Per review: docs represent current state. Every "used to" / "no longer"
clause this branch introduced is gone — including the History section in
network.md, which was a whole subsection about a sync mechanism that
doesn't exist.
Where the removed clause was carrying a real constraint, the constraint
stays and is stated in the present tense instead of as a delta: nothing
narrows what the gateway's nginx can reach except the directory
permissions in front of a socket, and nothing bounds `ReloadGatewayNginx`
except the hard-coded unit name. Those read as rules now rather than as
the story of how they came to be rules.
The gateway's nginx + dnsmasq no longer run in their own nspawn container.
`nix/host-modules/hive-gateway/default.nix` loses the
`containers.hive-gateway` wrapper and everything that existed only to punch
holes in it: `privateNetwork = false`, `CAP_NET_ADMIN`, five bind mounts,
its own `stateVersion`, `networking.firewall.enable = false`,
`networking.resolvconf.enable = false`, and the `hive-gateway-resolv`
path+service pair. 465 -> 303 lines.
The container never bought isolation here. It shared the host netns by
necessity — nginx binds the host's :80/:443, dnsmasq answers on the bridge —
so each of those settings was undoing a boundary the gateway could not
afford in the first place.
Four things made it more than a deletion, none of them visible in the nix
diff:
- The self-signed cert service also imports the hive CA leaf, so removing it
with the container would have left nginx naming a missing cert file, which
it refuses to load at all.
- The nginx reload is a hive-priv verb. It still needs root, but no longer
for the reason its doc gave, and `--machine=` was both transport and
scope — so the unit name is now hard-coded in the helper as the
containment.
- The lifecycle verb named a container that stops existing.
- `journalctl -M hive-gateway` had no machine to enter.
Per the operator's ruling, the operator verb keeps working and agents lose
it. `InfraContainer` answered three questions that used to share an answer;
it now splits into `name()` (identity), `target()` (Container vs HostUnit),
`service_unit()` (the systemd unit), and `agent_restartable()`, which the
MCP restart path checks before the capability so the refusal cannot read as
"ask for infra_admin". `SIBLING_CONTAINERS` drops the gateway — it gates the
requests that name a container as a string — while `FromStr` still accepts
it, because that answers what a name is, not who may act on it. The
dashboard's gateway journal reads host journald filtered to `nginx.service`.
Prose was corrected where it only named a location, and re-argued where the
container was doing security work: a `0666` per-agent socket was safe
because only the gateway container had the directory bind-mounted. There is
no mount now, so the directory permissions are the whole of the access
control — the constraint holds, its mechanism doesn't.
Gate: nix fmt / clippy --all-targets -D warnings / cargo test all clean (710
tests); hivectl-cli.md regenerated from the clap tree. The nix eval was run
in both TLS shapes at this commit: every delta in the rendered
virtualHosts is one of the three intended path moves, dnsmasq settings are
byte-identical, and the absence probe flips true -> false with bindMounts
emptied.
Extends the SweepHealth/warnings registry (already landed for
knowledge_pull) to two more background sweeps:
- gateway_nginx::reload_gateway_nginx: raises a warn-level banner
immediately on the first failed reload (routing changes silently
not taking effect is user-visible right now, so no debounce).
- forge::repos::ensure_config_repo: raises a crit-level banner
listing every agent whose config-repo branch protection is
currently unapplied (security-relevant — bypasses the deploy
pipeline), clearing agents out of the message as they recover.
Journal warn!/error! logging is left in place; the registry adds a
dashboard-visible signal on top. forge::ensure_all() and
matrix::ensure_all() sweeps are deliberately left for a fast-follow.
Agents run in private netns (always-on isolation). The TCP fallback
to 127.0.0.1:<port> was unreachable from the gateway's host netns
regardless of whether the per-agent socket marker existed.
Remove the conditional entirely: always use the unix socket path.
If the socket is not yet bound, nginx returns 502 which is already
handled by the error_page 502 503 504 = /__hive_agent_unreachable
directive in every location block.
