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>
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>
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>
Three stacked bugs made "switch version" silently fail / crash-loop, and
the data-access mismatch corrupted a node's index during recovery attempts.
Backend renderer:
- sync_quadlet_unit ignored the per-app pinned version and re-rendered the
quadlet with the manifest's :latest every reconcile tick, reverting any
switch. Factor the install-time catalog/pin resolution into a shared
resolve_catalog_image() and call it in BOTH install_fresh and
sync_quadlet_unit.
- The renderer folded manifest `entrypoint: ["sh","-lc"]` into Exec=, which
only worked when the image entrypoint was a passthrough shell wrapper. The
versioned images use ENTRYPOINT ["bitcoind"], so Exec=sh -lc ... became
`bitcoind sh -lc ...` and crash-looped. Emit a real Entrypoint= override;
exec_changed now also compares Entrypoint=.
Images:
- Build all bitcoin images (Core + Knots, every version) as container-root
(USER removed) like the legacy :latest image. Chain data is owned by the
data_uid (container uid 102); root reads it via CAP_DAC_OVERRIDE (granted in
the manifest). A non-root USER (the previous uid 1000) can't read existing
chain data → "Error initializing block database". Still fully rootless:
container-root maps to the unprivileged host service user.
Catalog:
- bitcoin-knots versions[]: 29.3.knots20260508/20260507/20260210 +
29.2.knots20251110, "latest" tracking newest.
- bitcoin-core versions[]: add 29.2 + a "latest" entry. All images rebuilt
root and published to the mirror.
Frontend:
- AppSidebar version dropdown: rename the latest option to "Always use the
latest version" (no v prefix), fix right padding, and guarantee the current
selection matches a real option (was rendering blank).
- New InstallVersionModal: full-screen version chooser shown from the App
Store / Discover install button for multi-version apps (Bitcoin Knots/Core),
app icon + "Install <name>", latest pre-selected.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Uninstalling immich/grafana could hang with a frozen full-red progress
bar, leave a ghost entry stuck in My Apps, and then refuse reinstall.
Single root cause: quadlet::disable_remove() — called first in the
uninstall task (via companion + orchestrator teardown) — ran
`systemctl --user stop`, daemon-reload, and `podman rm -f` with NO
timeout. On rootless podman a generated unit can wedge in "deactivating"
while podman hangs underneath, so `systemctl stop` blocks forever. The
spawned uninstall task then never returns Ok or Err, so:
- set_uninstall_stage() (after the stop) never fires → progress frozen;
- remove_package_state_entry() never runs → entry stranded in
`Removing` → ghost in My Apps;
- the install guard rejects reinstall with "already Removing".
The spawn wrapper already reverts state on Err and removes the entry on
Ok — the only failure mode was a hang that returns neither. Bound the
teardown so it always terminates:
- systemctl stop → QUADLET_STOP_TIMEOUT, escalate to kill+reset-failed
on timeout (reuses the existing helpers);
- daemon_reload_user() → bounded systemctl_user_status (30s);
- defensive `podman rm -f` → wrapped in tokio timeout.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Add `container.network_aliases: Vec<String>` (serde default, DNS-label
validated) so a stack member can answer to short hostnames its peers bake
in, beyond its own container name. Rendered in both runtime paths:
- podman_client: merged (deduped) into the custom-network aliases array.
- quadlet from_manifest: appended after the container name; emitted only
for Bridge networks (slirp/pasta reject aliases).
Needed for the indeedhub migration: its frontend nginx proxies to
`api:4000` / `minio:9000` / `relay:8080`, so those members declare
`network_aliases: [api|minio|relay]` to keep the short names resolvable on
the dedicated indeedhub-net (vs. colliding generic aliases on archy-net).
Also fixes 4 pre-existing from_manifest test failures (unrelated to this
change, surfaced now that the quadlet suite runs green): test manifests
used the long-invalid `network_policy: archy-net` (allowlist is
isolated/bridge/host → moved to network_policy: isolated + container.network)
and bind sources outside /var/lib/archipelago.
Tests: container crate 53 pass; archipelago quadlet+alias 47 pass.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
fmcd crash-looped "Operation not permitted (os error 1)" on .116 (kernel
6.12.74): the default rootless seccomp profile blocks a syscall its Mainline-DHT
/ iroh transport needs, so the REST API never came up (:8178 → HTTP 000) and
federations couldn't be joined. Verified: with seccomp=unconfined fmcd boots and
answers /v2/* (HTTP 401 instead of dead). fmcd works on other nodes, so this is
kernel/seccomp-specific — but the relaxation is safe for an outbound-networking
daemon and harmless where not needed.
