The reconciler doesn't take the RPC layer's per-app FIFO op lock (known
limit of 891cbba4): between a restart worker's stop and start halves it saw
the mempool frontend "missing", repair-recreated it behind systemd's back,
killed the worker's fresh container 11s after start, and left the unit down
for ~3.5 min until the next heal — gate test 123 measured exactly that
window (.228 iteration 3, 2026-07-09).
New crate::app_ops module owns the op-lock registry + stack member table
(runtime.rs and dependencies.rs now delegate) so the reconciler can probe
lifecycle_op_in_flight(app_id) — covering both the app's own key and its
owning stack package — and skip that app for the cycle. The ownership-sweep
podman restart gets the same guard. Health monitor is a follow-up.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
mempool-api exits at startup while electrumx is catching up to the daemon
(7-day initial-sync ETA on .228), flapping through 'activating' for ~2 min
after each lifecycle cycle before systemd heals it. The instant asserts
probed inside that window both iterations. Poll up to 180s for
converges-to-active — a genuine crash-loop still fails.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
lnd resolves the bitcoin RPC address once at startup and never re-resolves;
when bitcoin-core/knots restarts (new container IP) lnd spins on "dial tcp
<old-ip>:8332: no route to host" and its own RPC wedges until lnd restarts
(gate lnd getinfo test 75 failed every iteration on .228, 2026-07-09;
hardening plan §C cascade item). electrumx is already restarted by the
version-switch path — lnd was the only dependent left stranded.
After a successful package.start/restart of a bitcoin backend, restart
running, not-user-stopped address-caching dependents under their own app-op
lock (lock order backend→dependent only, so no deadlock).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
wyoming-piper/wyoming-whisper are being integrated and are not yet part of
the platform contract; their hand-staged units (companion-style
Restart=always) failed gate test 121 on .228. Exclude wyoming-* from
backend_quadlet_units() until their packaging lands.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
package.stop/restart on orchestrator-managed stacks (immich, indeedhub,
btcpay, netbird, mempool) addressed live CONTAINER names; unknown app ids
fell through to raw podman stop/start, racing systemd's --rm cleanup of the
quadlet unit — the container vanished on stop and the start half found
nothing (immich RESTART FAIL + indeedhub stuck restarting, gate 2026-07-09).
Now stack ops resolve member APP ids (reverse start order for stop) so the
orchestrator drives the quadlet .service; user-stop markers are written for
member app ids too (covers absent containers), and the orchestrator stop
fan-out treats an already-missing container as stopped.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
All three landed 2026-07-09 (dd3afbba, c3f0a306, 2683ad4f0). Files the
repair-wait user-stop preemption as the open follow-up of the stick fix.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
/var/log/archipelago/container-installs.log has been 0 bytes since April:
the service sandbox (ProtectSystem=strict class) leaves /var/log
read-only for the unit, the open() fails, and install_log() drops every
line by design (fire-and-forget). These START/STOP/RESTART/FAIL
breadcrumbs are the primary forensic trail for gate failures, so mirror
every line to tracing (journald) unconditionally and keep the file
append as best-effort for hosts where the path is writable.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
backend_quadlet_units() enumerates every *.container file on disk and the
active-state tests asserted each unit active + container running. A
user-stopped app keeps its unit file (e.g. the inactive half of the
bitcoin-core/bitcoin-knots multi-version pair after tonight's crash-loop
wave left core's unit on disk), so its inactive service / absent container
is the CORRECT state — gate run E false-failed on it (tests 123/124,
.228 2026-07-09). Honour the orchestrator's user-stopped.json intent
marker and skip those units in both tests.
Verified on .228: bitcoin-core's lingering unit now exempted; suite parses
(bats --count = 6) and the two fixed tests pass user-stopped filtering.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
One legacy app per gate run stuck at 'stopping'/'starting' past the test
window while its container was already settled (vaultwarden:stop run C,
jellyfin:stop run D, uptime-kuma:start run E — .228, 2026-07-09). The
package scanner already sees the truth every 60s but
merge_preserving_transitional refused to report it until the RPC worker
wrote the final state, and the workers can legitimately trail the
container by minutes:
- stop workers queue behind the orchestrator per-app lock, which the
reconcile host-port repair path holds through multi-minute stability
waits (repair_manifest_host_ports_after_stability: 5s + 5s probe +
restart + 60-420s port wait + 15-90s stable-running);
- start workers hold Starting through the full readiness wait —
host_port_wait_timeout_secs is 420s for uptime-kuma (HTTP probe),
longer than any UI/test patience — and the Starting stuck-timeout is
the 20-minute INSTALLING one.