Also removes the unused lifecycle::agent_web_port call and the
now-misleading '.bound state' mention from the render doc comment.
systemctl --machine=hive-gateway requires root (machine-bus transport
enters the container namespace). hive-c0re is unprivileged, so every
call to nginx_active_state() and gateway_systemctl() silently failed
with exit 1, causing a continuous 30s retry loop without ever
syncing nginx.
Fix:
- Move state-aware nginx logic into hive-priv ReloadGatewayNginx:
check ActiveState, then reload/reset-start/start accordingly.
hive-priv already runs as root and has machine-bus rights.
- Remove nginx_active_state() and gateway_systemctl() from
gateway_nginx.rs (they were always running unprivileged, always
failing silently).
- Make write(), reload_if_pending(), reload_gateway_nginx() async so
they can call the async priv_client without a blocking bridge.
- Update callers in agent_sockets::spawn_poll and meta::sync_agents
to await the now-async functions.
The priv_client::reload_gateway_nginx() call and PrivRequest::ReloadGatewayNginx
wire type already existed — the gateway_nginx module was just not using them.
RELOAD_PENDING stays true on any failed reload, and reload_if_pending
fires on every 10-second spawn_poll tick. When the gateway is down or
nginx config is bad this hammers systemctl indefinitely.
Fix: track LAST_FAILED_RELOAD (unix timestamp). reload_if_pending backs
off to one attempt per RELOAD_RETRY_SECS (30s) after a failure. Fresh
write() calls reset the backoff so new topology changes still attempt
reload immediately.
Fixes#1113.
All crates now pass `cargo clippy --workspace -- -D warnings` cleanly.
Fixes span six crates (hive-sh4re, hive-ag3nt, hive-c0re, hive-forge,
hive-priv, hive-matrix-mcp was already clean):
- doc_markdown: wrap snake_case, type names, constants in backticks
- collapsible_if / collapsible_match: fold nested ifs into let-chains
- duration_suboptimal_units: Duration::from_secs(N) → from_mins/from_hours
- implicit_hasher: allow on HashMap-param fns where generalization is risky
- items_after_statements: hoist use to function tops
- map(f).unwrap_or(x) → map_or(x, f); map(f).unwrap_or_else(g) → map_or_else
- is_ok_and / is_none_or in place of map().unwrap_or(bool)
- needless_continue: {} instead of continue in loop match arms
- match_same_arms: Ok(None) | Err(_) merged
- format_push_str: write!() instead of push_str(&format!())
- while let replaces loop { let Some(..) = x else { break } }
- struct_excessive_bools / dead_code: allow on purpose-built structs
- too_many_lines / too_many_arguments: allow where refactor not worth it
- unused_async: remove async from poll_once in bash_runner
- needless_borrow: fix &repo deref in hive-forge comments verb
- cast_possible_truncation: allow u64→usize in fetch_tail
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
nginx -s reload signals a running master process. When nginx enters
failed state (start-limit-hit from repeated nginx -t failures on a
bad agents.conf), there is no master and the reload is a silent no-op.
c0re kept re-firing the same no-op reload forever via RELOAD_PENDING.
Fix: probe the nginx unit's ActiveState before sending the reload:
- active → nginx -s reload (existing zero-downtime path)
- failed → systemctl reset-failed nginx + systemctl start nginx
- other → systemctl start nginx
This makes c0re self-healing: once a corrected agents.conf is published,
the next reload_gateway_nginx call clears the start-limit and restarts
nginx automatically without operator intervention.
