

Sanity-check worker password and difficulty. Some pools accept any password (`x` is standard), some accept `d=N` to set vardiff floor. On F2pool HNS, password `x` is correct. On Dxpool, password `x`. On NiceHash, you need the specific extranonce params — if you're determined to use NiceHash, use a stratum proxy between the miner and NiceHash to translate. Cleanest path is just to use Dxpool or F2pool.

Reboot the router and the miner together. Power off both, wait 90 s, power router on first, wait 2 min for DHCP/DNS to stabilise, power miner on. Flap stops = NAT table or DHCP lease churn was the cause. Consumer routers age badly and accumulate half-open connection-tracking entries; a monthly reboot is not a crazy maintenance item in a mining household.

SD-card recovery flash. For BOX miners (KD-BOX, Mini-DOGE, HS-BOX, ST-BOX, CK-BOX): if the flap survives every software fix and you suspect corrupted control-board firmware, pull the SD card from the back slot, image the correct recovery `.img` using Balena Etcher on a fresh microSD, reinsert, boot. Watch for a single-blue-LED flash-complete signal. Then log in, upload the latest stable firmware via the web UI. Recovery images aren't on GitHub — you have to email `hello@goldshell.com` for the burn image for your specific model. The James Chambers guide is the community canonical reference for the procedure. A D-Central workshop note: make sure the SD card is the correct class (class 10 or better) and not a counterfeit sourced from a sketchy marketplace; bad flash media is a recurring cause of "recovery didn't take."

Open the chassis and check hashboard connectors. Power off, unplug, wait 5 minutes. Torx T10 (occasionally T8 or Phillips on older BOX chassis) to remove the top cover. On LT5/LT6/CK6: unplug each hashboard ribbon and power connector, inspect for carbonisation, oxidation, bent pins, blackening. Reseat firmly. On BOX miners: the hashboard is soldered; the connectors to check are the ribbon to the control board and the power to the PSU-side. Cold boot, watch 15 min.

Isolate a bad hashboard. On multi-board models, disconnect boards one at a time (power off for each swap). If disabling board N stops the flap, that board is partial/dead. Zeus Mining, LYS-SZ, and BT-Miners all sell Goldshell test fixtures that scan the ICT560/KD5-chip/HS5-chip hashboards for bad chips. A D-Central ICT560 / chip-level repair on an LT5 hashboard recovers the board's full hashrate at a fraction of the full-board replacement cost.

Deep-clean dust. 50% of thermal complaints on KD5/KD6/LT5/LT6/CK6/HS5 in D-Central's repair queue are "they never cleaned the miner." Pull the fans (2–4 screws each depending on model), blow out the heatsink fins with compressed air, vacuum the intake grille, replace fans if bearings are whining. A dust-clogged heatsink pushes chip junction temp up, triggers the thermal watchdog, forces chip re-init, and the dashboard flaps to `0 H/s` every 2–5 minutes on a reliable clock. This alone fixes the flap on 15–20% of aged Goldshell miners.

Reflow / chip-level repair. If per-hashboard testing isolates 1–2 dead chips on an otherwise-good board (LT5/LT5 Pro ICT560, CK6 BM1485-class, HS5 custom), reflow the dead position with a preheat-plus-hot-air setup (preheat bottom ~150 °C, top-side 310–330 °C for ~30 s). If reflow doesn't revive, replace the chip from D-Central's salvaged-grade inventory. This is bench work — if you've never done it before, the board is worth more to you shipped to D-Central than attempted in a home workshop.

When to stop DIY. You've done Tiers 1–3, and: (a) a hashboard is isolated bad on the bench and needs chip-level repair, (b) the control board isn't holding a stable IP even after SD-card recovery, (c) you see burnt-component smell, lifted traces, or visibly damaged caps near the PSU-input or hashboard power path, or (d) you have >2 miners of the same model showing the same flap simultaneously (points at a batch firmware or hardware issue beyond DIY scope). Book a [D-Central ASIC Repair slot](https://d-central.tech/services/asic-repair/).

What D-Central does at the bench. ICT560 / KD5 / CK6 / HS5 chip-family test fixture runs per-chip isolation, maps failing positions, we reflow or replace, burn-in under real pool load for 24 hours, and only ship back once the miner holds steady-state at nameplate with `<1%` stale share rate. Control-board swaps use matched firmware bundles — wrong firmware on a late-rev control board is itself a brick event.

Ship safely. Anti-static bag on each hashboard, double-box with ≥5 cm foam on every side, include a note with: model, firmware version, pool and worker string, flap cycle period, symptom history. We reply in 1–2 business days with a repair quote. Canada / US / international welcomed. Ship-for-diagnosis is available for cases where you're not sure if repair is worth it vs. selling for parts.