New merge rules, both truth-driven by podman's live view:
- (Stopping, Stopped) + user-stop marker → Stopped: the scanner only
normalizes exited→Stopped for user-stopped apps, so this is exactly
"the user asked for a stop and the container has exited".
- (Starting, Running) → Running: the start visibly succeeded; live
health readings are merged separately and keep reporting readiness.
Restarting is deliberately NOT resolved by a Running scan — mid-restart
readings are the pre-stop container.
Tests: merge_tests 13/13 (4 new).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
package.start/stop/restart reply immediately and run their multi-container
sequences in spawned tasks with no coordination. Back-to-back RPCs on the
same app interleave those sequences: gate run D (.228, 2026-07-09 04:28)
fired stop→start→restart on mempool; the start's member bring-up raced the
still-running stop's member shutdown — archy-mempool-db was stopped 2s
after it started and the stack finished with ZERO containers (tests 95/123/
124). A user double-clicking restart in the UI can do the same.
Workers now take a per-app tokio Mutex (fair/FIFO, keyed by the normalized
orchestrator app id) as their first await, so queued ops run in RPC arrival
order and the final state matches the last request. The RPC reply stays
immediate.
Known limit: container lists are still computed at RPC time (pre-lock), so
a stop landing mid-start can still return 'No containers found' — it can no
longer interleave/corrupt, only error. Reconciler/health-monitor restarts
don't take the lock yet (separate paths, same candidate follow-up).
Also from run D, filed separately: jellyfin/vaultwarden (legacy containers
on a quadlet node) stop fine but the package state sticks at 'stopping' —
'Failed to stop jellyfin.service: Unit not loaded' error path suspected.
Tests: runtime 9/9; archipelago builds clean.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The dd61a204 bind hardening published Bitcoin RPC on the archy-net gateway
10.89.0.1 — but rootlessport binds in the HOST netns, where that address
does not exist. First real deploy (.228, 2026-07-09) crash-looped
bitcoin-knots AND bitcoin-core the moment the gate's stop→start regenerated
the unit from the re-signed catalog: 'rootlessport listen tcp
10.89.0.1:8332: bind: cannot assign requested address', restart counter 132.
Hand-edits to the unit don't survive — the orchestrator regenerates it from
the signed catalog manifest within seconds.
- New archipelago_container::manifest::host_can_bind_publish_ip(): empty/
wildcard/loopback accepted without probing, anything else ephemeral-bind
probed. Applied at all three publish paths — quadlet from_manifest
(PublishPort), podman API create (host_ip), and the legacy -p string
table loop — each dropping the publish with a warn instead of taking the
container down. This neutralizes the bad binds already in the SIGNED
catalog, so nodes recover on binary deploy alone (no ceremony needed).
- Remove the gateway publishes from bitcoin-knots/-core manifests and the
legacy config.rs tables: verified on .228 that every in-node consumer
(lnd, btcpay/nbxplorer, fedimint, mempool-api) dials the container's
archy-net alias directly (bitcoin-knots:8332) and lnd uses RPC polling
(no ZMQ) — the gateway publish had zero consumers. Loopback-only RPC/ZMQ
(the approved LAN lockdown) stands; P2P 8333 stays public.
Catalog still ships the gateway binds until the next signing ceremony
regenerates it from these manifests; the guard makes that non-urgent.
Tests: container crate 65/65 (2 new guard tests), quadlet 40/40.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Second instance of the stop→start destruction class from tonight's gate
runs (.228): package.stop of a multi-container stack marks every member
user-stopped and quadlet-removes their containers; package.start's fallback
then resurrected only the bare package id, leaving the other members absent
AND user-stopped — indeedhub lost all 7 containers this way (mempool needed
its umbrella alias for the same reason).
