[OisinKyne]

fix(network): Silver: info logs, HTTP readiness, refuse non-amd64 hosts

Found on the v0.18.0-rc0 lighthouse → Silver switch on Hoodi:

• Logs: Silver logs only errors unless RUST_LOG is set. Its first run was silent for 7 minutes, and a later
  9-minute run produced one line. Set RUST_LOG=info.

• Readiness: the ethpandaops chart's probes are tcpSocket. Silver's API accepted connections but answered nothing
  while it waited on a checkpoint, and the pod still showed 1/1 Ready, so the ethereum-beacon Service kept routing
  to it.

  • Readiness is now GET /eth/v1/node/health, which returns 206 while syncing and counts as ready. tcpSocket: null
    replaces the chart default rather than merging into it.
  • Startup and liveness stay as they are, so a long boot or catch-up isn't killed.

• Architecture: Silver publishes linux/amd64 images only, and the k3d node runs the Docker host's architecture.
  obol network install ethereum --consensus-client silver and obol network sync now refuse on any other host
  (e.g. Apple Silicon). The message names the other clients. Sync checks too, because values.yaml can be edited by hand.

Only applies when consensusClient: silver. A rendered lighthouse deployment has neither change.

Tested with go test ./internal/embed/... ./internal/network/... (including a checkClientArch table test), golangci-lint --new-from-rev clean,, plus helmfile template for silver (env and probe
as intended) and for lighthouse (unchanged).

🤖 Generated with Claude Code