Migrating cluster pairings from v2.9 to v2.10
v2.10 replaces the global “shared cluster token” model with per-peer shared secrets that are exchanged automatically during pairing. This means:
- The “Use cluster’s token” step is gone — every fresh pairing is one-shot via a 5-minute single-use pairing code.
- Revoking a peer no longer invalidates anyone else’s secrets.
- Existing v2.9 pairings keep working until you re-pair (the legacy cluster_token is still accepted as a fallback during the migration window).
When to migrate
You can migrate at your own pace. Each v2.9 pairing is flagged migrationRequired in the Cluster page. The flag is informational — the pair still works — but you should re-pair to switch to per-peer secrets and unlock per-peer revocation.
Re-pair workflow
For each peer pair:
- Open
Maintenance → Clusteron either peer. - Click Issue pairing code. A short 8-character code appears.
- On the other peer, open
Maintenance → Cluster → Add peer. - Paste the URL of the first peer + the pairing code.
- Click Pair. Both sides install a fresh per-pair secret and the
migrationRequiredflag clears within seconds.
That’s it. The old pair’s shared_secret slot fills in; nothing else needs to change.
What if I rotate the global cluster_token?
After v2.10, rotating the global token only affects:
- New peers being onboarded via “Use cluster’s token” (legacy fallback for v2.9 setup; deprecated).
- Pairing codes minted FROM the global token — they’ll stop validating immediately.
It does NOT affect already-paired peers that hold per-pair secrets.
Removing legacy support
The legacy global-token fallback in HMAC verification stays on for the v2.10 release cycle. v2.11 will remove it; any peers still flagged migrationRequired at that point will fail to authenticate cross-peer calls and need to re-pair.
Other v2.10 changes you’ll see
- Real-time WS push — paired peers maintain a persistent
/ws/clusterlink so catalog updates appear in the gallery in under a second instead of waiting on the 30-second polling cycle. - LAN auto-discovery — peers on the same network broadcast their presence on UDP port 28910. The Cluster page surfaces “Discovered” peers; click Pair to skip pasting the URL.
- Backup peer + failover — group editor gains a Backup peer dropdown. If the owner has been silent for 5 minutes, the backup takes over downloads automatically.
- Cross-peer file delete — sweep’s “resolve conflict” now actually deletes the loser on the remote peer (was queue-only in v2.9).
- Live config sync — set
cluster.replicate.<key>toclusterin advanced settings to mirror that setting across paired peers. - Cluster-wide search — gallery search fans out to every paired peer and merges results.
- Relay-through-peer — if peer A can’t reach C directly but B can, A’s calls to C are forwarded through B with end-to-end signing (B can’t tamper).