New helpers: nginx_active_state() (systemctl show --property=ActiveState
--value) and gateway_systemctl() (host-side systemctl --machine=hive-gateway).
nginx forbids a URI part on proxy_pass inside named locations, so the
split-mode `@<name>_dynamic` fallback (introduced when the gateway
started serving static dist directly) failed config-test with:
"proxy_pass" cannot have URI part in location given by regular
expression, or inside named location
Render two upstream forms: prefix locations keep the trailing slash so
nginx auto-strips the location prefix; the named-location path strips
`/agent/<name>/` via `rewrite ... break` and uses a bare upstream.
gateway_nginx.rs reads HIVE_AGENT_FRONTEND_DIR (injected by hive-c0re.nix
as ${cfg.frontend}/agent). When set, agents.conf emits per-agent split
blocks instead of the old single proxy_pass:
# Compiled assets — immutable nix store path, cache 1y
location ^~ /agent/<name>/static/ {
alias <frontend>/static/;
expires 1y; add_header Cache-Control "public, immutable, ...";
}
# Static dist + proxy fallback
location /agent/<name>/ {
alias <frontend>/;
try_files $uri $uri.html $uri/index.html @<name>_dynamic;
}
location @<name>_dynamic {
proxy_pass <upstream>; # api, events, icon, login, …
…proxy headers unchanged…
}
try_files path resolution (nginx applies alias mapping first):
$uri — exact file (/static/app.js → static/app.js)
$uri.html — bare-path fallback (/stats → stats.html)
$uri/index.html — directory index (/ → index.html)
@<name>_dynamic — proxy catchall for anything not in the dist
Adding pages to the frontend dist works automatically — no generator
change needed. Per-agent extraFiles (in mergedDist, not in the base
nix-store path) continue to proxy to the agent daemon.
frontend is a nix store path injected at build time — only [a-z0-9/._-],
no shell metacharacters — safe to interpolate without sanitization;
comment added documenting this assumption.
Without HIVE_AGENT_FRONTEND_DIR the existing single-proxy block is
emitted unchanged — backward-compatible for deployments without the env.
render() takes frontend_dir as a parameter so tests exercise both code
paths safely in parallel. 13 tests: 7 legacy, 6 split-mode. No clippy
warnings in changed files.
nix/modules/hive-c0re.nix: inject HIVE_AGENT_FRONTEND_DIR = "${cfg.frontend}/agent".
The #872 path-unit approach was silently broken: IN_MOVED_TO from an
atomic rename on the host does not propagate across the nspawn
mount-namespace boundary into the container, so the watcher inside
the gateway container never fired.
Fix: after each agents.conf write, c0re calls
`systemd-run --machine=hive-gateway -- nginx -s reload` from the host.
The reload is best-effort (logged on failure, not fatal).
Remove the now-unused `hive-gateway-agents-conf.path` +
`hive-gateway-nginx-reload.service` from the gateway container config.
Update docs/gateway.md + comments to reflect the host-side approach.
- gateway_nginx.rs: HOST_CONF_PATH → /var/lib/hyperhive/gateway/agents.conf
- hive-gateway.nix: hostPath = /var/lib/hyperhive/gateway (not whole parent
dir — avoids exposing forge tokens or other credentials to the gateway
container)
- tmpfiles: add /var/lib/hyperhive/gateway/ dir rule + seed agents.conf there
- scrub "(#869)" from hive-gateway-nginx-reload service description
Replace eval-time per-agent nginx location baking with a runtime
include file. c0re writes /var/lib/hyperhive/agents.conf (nginx
location blocks, UDS or TCP per agent) on every topology change and
on the 10s marker poll. The gateway container bind-mounts
/var/lib/hyperhive/ at /run/hive-state/ and nginx includes
/run/hive-state/agents.conf. A systemd path unit inside the container
watches the file for changes and fires `nginx -s reload` on each
atomic rename from c0re — no nixos-rebuild switch needed when agents
start, stop, or flip useUnixSocket.
- new hive-c0re/src/gateway_nginx.rs: write() + render()
- lib.rs + meta.rs + agent_sockets::spawn_poll: hook in write()
- hive-gateway.nix: drop agentPortsTable/agentSocketsTable/
agentUpstreamFor/lib.mapAttrs', add /run/hive-state bind-mount,
include directive, systemd path unit + reload service, tmpfiles
for /var/lib/hyperhive + agents.conf seed
- docs/gateway.md: update vhost table + Per-agent UDS section