- New stack_member_app_ids map (real manifest app ids, dependency order) —
used by the start fallback when a stack has no live containers, and by
package.restart instead of hard-failing "No containers found".
- orchestrator_uninstall_app_ids grows the missing indeedhub arm (same
reconciler-ghost rationale as the immich arm).
- do_orchestrator_package_start no longer aborts the whole sequence when one
member's unknown-app-id fallback fails (the #73/#74 wish) — collects and
continues.
NOTE: extends the known §G per-app-map debt (tracker updated) — the proper
home for stack membership is a manifest field; parked post-tag by design.
Tests: package:: 48/48 incl. new fallback + indeedhub-uninstall coverage.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Regenerated from disk manifests and signed by the release root (verified
against the pinned anchor). Carries: bitcoin-knots/-core RPC/ZMQ bind
hardening (dd61a204), immich ownership-fix capabilities, lnd templated
BITCOIN_HOST, bitcoin:archival dependency tags for electrumx/mempool stack.
Nodes pick this up on their next catalog refresh; running bitcoin containers
converge on their next restart (restart-sensitive guard) — restart lnd after.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
USER DECISION 2026-07-08: accept breaking external wallets pointed at
nodeIP:8332. The 0.0.0.0 publish exposed auth-only RPC (and unauthenticated
ZMQ 28332/28333 on the legacy path) to the whole LAN.
- New `bind` field on manifest port mappings (validated as an IP; the same
host port may repeat with distinct binds). Rendered as
PublishPort=<ip>:<host>:<container> in quadlet units and host_ip in the
podman API create — unbound ports render byte-identical to before, so no
fleet-wide false-drift wave.
- bitcoin-knots/-core manifests publish 8332 on 127.0.0.1 + 10.89.0.1 only.
In-node consumers (lnd, fedimint-gateway, btcpay/nbxplorer) are unaffected:
they dial host.archipelago / host.containers.internal, which the
orchestrator pins to the archy-net gateway 10.89.0.1. P2P 8333 stays public.
- Legacy config.rs port strings get the same treatment incl. ZMQ.
Tests: new quadlet bind-render + manifest bind-validation tests;
container:: suite 167/167, archipelago-container 63/63.
DEPLOY NOTE: PublishPort strings change for bitcoin containers → one
planned recreate per node; restart lnd afterwards (it caches the backend
IP — see the backend-recreate cascade tracker item). Catalog manifests for
bitcoin apps must be regenerated + re-signed for catalog-covered nodes.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
lnd.bats retry loop had no per-exec timeout — a wedged lnd RPC (chain-blind
after bitcoin-knots was recreated under it) hung one podman exec, and the
whole gate, indefinitely (.228, 30+ min at test 75). timeout 10 per attempt
keeps the intended ~4min bound. Tracker: new §C item for the root cause —
apps cache their backend container's IP; a backend recreate must cascade a
dependent restart/alert or lnd goes chain-blind silently.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Found by the .228 lifecycle gate (2026-07-08, first quadlet-mode run):
stop→start on the legacy `mempool` umbrella app id destroyed the mempool
deployment. Under quadlet, package.stop removes the containers; package.start
then failed with `unknown app_id: mempool` because load_manifests drops the
umbrella manifest whenever the three split-stack members are loaded — leaving
nothing to recreate from and the stack down.
Two fixes:
- prod_orchestrator start/stop/restart now resolve the historical
`mempool`/`mempool-web` id to the split members (archy-mempool-db,
mempool-api, archy-mempool-web) whenever the umbrella manifest was dropped —
the same alias install already implements ("installing mempool assembles the
split stack"). Restart falls back to start for members whose container/unit
is gone. Legacy umbrella-only nodes are unaffected (alias inactive while the
umbrella manifest is live).
- is_missing_container_error (3 sites) now recognizes podman 5.x's
`no such object` inspect phrasing, so a genuinely absent container is
classified as missing instead of surfacing as a hard inspect error.
Tests: 2 new alias tests + 2 classifier tests; prod_orchestrator suite 61/61
green.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
- multinode suites no longer carry real node passwords as defaults: *_PW
vars are required, auto-loaded from git-ignored tests/multinode/.env
(lib sources it; .env.example documents the shape). Suites fail fast
with a clear message when unset.
