docs(node): document the Sei Labs seeds, retire the random peer list

[monty-sei]

Important

Draft — do not merge until sei-chain#3885 ships in a seid release.

This page tells operators that seid init populates bootstrap-peers automatically. That is only true once the release carrying sei-chain#3885 is out. Merging earlier documents behaviour the shipped binary does not have. Once that release lands, mark ready and merge.

The schema corrections below are correct today and independent of the release — they can be split out if we want them sooner.

Why

snippets/random-peers.jsx shipped 17 hardcoded peer addresses — mostly bare IPs of nodes we do not operate — and handed each reader a random five:

d53ab768…@45.250.253.40:51656
4ee8aea5…@202.8.8.183:11956
81409623…@57.128.230.96:51556
…

Those rot silently as operators re-key or renumber, and a wrong NodeID is rejected at the secret-connection handshake rather than degraded — so the failure is opaque to the operator following our docs. Handing out a hand-maintained peer list is also the practice we are trying to retire.

The Sei Labs seeds replace it: stable identities behind stable DNS, three per network across three regions, gossiping the rest of the network over PEX.

What changed

• snippets/seeds.jsx (new) replaces snippets/random-peers.jsx (deleted). Shows all three seeds — no shuffle, since the set is small and stable — and emits bootstrap-peers rather than persistent_peers.
• node/index.mdx — a new Peer discovery section explaining that init does this for you, with the manual values kept for nodes initialised before seeds shipped.
• node/troubleshooting.mdx — the peer-handshake section now points at the seeds and explains that expected to connect with peer X, got Y means the ID/address pair is wrong and should be replaced, not retried.

Schema corrections

The documented [p2p] block listed keys that do not exist in sei-tendermint. Verified against sei-tendermint/config/toml.go:

documented	actual
seeds	bootstrap-peers
persistent_peers	persistent-peers
max_num_inbound_peers, max_num_outbound_peers	max-connections, max-incoming-connection-attempts
allowed_pools	(no such key)
addr_book_strict	(no such key)

The last one matters beyond tidiness: it appears in the validator sentry-node example in node/validators.mdx, alongside four underscore-spelled keys. As written that config would not have applied. Fixed there too.

Note

llms-full.txt still contains the old content — it is a generated artifact (scripts/generate-llms.mjs) and will refresh on its next regeneration; I have not hand-edited it.

Why bootstrap-peers and not persistent-peers

Seeds populate the address book via PEX and may then be dropped. persistent-peers holds connections open indefinitely, which is only appropriate for nodes an operator runs themselves — and pointing every operator's persistent connections at our seeds would be a load multiplier for us.

[mintlify]

Preview deployment for your docs. Learn more about Mintlify Previews.

Project	Status	Preview	Updated (UTC)
sei-docs	🟢 Ready	View Preview	Aug 10, 2026, 12:56 AM

💡 Tip: Enable Workflows to automatically generate PRs for you.

[monty-sei]

Closing in favour of #96.

This PR predates the seed documentation that has since landed on main: technical-reference.mdx and validators.mdx already describe the auto-populated bootstrap-peers, and with more detail than this branch had (including the v6.6.2 version reference, which this one was missing). It had also fallen into conflict across all four pages.

It would also have regressed two values that main now has correct, max-connections (64 vs the actual 100) and handshake-timeout (20s vs the actual 10s).

#96 is built fresh off main and does only the part nobody has covered yet: retiring the random peer list, which currently serves 17 hardcoded community addresses of which only 6 still answer. The seeds.jsx snippet carries over from here unchanged.