- new `security.seccomp_unconfined` manifest flag (SecurityPolicy);
- libpod backend sets `seccomp_profile_path: "unconfined"` (== --security-opt
seccomp=unconfined); quadlet backend emits `SeccompProfile=unconfined`;
- enabled in apps/fedimint-clientd/manifest.yml.
NOTE: manifests live on-disk at /opt/archipelago/apps/<id>/manifest.yml, so the
node needs the updated manifest deployed + the fmcd container recreated to apply.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Bug surfaced live on .228 2026-05-02 — every backend Quadlet unit
(lnd, electrumx, fedimint, btcpay-server, mempool-api, bitcoin-knots)
hit systemd's default 90s start timeout because Notify=healthy makes
systemctl wait for the first green health probe, but
HealthInterval=30s × HealthRetries=3 = 90s minimum even on a healthy
service. Race: timeout fires the moment the third probe MIGHT succeed.
Result was three different post-states (inactive+running, failed+missing,
inactive+stopped) depending on whether systemd's ExecStopPost ran
podman rm before the orchestrator's adoption logic re-grabbed the
container.
Fix: when health is set, render TimeoutStartSec=600 (10 minutes) into
[Service]. Long enough for slow-starting backends (electrumx index
replay, lnd wallet unlock) without being so long that a truly stuck
unit hangs forever. Companions stay unchanged (no health → no override,
default 90s applies).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Two bugs surfaced by the first real-node validation of Phase 3.2-3.4
on .228 (2026-05-02), both caught before flipping the default.
Bug 1 — translate_health_check double-prefixed http://. Manifests in
the wild carry the scheme inside the endpoint string ("http://localhost:8175"),
and we were prepending another http:// unconditionally. Result on .228:
every backend HealthCmd read `curl -fsS -m 5 http://http://localhost...`,
every probe failed, fedimint hit a 14-restart loop. Now we accept either
form and skip appending hc.path when the endpoint already carries one.
Regression test asserts no double-prefix and that an in-endpoint path
is honoured.
Bug 2 — Phase 3.3 migration ran for UI companions (bitcoin-ui /
electrs-ui / lnd-ui) that have shipped via Quadlet since v1.7.41.
Migration tore down the running companion + raced companion.rs render,
producing "Phase 3.3: re-install archy-bitcoin-ui via Quadlet" reconcile
errors and leaving archy-bitcoin-ui down. Companions now short-circuit
out of migrate_to_quadlet_if_needed before any IO. Also: when try_exists
returns Err for an unrelated reason (permissions, EIO), we now skip
migration instead of treating "I can't tell" as "go ahead and migrate" —
migrating on top of a possibly-existing unit is destructive.
What this does not fix yet:
* the orchestrator's reconciler iterating every manifest in
/opt/archipelago/apps/, not just installed apps. Pre-existing
behavior (also affects the legacy path) — separate scope.
* fedimint /data UID mismatch surfaced when Quadlet started fedimint
fresh. Likely orthogonal — defer.
* no rollback when install_via_quadlet fails after a remove_container.
Tracked as Phase 3.3.1 — defer.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
QuadletUnit gains an optional HealthSpec; from_manifest translates the
manifest's health_check (tcp/http/cmd) into a HealthCmd= directive and
emits Notify=healthy alongside it. systemctl start <unit>.service then
blocks until the container's first green probe — eliminating the
"container up but RPC not ready" race the orchestrator currently papers
over with post-start polling.
Translation policy:
* tcp, endpoint "host:port" -> nc -z host port
* http, endpoint "host:port", path -> curl -fsS -m 5 http://endpoint<path>
* cmd, endpoint "<shell command>" -> verbatim
* unknown type / malformed endpoint -> None (skip Notify=healthy rather
than emit a HealthCmd that hangs the unit start forever)
Companion units leave health: None and remain byte-identical to before
this PR — the renderer only emits the Health* / Notify= block when set.
+4 quadlet unit tests (19 total). Dropped a never-used test setter that
was generating a dead_code warning.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
prod_orchestrator::install_fresh now branches on the new
Config::use_quadlet_backends flag (default false):
* off (today's production behavior) — unchanged: runtime.create_container
+ start_container, container parented under archipelago.service's
cgroup, FM3 cascade SIGKILL on every archipelago restart.
* on — install_via_quadlet renders the manifest as a Quadlet unit via
QuadletUnit::from_manifest, writes it atomically into
~/.config/containers/systemd/, calls daemon-reload, and starts the
generated <name>.service. Container ends up under user.slice — no
more cgroup parented under archipelago, so archipelago restarts
don't touch the container's lifetime.
Default off so this commit is structurally safe to ship: nothing
changes at runtime until an operator opts in. Flip the default once
tests/lifecycle/run-20x.sh has gone green against the new path on
.228 + .198 (the v1.7.52 release gate).