- node_login/node_rpc get --connect-timeout 10 --max-time 120 (override:
MULTINODE_RPC_TIMEOUT) so one hung server-side RPC can't stall the whole
suite. Verified live: login+rpc against .116 OK; unroutable node fails
in 10s instead of hanging.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
- create-release.sh: assert the freshly built frontend dist actually embeds
the bumped version before packaging — the ui-dist-version guard in
tests/release/run.sh is behind --with-build, which the release path never
passes, so a silently no-opped npm build could ship a stale dist with a
valid sha256. Verified against the real dist (passes on current version,
trips on a missing one).
- publish-release-assets.sh: verify published assets by downloading and
comparing sha256 against the manifest, not just Content-Length — a
size-correct/content-wrong mirror asset now fails the publish gate.
Verified live against the v1.7.99-alpha assets on the vps2 mirror.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
- operations-runbook: redact the plaintext sudo password (rotate it — it
stays in git history); replace the nonexistent test-cross-node.sh /
test-reboot-survival.sh with the real entry points (lifecycle gate,
multinode smoke, e2e/post-install scripts)
- hotfix-process: de-hardcode the v1.0.x framing, point releases at
create-release.sh + primary Gitea, correct the rollback backup paths
(/opt/archipelago/rollback + updater backup dir, not /usr/local/bin)
- troubleshooting: connectivity probe start9.com -> debian.org
- INSTALL.sh: drop removed endurain app, point at developer-guide.md
(development-setup.md never existed)
- tests/lifecycle/TESTING.md: replace the stale 2026-06-21 mid-session
resume block with a resolved note; release-gate item 8 reflects the
2026-07-08 version decision (1.8.0-alpha)
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Second false-drift source found live on .228 right after the entrypoint
split fix: a YAML `>-` script with more-indented continuation lines
keeps literal newlines in custom_args, but shell_join flattens them to
spaces when writing the quadlet Exec= line, so Config.Cmd stores spaces
where the manifest has newlines. bitcoin-knots and fedimint-gateway
read as permanently command-drifted over line breaks alone (only the
restart-sensitive guard kept them from recreate-looping like electrumx).
Normalize both sides the way shell_join does before comparing.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
- app-manifest-spec.md: full rewrite from the real schema in
core/container/src/manifest.rs — it was 5 months stale and missing the
entire modern feature set (build, network/network_aliases, derived_env,
secret_env, generated_secrets/certs, data_uid, files, interfaces, hooks,
extensions flatten) and documented wrong network_policy values.
- app-developer-guide.md: add generated_secrets/generated_certs/
network_aliases/hooks to the field table.
- APP-PACKAGING-MIGRATION-PLAN.md: phase status stamped (1-3 done, 5 mostly,
4+6 open); deleted meshtastic app removed from regression-proof lists.
- registry-manifest-design.md: status design → implemented (phases 1-3),
stale manifest_dir:Option line fixed.
- marketplace-protocol.md: reframed proposal → as-built (marketplace.rs +
RPCs + UI shipped; create-invoice noted; schema disambiguated).
- manifest-hooks-design.md: phase 3 (indeedhub) done, phase 4 resolved via
orchestrator+generated_secrets instead of hooks.
- README/architecture: restore the NIP-07 signer-bridge claim — it is real
(neode-ui/public/nostr-provider.js), the earlier audit only grepped Rust.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
developer-guide: RPC dispatch moved from mod.rs handle_rpc_call to
dispatcher.rs dispatch() long ago — the registration instructions now point
at the real file; module tree updated (package/, federation/, mesh/ are
directories; mesh is tri-protocol); roadmap pointer loop/plan.md → docs/ROADMAP.md.
CONTRIBUTING: docs/development-setup.md never existed → developer-guide.md;
GitHub-specific fork URL generalized to the Gitea instance.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Fixes: phantom parmanode crate removed and crate table now matches the real
workspace (5 members + 4 standalone), Debian 12 box → 13, LOC/file counts
were understated ~2x (49k→117k Rust, 47k→69k TS/Vue), 30→51 apps, dispatcher
is dispatcher.rs with ~380 methods (not mod.rs '100+'), NIP-07 claim replaced
with the NIPs actually implemented (33/44/04), Meshcore-only mesh → tri-protocol
with Reticulum daemon, dead MASTER_PLAN.md/BETA-PROGRESS.md links replaced.
