Docs / Core concepts

Chain status

Each chain has one status value. It describes Farbeam's ability to read the chain, measured by this service. It is not a statement about the chain itself. A chain can be healthy while the RPC providers are not answering, and Farbeam reports degraded or down in that case.

#The rule

The state is computed each time you ask for it, in this order. poll is the chain's poll_seconds.

Order	State	Condition
1	`starting`	No poll has completed since the process started
2	`down`	Seconds since `last_ok` is greater than `max(poll * 6, 180)`
3	`degraded`	`consecutive_failures` is 2 or more, or `lag_seconds` is greater than `max(poll * 3, 90)`
4	`ok`	None of the above

With the current poll intervals, the max always gives the same numbers.

Chain	`poll_seconds`	`down` after	`degraded` when lag exceeds
Ethereum	12	180 s	90 s
Base	16	180 s	90 s
BNB Chain	16	180 s	90 s
Arbitrum One	16	180 s	90 s
OP Mainnet	30	180 s	90 s
Polygon PoS	30	180 s	90 s
Solana	30	180 s	90 s

#The fields behind the state

JSON

{
  "chain": "ethereum",
  "name": "Ethereum",
  "kind": "evm",
  "head": 26343493,
  "head_time": "2026-11-04T19:18:23.000Z",
  "lag_seconds": 6.4,
  "rpc_latency_ms": 602,
  "last_ok": "2026-11-04T19:18:29.782Z",
  "consecutive_failures": 0,
  "last_error": null,
  "indexed_since_start": 353,
  "status": "ok",
  "head_age_seconds": 15
}

This object is cut to the status fields. It was captured from GET /status on 2026-11-04 UTC.

Field	Meaning
`lag_seconds`	Age of the head block when it was read. It is not refreshed between polls.
`head_age_seconds`	Age of the head block at response time. It grows between polls, so 15 seconds is normal on a 12 second poll.
`last_ok`	Time the last poll finished without error
`consecutive_failures`	Polls that threw an error since the last good one
`scan_lag_blocks`	Blocks between the head and the log cursor, on the four EVM chains with an index. Absent elsewhere.
`last_error`	`upstream failed` when a source failed, `poll failed` for any other error

#What each state means for you

State	Stored data	New data	What to do
`ok`	Readable	Arriving	Nothing
`starting`	Readable	Not yet	Wait. This follows a restart and normally lasts for one poll.
`degraded`	Readable	May be late or partial	Treat recent gaps as possibly incomplete
`down`	Readable	Not arriving	Do not read silence as a quiet chain. Rows for this period will not be filled in later.

The log cursor catches up after a short outage. A down period longer than 2,000 blocks leaves a permanent gap, and so does a restart, because the cursor is held in memory. See indexing.

#Edge cases

**A fresh head with degraded.** The head is read before eth_getLogs. If the log call fails, the head still moves, but the poll counts as a failure. Two in a row give degraded even though lag_seconds is small. It means rows may be late.
**ok is not "complete".** An indexed chain can read ok while scan_lag_blocks is above 0, and Solana reads ok while sampling. The state speaks to the poller, not to coverage.
Chains with no index. OP Mainnet and Polygon have a status too. indexed_since_start stays at 0 by design.
**/health is narrower.** Service health returns status: "ok" whenever the process can answer, plus chains_ok and chains_total. Use chains_ok for a quick check, and use Status for detail.

#History

The service stores one sample per chain per 5 minute bucket for 30 days. Each sample holds the state at the moment of writing, lag_seconds and latency_ms. Read them with Status history, or see them on the status page. hours accepts 1 to 720.

Samples show measured state. They do not make up a published uptime figure, and Farbeam has no SLA. An SLA is Planned. measuring_since on /status gives the time of the first stored sample, so you can see how short the record is.

#Sources

GET /status also returns latest_event (the newest stored row per chain), watch_list (the Solana exchange watch list), and keyed_rpc_daily_limit_reached. It returns one row for each outbound source: name, ok, calls, failures, latency_ms, last_ok, last_error_at and error. The counters reset on restart, and a source appears only after its first call. ok reflects the last call only. See data sources for the list.

To build your own monitor, follow the status monitoring recipe.

PreviousIndexing NextFreshness