Plumbing:
* config.rs — `use_quadlet_backends: bool` w/ Default false
* prod_orchestrator.rs — flag stored on the struct, threaded through
new(), with set_use_quadlet_backends(bool) test setter
* prod_orchestrator.rs — install_via_quadlet helper
* dropped the Phase-3.1 #[allow(dead_code)] markers on from_manifest /
parse_memory_mib / RestartPolicy::OnFailure now that the call path
exists; if a future revert removes the wiring, the warnings come back.
Tests: 624 passing, cargo check clean (0 warnings). Existing companion
behavior unaffected — render_skips_backend_directives_when_default
still passes byte-equal to before quadlet.rs grew the new fields.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The QuadletUnit struct now covers everything a backend manifest needs
(ports, environment, devices, add_hosts, entrypoint+command, read-only
root, no_new_privileges, cpu_quota, restart policy choice). Adds
QuadletUnit::from_manifest(&AppManifest, name) that translates a parsed
manifest into a unit, plus parse_memory_mib for "1g"/"512m"/raw-MiB
forms. The renderer skips empty/false directives so existing companion
units render byte-identically — no behavior change for shipping
companions; the backend renderer is dead code until Phase 3.2 wires it
into the orchestrator.
Eight new unit tests cover:
* parse_memory_mib forms (1024, 512m, 2g, garbage)
* shell_join quoting (whitespace, embedded quotes)
* RestartPolicy → systemd string mapping
* render emits backend directives when set
* render skips them when defaulted (companion regression gate)
* from_manifest happy path on a bitcoin-knots-shaped manifest
* from_manifest read-only volume detection
* from_manifest tmpfs filtering
* end-to-end manifest → render bytes assertion
Tests: 615 → 624 (+9 net; one pre-existing parse_memory_mib path was
implicitly covered before but is now explicit). Cargo warnings: 0.
`from_manifest`, `parse_memory_mib`, and `RestartPolicy::OnFailure` are
marked allow(dead_code) with explicit references to Phase 3.2 — if
3.2 doesn't wire them, the dead-code warning resurfaces.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Companion UI containers (archy-bitcoin-ui, archy-lnd-ui,
archy-electrs-ui) used to be launched as fire-and-forget tokio::spawn
blocks from install.rs. If archipelago crashed mid-spawn or the
container's cgroup was reaped, companions vanished from podman ps -a
and only a manual rm/run could bring them back (the .228 incident).
Now each companion is rendered as a Quadlet .container unit under
~/.config/containers/systemd/, daemon-reloaded, and started via
systemctl --user. systemd owns supervision from that point on:
- archipelago can crash, restart, or be uninstalled without touching
any companion.
- Quadlet's Restart=always + RestartSec=10 handles container exits.
- A 30s reconcile tick in boot_reconciler enumerates expected
companion units and re-installs any whose unit file or service
vanished — defense-in-depth against external tampering.
New module layout:
- container/quadlet.rs: pure unit renderer + atomic write_if_changed
+ systemctl helpers (daemon_reload_user / enable_now / disable_remove
/ is_active). 6 unit tests, no I/O in the renderer.
- container/companion.rs: per-app companion specs, install/remove/
reconcile, image presence (build local first, fall back to insecure
registry only via image_uses_insecure_registry whitelist). 2 tests.
install.rs handle_package_install now ends with a single call to
companion::install_for(package_id), replacing 287 lines of spawn-and-
hope shellouts plus a ~120-line nginx auth-injector helper that worked
around per-node RPC password baking. The helper is gone too — the
pre-start hook renders the per-node nginx.conf to /var/lib/archipelago/
bitcoin-ui/nginx.conf and the Quadlet unit bind-mounts it read-only.
runtime.rs handle_package_uninstall now disables companions before
the container rm loop. Otherwise systemd's Restart=always would
respawn each companion within ~10s of removal.
Tests: 53 container tests pass, including 6 quadlet renderer tests
(host network, bridge network, capability set, atomic write idempotence)
and 2 companion specs (per-app companion lookup, build_unit shape).
boot_reconciler tests gain a #[cfg(test)] without_companion_stage()
flag so the paused-clock fixtures don't race the real systemctl I/O.
A bats regression test (companion-survives-archipelago-restart.bats,
gated on ARCHY_ALLOW_DESTRUCTIVE=1) asserts the .228 failure mode
cannot recur: every installed companion has a unit file, services
stay active across systemctl --user restart archipelago, and a
deleted unit file is recreated within one reconcile tick.
Net delta: +941 / -363, but the +941 is mostly tests (~440 lines)
and the new declarative layer; the imperative tokio::spawn block and
its nginx-auth helper are gone, removing two failure classes
(orphan companions on archipelago crash, and post-start exec races
under tightly-confined cgroups) that previously needed manual SSH
recovery.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