Adds as-built sections for the app platform (manifest → Quadlet → signed
catalog overlay → marketplace) and drops the environment-specific node
IP inventory from a public doc.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
container_command_drifted compared the manifest's entrypoint and
custom_args against Config.Entrypoint and Config.Cmd separately, but
quadlet's Entrypoint= takes a single value, so a manifest entrypoint of
[sh, -lc] is written as Entrypoint=sh + Exec=-lc ... and podman reports
entry=[sh], cmd=[-lc, script]. Same argv, different split — every
quadlet-created app with a multi-element entrypoint read as drifted
forever. On .228 this recreated electrumx on every reconcile pass (114
times in 6h); bitcoin-knots and fedimint-gateway showed the same false
drift and were only spared by the restart-sensitive guard.
Compare the concatenated argv (entrypoint ++ args vs Entrypoint ++ Cmd)
instead — that is what actually runs. When the manifest declares no
custom_args, only the entrypoint prefix must match, since the trailing
Cmd may be the image's baked-in CMD.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
README fixes: version badge 1.3.1→1.8.0-alpha, real LOC (117k Rust/69k TS-Vue
vs the stale 49k/47k), 51 apps not 29, tri-protocol mesh
(Meshtastic/MeshCore/Reticulum) not Meshcore-only, Nostr NIP-33/44/04 (NIP-07
was never implemented), Ed25519 signed catalog + pinned release-root anchor,
OTA host de-hardcoded (tx1138 retired), dead docs/MASTER_PLAN.md link removed,
crate map corrected (openwrt added, phantom crates dropped), new Philosophy and
Roadmap sections reflecting the manifest-driven/signed-registry north star.
docs/ROADMAP.md is the public-facing shipped/in-progress/planned summary,
synthesized from the master plan, unified tracker, and 1.8.0 hardening plan —
including the completed signing ceremony and the 1.8.0-alpha version decision.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Session logs, handovers, point-in-time snapshots, the v0.1.0 security audit,
stale generated HTML guides, and shipped design docs (three-mode UI, install
screens) move out of the living docs/ tree. docs/archive/README.md explains
what each was and why it's archived; all in-repo links updated.
(docs/UNIFIED-TASK-TRACKER.md links intentionally untouched — it has another
agent's uncommitted work in flight.)
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The packaged reticulum-daemon is a PyInstaller one-file binary: the
bootloader extracts a ~48M _MEI* dir into TMPDIR and only deletes it on
clean exit. ReticulumLink's shutdown SIGKILLed the process group
immediately (kill_on_drop + instant SIGTERM+SIGKILL in Drop), so the
cleanup never ran and every reconnect/restart stranded another 48M in
/tmp — filling .116's 12G tmpfs in ~4h.
Two-pronged fix:
- terminate_group(): group-wide SIGTERM now, SIGKILL from a detached
backstop thread 5s later, used on every shutdown path (Drop and all
spawn-failure paths, which previously used bare start_kill and could
orphan the forked Python child too).
- Per-interface private TMPDIR under the runtime dir, wiped on every
(re)spawn — even a hard-killed or power-lost daemon can never
accumulate stale extraction dirs, and they land on the data disk
instead of the small tmpfs /tmp.
Verified: unit tests + the ignored mesh_service_connects_over_
reticulum_tcp_client integration test (spawns two real daemons over
loopback TCP through the changed spawn/terminate path), clean teardown
with no stray processes.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
fedimint sat 'failed' for 7+ hours (exit 255 at 21:21, pre-gate) — the
reconciler repairs missing/drifted containers but never reset-fails +
starts a failed quadlet .service, and the health monitor can't see an
app whose container is gone. Concrete lever added to the Tier 2
container-flapping item.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
capabilities:[] was latent — the long-lived legacy container predated
strict manifest enforcement, so nothing noticed that a recreate against
this manifest produces a root process without DAC_OVERRIDE that
EACCESes on upload/encoded-video and crash-loops (49 systemd restarts
on .228 when the 2026-07-05 secret-env migration finally recreated
it). Any reinstall or reboot-repair would have tripped the same wire.
