Real-time Monad observability dashboard — testnet is fully live, mainnet runs in preview against public RPCs. Focus on the things other explorers don't show: parallel-execution metrics, a unified incident timeline, validator health scoring, network-wide decentralization insights, and a curated public-RPC catalog with live latency monitor.
🌐 Live: https://monad-tech.com · (legacy alias https://monad-tech.bee-hive.work)
Block explorers like MonadScan cover the basics (blocks, txs, gas). This dashboard adds the things operators and delegators actually need to see:
retry_pct— share of transactions re-executed per block due to parallel-execution conflicts. Unique to Monad's OCC engine. Surfaces which contracts break parallelism.- Block execution time breakdown —
state_reset/tx_exec/commitphases in microseconds. - Top contracts by retry rate — parallelism hotspots ranked.
- Unified incident timeline — reorgs · validator churn · retry spikes · block stalls · critical logs · 6 Monad-specific anomaly detectors, all in one chronological feed with persistence across restarts. Reorgs are enriched with replacement miner / tx count / detection lag, and merged with a 7-day InfluxDB history so the panel never blanks out after a deploy.
- Validator health score — composite of block-production, uptime, recency, with penalty for unregistered signers. Block attribution uses on-chain
ValidatorRewardedevents so operators with non-authAddress beneficiaries are still credited. - On-chain delegator lists — pulled directly from the staking precompile (
getDelegators(uint64,address), selector0xa0843a26) instead of block-scanning. Survives validator-set reshuffles and is correct on day one. - Network decentralization — Nakamoto 33/50/66 coefficients, peer geo distribution, client version tracking.
- Public-RPC catalog with live latency —
/tools/rpcspings 19 public Monad endpoints (testnet + mainnet) every 60 s, sorts by median latency, exposes "Add to MetaMask" per network.
The goal: be the single dashboard where an operator can diagnose "is my node lagging, or is the whole network halted?" and where a delegator can compare validators on dimensions that matter.
- Live KPIs: latest block, TPS, gas, block time, utilization
- Epoch progress
- Parallel execution panel: retry_pct avg/peak, effective TPS peak, stacked execution-time breakdown
- Top contracts by retry rate table
- TPS / Gas / Util chart with "pixelated" squared bars
- Latest blocks + transactions with pagination + search
- 309-row sortable table (197 in active set, 112 registered-but-not-producing or unregistered miners)
- Filter:
REGISTERED ONLYtoggle (hides block producers without a matched on-chain authAddress) - Score = 40% health + 40% uptime + 20% recency, with 0.7× penalty for unregistered
- Columns: moniker · address · stake · commission · score · health · uptime · blocks · share · txs · last-block
- Stake-weighted participation: short-window (
participationPct, 500-block sample) and long-window (participationLong, cumulative WS aggregator). Long-window is statistically stable after ~30 min of process runtime.
- Nakamoto coefficient breakdown
- Recent reorgs (usually zero — MonadBFT gives deterministic finality)
- Peer geo distribution (by country + ASN)
- Validator set changes log
- Unified feed of 13 event types:
- Standard: reorg · validator_added/removed/stake_decrease · retry_spike · block_stall · critical_log
- Anomaly detectors (Monad-specific, edge-triggered): state_root_mismatch · state_sync_active · consensus_stress · vote_delay_high · tip_lag · exec_lag
- Reorg cards now show replacement miner (with moniker), tx count, and detection lag, and merge in-memory ring with the last 7 days from InfluxDB
- Filter by severity (all / critical / warn / info)
- Filter by range (1h / 6h / 12h / 24h / 7d)
- Persistent: survives PM2 restarts (InfluxDB-backed)
/tools/rpcs— 11 mainnet + 8 testnet public endpoints from Foundation, Alchemy, Ankr, dRPC, OnFinality, Tenderly, thirdweb, bloXroute, Tatum, MonadInfra, Natsai- Live status: online dot, median latency over the last 5 samples, current tip block, WS support badge
- Per-network "Add to MetaMask" button using
wallet_addEthereumChain - Network switcher filters the catalog so the testnet tab only shows testnet RPCs and vice-versa
- Operator landing page with live infra telemetry
- Client version + sync status + peer count
- Commission / minimum-delegation details
- Delegate CTA (Twitter · Discord · website)
The hot path is push-based, not polling. A persistent WebSocket subscription to
monad-rpc's eth_subscribe(newHeads) fills a 1000-block ring buffer in RAM.
