archetech / archon / 38331516593

93%

push

github

fix(hyperswarm): retry IPFS peers after transient failures (#1388)

* fix(hyperswarm): retry IPFS peers after transient failures

Since #994 the mediator recorded each announced peer as added before
resolving its DID, so one failed attempt left that node unpeered for the
life of the process. On a cold start, Gatekeeper stalls timed out every
lookup and the node ran with zero IPFS peers until restarted. A peer whose
DID document lacked node info was dropped silently, and a peer with no
dialable address was logged as added and announced to other nodes without
being peered.

Peer state now lives in a PeerTracker keyed by DID suffix. Resolution
errors, missing node info and unreachable addresses are retried with
exponential backoff (1 to 30 minutes); pings and the connection loop both
trigger due retries, one pass at a time. Only peered nodes are announced.
KuboClient.addPeeringPeer reports whether it wrote the peer, failure
reasons are logged on every attempt, and a mediator_peer_candidates gauge
shows peers by state.

Fixes #1378

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* fix(hyperswarm): keep announcing resolved peers and ignore self aliases

Review follow-up for #1388:

- Announce every DID that resolved to IPFS node info, including one this
  node cannot dial yet, since other nodes may reach it. This also keeps
  mediator_known_nodes counting resolved DIDs, as documented.
- Mark a DID that resolves to this node's own IPFS node as ignored instead
  of reporting it as peered and announcing it.
- Fall back when describeError cannot serialize a failure, so the retry
  is still scheduled.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>

3567 of 3984 branches covered (89.53%)

Branch coverage included in aggregate %.

3 of 3 new or added lines in 1 file covered. (100.0%)

9848 of 10136 relevant lines covered (97.16%)

22080.96 hits per line