Cap set mirrors immich-postgres minus SETUID/SETGID.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Secret env used to merge into manifest.app.environment, landing in
'podman inspect' Config.Env on the API backend and — worse — as
plaintext Environment= lines in Quadlet unit files on disk. Now:
- expand_and_partition_env (container crate, pure + tested) expands
${KEY} placeholders and splits env into plain entries and
secret-bearing pairs. Plain entries that interpolate a secret
(btcpay's Password=${BTCPAY_DB_PASS} connection strings) are
tainted and travel as secrets too. Secret values themselves are
never expanded (a generated value containing '${' passes verbatim).
- values register as podman secrets: stdin (never argv/tempfile),
--replace, content-hash label to skip no-op rewrites; a per-app hash
cache in the orchestrator makes steady-state reconciles free of
podman secret calls. Registration goes through the runtime trait
(default no-op keeps mocks/docker inert).
- containers reference secrets by name: secret_env map in the libpod
create spec, Secret=<name>,type=env,target=<KEY> in Quadlet units.
Verified empirically on fleet podman 5.4.2: value absent from
inspect Config.Env, runtime injection works rootless.
- rotation detection: io.archipelago.secret-env-hash container label
(API) / the changed unit bytes (Quadlet). Pre-upgrade containers
lack the label, so every secret-bearing app recreates ONCE on the
first reconcile after deploy — deliberate, it scrubs the plaintext
secrets out of existing container configs. Data dirs untouched.
- docker dev fallback keeps plain -e injection (no secret store);
podman secrets persist across uninstall, matching the
preserve-credentials invariant (reinstall re-registers by hash).
In-container /proc/<pid>/environ is unchanged — env remains the
app-compat contract; the closed leaks are inspect output and unit
files on disk.
Tests: archipelago-container 61/61 (3 new: taint partition, verbatim
secrets, hash order-independence), archipelago container:: 160/160
(fedimint install test now asserts the secret arrives as a ref, not
env; quadlet render test asserts Secret=/Label= lines). NEEDS the
on-node gate re-run before the item counts as verified.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
New container::image_verify gates PodmanClient::pull_image and the
dev-only DockerRuntime::pull_image. Signature claims classify three
ways: absent/empty (pull unverified, logged), the literal
'cosign://...' placeholder every fleet manifest carries today (same —
enforcement stays dormant until the signing ceremony ships real
values), or a declared signature, which must verify via
'cosign verify --key /etc/archipelago/cosign.pub
--insecure-ignore-tlog=true' (plus --allow-insecure-registry
--allow-http-registry for the HTTP mirror; flags checked against
cosign's own docs) before anything is fetched. Missing key, missing
cosign binary, timeout, or verification failure all hard-fail the
pull — a declared signature cannot be skipped on either runtime. Key
path overridable via ARCHIPELAGO_COSIGN_PUBKEY for tests/staging.
Deletes security::ImageVerifier: zero callers, blocking
std::process::Command on would-be async paths, and a fantasy
'cosign verify --signature' invocation (that flag belongs to
verify-blob).
Activation ships with the Workstream B ceremony, in order: pin
cosign.pub on nodes + install cosign, then publish real
image_signature values in the catalog.
Tests: archipelago-container 58/58 (5 new), archipelago container::
159/159, security check clean.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
- mesh: verify_signature accepts a v2 preimage (t,v,ts,seq) alongside
legacy v1 (t,v,ts); signed_with_seq() is the v2 sender path, not yet
wired — senders stay v1 until the fleet verifies v2 (receivers
hard-drop bad sigs, so flipping send-side first would break
mixed-fleet alerts). Tests: v2 verify, v2 seq-tamper rejection,
v1 sign-then-set-seq compat.
- mesh listener: malformed radio-supplied DID shorter than the
'did🔑' prefix can no longer panic advert_name (slice -> .get()).
- auth: the pre-setup password123 dev login and the constant itself are
now #[cfg(debug_assertions)] — no release binary carries the bypass,
whatever its runtime config says.