Every "live tip" endpoint reads from this ring at request time — zero RPC calls
per user request. (Migration story: see feat(rpc): migrate from polling to WebSocket push. Eliminated 30,000+ WARN/30min on monad-rpc's triedb_env channel.)
Monad validator
───────────────
monad-execution ──── shared-memory event ring (642 MB, hugepages)
│
monad-rpc :8081 ──── eth_subscribe(newHeads) bridge
│ ws://, push every ~0.4s
▼
┌──────────────────────────────────────────────────────────┐
│ monad-stats (this dashboard) │
│ │
│ lib/wsBlockStream.ts │
│ ├─ Map<blockNum, RingBlock> ≤1000 (RAM) │
│ ├─ Map<miner, MinerAggregate> (cumulative since boot) │
│ └─ enrichBlock: 1 eth_getBlockByNumber per push (smooth)│
│ │
│ Hot endpoints (all 0 RPC at request time): │
│ ├─ /api/blocks ← ring │
│ ├─ /api/transactions ← ring (full tx data) │
│ ├─ /api/stats ← ring tip + 1 eth_gasPrice │
│ └─ /api/top-contracts ← ring (5m), RPC fallback (15m+) │
│ │
│ Background pollers (smooth-paced single calls): │
│ ├─ reorg detector (4s, depth-15) │
│ ├─ TPS per-second collector (1s) │
│ ├─ validator-set tracker (60s) │
│ ├─ exec-stats writer → InfluxDB (30s, parses Loki logs) │
│ ├─ anomaly detectors (30s, Prometheus + 2 RPC) │
│ └─ peer-geo refresh (30min) │
└──────────────────────────────────────────────────────────┘
│ │ │
▼ ▼ ▼
Loki Prometheus InfluxDB
(logs) (otelcol) (persistence)
Hot-path diagnostic: GET /api/ws-state shows live ring/aggregator state.
Long-range historical endpoints — Loki for __exec_block parsing,
InfluxDB for chain-metric aggregates and persisted incidents.
- Next.js 16 (App Router, Turbopack, client + server components)
- TypeScript strict
- Tailwind v4 for base + inline styles for dark gold theme
- Recharts for charts
- Nginx reverse proxy + Cloudflare CDN
- PM2 process manager, fork mode
# 1. Install
npm ci
# 2. Environment — copy template and edit
cp .env.example .env.local
$EDITOR .env.local # at minimum set MONAD_RPC_URL
# 3. Dev mode (hot reload)
npm run dev
# → http://localhost:3000
# 4. Production
npm run build
npm run start -- -p 3001| Feature | Requires |
|---|---|
| Blocks / txs / TPS chart | MONAD_RPC_URL only |
retry_pct + top-contracts |
Loki with monad-execution journald logs |
| Persistent historical charts | InfluxDB 1.x reachable |
/node (internal dashboard) |
PROM_URL pointing at an otelcol-exposed endpoint + NODE_AUTH_* |
/beehive landing |
PROM_URL + optional BEEHIVE_* env |
Without Loki / InfluxDB the dashboard still works, just with reduced historical depth.
