0.6.0 — support Logos node 0.2.1 (breaking release)

[ibeezhan]

This is the 0.2.1 support update, opened against the official repo. (Earlier drafts lived on my fork ibeezhan/logosup as PRs #1/#2; #1 was built on a stale pre-0.2.0 base and reimplemented work already landed upstream in 0.5.0/0.5.1, so it was discarded.) This PR is the true incremental 0.2.0 → 0.2.1 delta on top of current main (4a600dc).

The real 0.2.0 → 0.2.1 delta

0.2.1 is another genesis reset. The release notes don't say so explicitly, but it's verified empirically: a freshly synced 0.2.1 node reports slot 532,277 at 1 s/slot, which back-computes to a genesis of 2026-08-05 09:01 UTC — four minutes after the release was published (08:57 UTC). The chainsync protocol ID is also versioned per release (/logos-blockchain-testnet-0.2.1/chainsync/1.0.0), so 0.2.0 and 0.2.1 nodes can't sync from each other.

What actually changed for logosup:

1. --ibd flag removed — replaced by opt-out --skip-ibd (IBD from -p peers is now default). Passing --ibd hard-fails. Dropped from both init-config and migrate-from-0.1.2 invocations.
2. init-config now refuses to run when keystore.yaml exists (Keystore file exists. Use 'update' command.) — every 0.2.0 operator has one, so the old wipe-and-init reset flow would die. New docker_update_config() uses the node's update-config subcommand, which regenerates the config from the existing keystore and preserves wallet key identities across the reset (verified: keystore untouched, same known_keys, new 0.2.1 config fields present). cmd_reset now picks between migrate-from-0.1.2 / update-config / fresh init-config.
3. /cryptarchia/info schema changed — "mode":{"Started":"Bootstrapping"} became scalar "state":"Bootstrapping" plus top-level "phase" (InitialBlockDownload → ProlongedBootstrapPeriod). Without this fix, status/start show Mode: (unknown) and the Grafana consensus-mode panel goes dark. Updated cmd_status, cmd_start, and the monitoring exporter (old shapes kept as fallbacks).
4. 0.2.1 added to LOGOS_BREAKING_VERSIONS so logosup update diverts to the migration flow.
5. Bootstrap peers verified unchanged — same four fleet multiaddrs as 0.2.0 (comment updated in network.yml).
6. VERSION 0.5.1 → 0.6.0.

Explicitly not changed (verified identical at 0.2.1): release asset naming (logos-blockchain-node-linux-{arch}-{version}.tar.gz, circuits still bundled), ports (blend udp/3400 existed in 0.2.0), faucet/dashboard URLs, /network/info fields (n_peers/n_connections still present).

Drive-by fix (separate commit): the 0.5.1 status diagnostics used grep -c ... || echo 0, but grep -c prints 0 itself and only exits 1 — so any firing diagnostic corrupted counts to "0\n0" and sprayed (( : syntax error into status output. Visible on every 0.2.1 status call because stale 0.2.0 peers trigger the protocol-mismatch warning.

Testing (macOS arm64, Docker)

• docker build of the 0.2.1 image (aarch64 asset downloads and unpacks; same URL pattern) ✅
• bash -n + shellcheck on all touched scripts (no new warnings vs main baseline), py_compile on the exporter ✅
• init-config (fresh) and update-config (from a real 0.2.0-generated config + keystore) both produce configs that pass --check-config ✅
• Booted both configs against the live 0.2.1 testnet: node enters Bootstrapping/InitialBlockDownload, IBD pulls blocks (height 0 → 17k+ in ~1 min), then ProlongedBootstrapPeriod, 46 connected peers ✅
• logosup status against the running 0.2.1 container: Mode/phase/slot/height, network info, wallet keys with roles, clean diagnostics ✅
• Not tested: full sync to Online (operator will finish on the Raspberry Pi), monitoring stack end-to-end.

🤖 Generated with Claude Code