- orchestrator: canned host-facts under #[cfg(test)] — awaiting real
subprocesses under tokio's paused test clock deadlocks against
auto-advanced timers (the old blocking detection only worked by never
yielding).
- drop two now-unused std::process::Command imports left by 4c75bb3d.
Tests: mesh 110/110 (incl. 2 new), api 68/68, container 159/159,
archipelago-container check clean.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Was implicit across the Phase-3 Quadlet flip and Workstream F; now one
consolidated pre-tag item with the lever list, an observability proposal
(per-app restart counter + flap log line), and an already-landed list so
nothing gets re-done.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
extract_client_ip took X-Real-IP/X-Forwarded-For from any request, so
a client talking to the backend directly (the FIPS peer listener, or
any non-proxy path) could rotate a fake IP per request and never trip
the login rate limiter. The accept loop now records the TCP peer
address in request extensions, and forwarded headers are honored only
when the connection itself is from loopback — where nginx overwrites
X-Real-IP with the real client address. Direct connections bucket
under their socket IP.
§C of the 1.8.0 hardening plan; 3 new unit tests cover the
loopback/direct/no-header matrix.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
verify_pending_update previously cleared the rollback marker on any
2xx/3xx from GET / — a release with a dead RPC API or broken podman
access passed and never rolled back. Verification now requires, in the
same attempt: the frontend via nginx, backend RPC liveness (an
unauthenticated POST /rpc/v1 — 401 proves the stack is up, 5xx/404/
refused fails it), and rootless podman reachability. A pre-loop check
also asserts the running binary's version matches what the marker says
was applied, catching a silent or half swap deterministically.
Per-app container assertions are deliberately excluded: the
pre-Quadlet service restart legitimately takes containers down and the
boot reconciler can need minutes for heavy apps — that would
false-rollback healthy updates. Revisit after the Phase-3 flip.
§B of the 1.8.0 hardening plan; update suite 38/38 green.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Confirmed on the just-packaged binary: `archy-rnodeconf --info` printed
everything correctly, then crashed with NameError: name 'exit' is not
defined and returned exit code 1. rnodeconf.py's own graceful_exit() calls
the bare exit() builtin, which is only ever defined by site.py for
interactive Python — a frozen PyInstaller app skips that init, so any
bundled script relying on it hits this the moment it tries to quit cleanly,
after the real work already succeeded. Classic, well-documented PyInstaller
gotcha; the standard fix is a runtime hook pre-defining exit/quit as
sys.exit before the bundled script's own code runs.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
RNS's own rnodeconf utility (frequency/bandwidth/spreading-factor/coding-rate
read+write, firmware signature verification, board bootstrap) has been the
one tool that reliably diagnoses real RNode hardware — it's what finally
proved two live nodes were silently running at different spreading factors
(SF5 vs SF10, invisible from any of our own probe/logging code, and the
actual reason two correctly-flashed radios could detect each other's RF but
never decode a packet). Every node should have it, not just whichever one an
agent happened to build a throwaway venv on to debug a specific incident.
- reticulum-daemon/build.sh: also PyInstaller-package archy-rnodeconf
alongside the existing archy-reticulum-daemon, same --collect-submodules/
-d noarchive flags (same RNS.Interfaces __all__-glob requirement applies).
- scripts/deploy-to-target.sh: actually wire both packaged binaries into the
live deploy path (neither was wired in before — a pre-existing gap noted
in docs/RETICULUM-TRANSPORT-PROGRESS.md; this is why reticulum-daemon
previously only worked via manual dev-venv setup on rsync-deployed nodes,
not the ISO-imaged ones). Non-fatal on build/deploy failure — archipelago
already falls back to its dev-venv path if the packaged binary is absent.
Also installs the missing `python3.<minor>-venv` package when needed
(same "ensurepip not available" gap hit manually twice this session).
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Catalog- and manifest-supplied image refs reached pull_image without
ever passing the RPC boundary's validator — a malicious catalog entry
or manifest could pull from an arbitrary registry. The allowlist now
lives in container::image_policy (the RPC check delegates to it) and
both orchestrator pull sites (install_fresh and
ensure_resolved_source_available) refuse refs that fail it.