This repo runs in production behind Nginx + Cloudflare at https://monad-tech.com. The deploy pattern:
# On the server
cd /path/to/monad-tech
git pull
npm ci --omit=dev
npm run build
pm2 restart monad-statsNginx vhost example (with Cloudflare Origin Cert):
server {
listen 443 ssl http2;
server_name monad-tech.com www.monad-tech.com;
ssl_certificate /etc/nginx/ssl/monad-tech.com/cert.pem;
ssl_certificate_key /etc/nginx/ssl/monad-tech.com/key.pem;
# HSTS, CSP, rate-limit zones — see docs
location /api/ { proxy_pass http://127.0.0.1:3001; }
location / { proxy_pass http://127.0.0.1:3001; }
}Middleware in middleware.ts adds in-app rate limiting (300 req / min / IP). Cloudflare handles the outer layer.
- WebSocket push over polling. Hot endpoints (
/api/blocks,/api/transactions,/api/stats) read from a RAM ring buffer filled byeth_subscribe(newHeads), not by per-request RPC fetches. Eliminates burst patterns that overflow monad-rpc'striedb_envchannel. - WS aggregator for long-window stats. Each push updates a
Map<miner, MinerAggregate>in RAM. After ~30 min of process runtime this gives a 4500+ block window for stake-weighted participation calculations — variance ~1σ vs ~5σ on a 500-block window. - Staking-precompile slot 8 —
snapshotStakeper docs.monad.xyz/reference/staking/api. This is the canonical value gating active-set membership at epoch boundary. (Earlier code mis-named thistotalStake— display value is correct, but the field name confused readers.) - Unregistered signer handling — ~15 % of block producers use a separate signing
key, not their authAddress. We label them, don't hide them, and apply a 0.7× score
penalty. Their stake stays
null(we cannot resolve miner→authAddress without a registry of signing-key derivations). - InfluxDB dual-source —
/api/exec-stats≤ 15 min → Loki (freshest), > 15 min → InfluxDB (persisted). Writer polls every 30 s. - Anomaly detectors persisted to InfluxDB — 6 edge-triggered detectors
(state_root_mismatch, state_sync_active, consensus_stress, vote_delay_high,
tip_lag, exec_lag) write to
monad_anomaliesmeasurement. Survives PM2 restarts so the public IncidentTimeline doesn't blank out on deploys. - Per-network state isolation —
tipCache, validator registry, and moniker map are all keyed byNetworkIdso the mainnet preview cannot leak testnet data into a mainnet view (or vice versa). - Mainnet preview via public RPC — testnet talks to our own validator;
mainnet falls back to
https://rpc.monad.xyz. Mainnet-only sections that rely on data we cannot get from a public endpoint (parallelism, top contracts, retry charts, beneficiary attribution, our infra telemetry) render a "MAINNET — COMING SOON" placeholder rather than recycling testnet numbers.
- Tip-lag vs reference RPC — shipped as
tip_laganomaly detector - Stake-weighted participation metric — shipped (short + long window)
- Validator comparison tool (pick 2-5, side-by-side) — shipped
- Per-validator delegator list — shipped via
getDelegatorsprecompile (selector0xa0843a26) - Public-RPC catalog with live latency — shipped at
/tools/rpcs - Mainnet preview via public RPC — shipped (testnet stays on our validator)
- Beneficiary-based block attribution — shipped via
ValidatorRewardedevents - Reorg history persisted across restarts — shipped via 7-day InfluxDB merge
- Persist WS aggregator to InfluxDB so
participationLongsurvives PM2 restart - Telegram / Discord bot for critical incidents
- AS / ISP concentration detector → new incident type (mainnet, via Decentra)
- Light-theme toggle
- Public API docs (OpenAPI)
- Native mainnet validator (replace public-RPC fallback once BeeHive's mainnet node is live)
docs/METRICS.md— full reference of every metric, badge, chart and parameter shown on the site. Written for newcomers (with glossary) as well as operators / delegators.
MIT — see LICENSE.
Built and operated by BeeHive — a node-infrastructure team running validators across Lido, Obol, SSV, Mina, Provenance, Stellar, and Monad.
- Twitter / X — @BeeHive_NT
- Discord —
mav3rick_iphone