The shared policy accepts trusted-registry refs and registry-less
Docker Hub shorthand (grafana/grafana etc., used by 8 shipped
manifests — a registry-less ref cannot name an attacker host), and
rejects explicit non-allowlisted hosts, shell metacharacters, and
malformed refs. §A of the 1.8.0 hardening plan.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Server.vue rendered the backend-generated WireGuard peer QR with raw
v-html while the analogous TOTP QR was DOMPurify-sanitized — both now
use the same svg-profile sanitizer. Mesh.vue's 5s poll interval gets
the same start-guard as the arch poll (no leak on double-mount) and is
nulled on unmount. curatedApps.ts catalog fetches no longer fail
silently: each failed source logs a console.warn, including the final
all-sources-failed fallback to the hardcoded list.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Every production process spawn reachable from a tokio worker now uses
tokio::process: the install path's podman-port probe, the dependencies
disk check, factory-reset restart, config host-IP detection, the
orchestrator's host-facts helpers (resolve_dynamic_env and its call
sites made async to carry it through), and AutoRuntime's podman/docker
probes.
The FIPS transport probe is the special case: is_available() is a sync
trait method called from async route(), so instead of blocking ~50ms
on systemctl per stale-cache hit it now serves the cached value and
refreshes on a background thread (stale-while-revalidate) — bounded
staleness, zero stalled workers.
§C of the 1.8.0 hardening plan; container/transport/config/package
suites green.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Confirmed on real hardware: a genuine Heltec V4 RNode (firmware 1.86,
verified via rnodeconf --info against the same port) prints extra
boot/status chatter over the same serial line before answering KISS
commands. Replaying the exact probe bytes and timing budget our Rust
probe_rnode() uses showed DETECT_RESP arriving ~1.05s after the write —
past the old 800ms deadline, so a real, correctly-flashed, correctly-
responding RNode was being timed out and misclassified as "not an RNode".
2.5s leaves comfortable margin. Non-RNode devices (Meshcore/Meshtastic)
already budget ~5s each, so this doesn't meaningfully slow down detection
when the port turns out to be something else.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Two MeshPeer rows were being created for one physical Reticulum node: a
radio twin (keyed by the RNS dest_hash, arch_pubkey_hex always None) and a
pseudo-federation twin (keyed by the archy ed25519 pubkey, created via the
generic identity-broadcast path meant for Meshcore/Meshtastic). The generic
path relies on bind_federation_twins matching both twins' advert_name, but
the Reticulum radio twin's display_name is deliberately never the identity
text — so the two rows could never merge, and the generic send path (keyed
off whichever twin the caller resolves) ended up looking up the archy
pubkey's prefix in ReticulumLink's `prefix_to_hash` map, which is only ever
populated with RNS dest_hash prefixes. Every send failed with "Unknown
Reticulum prefix ... peer hasn't announced yet", confirmed live between two
real nodes (archy-x250-exp / archy-x250-pa) that could see each other's
adverts but never exchange a message.
Unlike Meshcore/Meshtastic, Reticulum's ARCHY identity blob arrives in the
same announce event as the destination hash, so there's no ambiguity about
which peer it belongs to — bind it directly onto the RNS-hash-keyed radio
peer instead of relying on name-matching. Threaded through a new
`ParsedContact::arch_pubkey_hex` (None for Meshcore/Meshtastic, unchanged
behavior there) so `refresh_contacts` can set it on the correct peer row
without touching `bind_federation_twins`/`group_peer_twins`, which already
know how to collapse twins once they share an arch_pubkey_hex.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Confirmed on real hardware (Heltec V4, RNode firmware): the board answers
the exact KISS detect probe correctly and instantly on a fresh port open,
but auto_detect_and_open (and open_preferred_path's unpinned branch) tried
Meshcore (~5s timeout) then Meshtastic (~5s timeout) first, leaving the
RNode firmware unresponsive by the time Reticulum's turn came ~10.6s later.
ReticulumLink::open() already gates its expensive daemon-spawn behind a
cheap ~1s probe_rnode check, so trying it first only costs ~1s extra when
the device turns out to be Meshcore/Meshtastic instead — the previous
comment's "most expensive, so goes last" reasoning only applies to a
successful match, not a failed one.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
