hyperhive/nix/modules/hive-network.nix
atlas 78482cb624 nix: remove the non-isolated (shared-netns) mode
Network isolation is validated working end to end (forge + matrix reach
the gateway under private netns), so the shared-host-netns mode is
removed — there is now one mode.

- hive-network.nix: the bridge/DNS block and the isolation overlay are
  now unconditional whenever hyperhive is enabled (gated on
  services.hyperhive.enable, not the per-feature toggles). The redundant
  isolateContainers-implies-enable guard is dropped.
- network.enable and isolateContainers are kept as DEPRECATED no-op
  options so existing configs that set them to true still evaluate
  cleanly; setting either to false now warns (gated on hyperhive being
  enabled, so a non-hyperhive host gets no spurious warning) and has no
  effect. Both are slated for removal in a future release.
- hive-c0re.nix: the HIVE_FORGE_URL / HIVE_MATRIX_URL conditionals drop
  the "&& isolateContainers" clause — agents always reach forge/matrix
  through the gateway vhost now.
- hive-matrix.nix: refresh one stale comment.

The Rust lifecycle is env-driven (HIVE_NETWORK_ISOLATION), so no Rust
change is needed: the env is now always set, so the existing
private-netns path always runs.

Supersedes the default-on flip (the option it would have defaulted is
gone).
2026-06-17 15:29:52 +02:00

245 lines
9.9 KiB
Nix

{
lib,
config,
...
}:
let
cfg = config.services.hyperhive.network;
in
{
# Hive-internal network — host-side bridge + per-agent DNS resolver.
# Containers stay on shared host netns at v1; this module stands the
# bridge + resolver up so the endpoint is in place before network
# isolation flips containers to private netns. Full design: docs/network.md.
options.services.hyperhive.network = {
enable = lib.mkOption {
type = lib.types.bool;
default = config.services.hyperhive.enable;
defaultText = lib.literalExpression "config.services.hyperhive.enable";
example = false;
description = ''
**DEPRECATED ignored.** The hive network (bridge + dnsmasq
resolver + private-netns isolation) is now always on whenever
hyperhive is enabled; setting this to `false` warns and has no
effect. Retained as a no-op so existing configs eval; will be
removed in a future release.
The network requires `services.hyperhive.domain` to be set the
dnsmasq resolver is authoritative for `<hive-domain>` and its
sub-domains. A bridge interface (`bridgeName`) appears on the host
with `bridgeIp` assigned, the hive-gateway container runs a dnsmasq
on that IP, and each agent container runs in a private netns with a
veth pair on the bridge.
'';
};
bridgeName = lib.mkOption {
type = lib.types.str;
default = "hive-br0";
example = "h0";
description = ''
Name of the host-side bridge interface the hive uses for
inter-container traffic. Kept short so it survives the
IFNAMSIZ (15-char) cap, and prefixed so it's obviously
hive-managed in `ip link` output.
'';
};
bridgeIp = lib.mkOption {
type = lib.types.str;
default = "10.42.0.1";
example = "172.30.0.1";
description = ''
IPv4 address assigned to the bridge interface on the host
side. Becomes the DNS server address agents point at (and
the upstream the gateway proxies to once netns isolation
lands). Default `10.42.0.1` is in RFC 1918 space and
unlikely to clash with operator's existing setup; override
if a different range is already in use.
'';
};
bridgePrefixLength = lib.mkOption {
type = lib.types.int;
default = 24;
example = 16;
description = ''
Netmask prefix length for the bridge subnet. Default `/24`
gives 254 usable per-agent addresses, enough for any
single-host hive. Operator with a larger swarm or a tighter
addressing scheme overrides.
'';
};
upstreamDns = lib.mkOption {
type = lib.types.listOf lib.types.str;
default = [
"1.1.1.1"
"9.9.9.9"
];
example = [
"192.168.1.1"
"8.8.8.8"
];
description = ''
Upstream DNS servers dnsmasq forwards non-hive queries to.
Defaults to Cloudflare + Quad9. Override for operators on
private networks who need a specific resolver (corporate
DNS, pi-hole, etc.). The hive resolver itself stays
authoritative for `<hive-domain>` and its sub-domains
regardless of upstream choice.
'';
};
isolateContainers = lib.mkOption {
type = lib.types.bool;
default = true;
example = true;
description = ''
**DEPRECATED ignored.** Network isolation is now the only mode and
is always on whenever hyperhive is enabled; the shared-netns path was
removed. This option is retained as a no-op so existing configs eval;
setting it to `false` warns and has no effect. It will be removed in
a future release.
Each agent container gets a dedicated veth
pair attached to `bridgeName` and a deterministic IP from
the bridge subnet. The bridge (already up when `enable = true`)
becomes the sole routed path between the host and agent
containers.
The host-side nix effect (this option) is:
- Sets `HIVE_NETWORK_ISOLATION=1` in the c0re service env so
the Rust lifecycle knows to pass `--private-network` +
bridge settings when creating/updating containers.
- Enables IP forwarding + NAT so agents can reach the internet
through the host.
- Adds a firewall rule DROP'ing traffic from the bridge subnet
to the host's loopback addresses defence-in-depth so a
compromised agent can't reach the c0re dashboard (already
bound to 127.0.0.1) or other host-loopback services even if
the routing table somehow leaks.
- Allows HTTP/HTTPS (80/443) traffic from the bridge subnet to
the host so agents can reach the gateway container (shared
host netns, proxies the operator's per-agent UI).
**Prerequisite**: all agents must have
`hyperhive.web.useUnixSocket = true` before enabling isolation.
Agents that still bind TCP on `0.0.0.0:<port>` will be
reachable at their bridge IP from other agents on the same
subnet defeating the isolation goal. The gateway routes via
unix sockets so gateway reach still works regardless.
**Migration**: containers are destroyed and re-created when
the network isolation flag flips. Operator state under
`/agents/<name>/state/` is bind-mounted and survives; the
container rootfs (nix store paths) is recreated cleanly.
**Rust counterpart**: `hive-c0re` reads `HIVE_NETWORK_ISOLATION`
and `HIVE_NETWORK_BRIDGE` from its service env and uses them in
`lifecycle::set_nspawn_flags` to configure `PRIVATE_NETWORK`,
`LOCAL_ADDRESS`, and `HOST_BRIDGE` in each container's
`nixos-containers/<name>.conf`. See `docs/network.md` for the
full design.
'';
};
};
config = lib.mkMerge [
# The hive network + container isolation are unconditional whenever
# hyperhive is enabled: the shared-netns mode was removed, so there is
# one mode (private netns behind the bridge). `network.enable` and
# `isolateContainers` are kept as deprecated no-op options (see the
# warnings block below) so existing configs that set them still eval.
(lib.mkIf config.services.hyperhive.enable {
assertions = [
{
assertion = config.services.hyperhive.domain != null;
message = ''
hyperhive requires services.hyperhive.domain to be set the
hive resolver is authoritative for `<hive-domain>` and its
sub-domains, and agents reach the forge/matrix through the
gateway by that domain. Pin a hostname
(`services.hyperhive.domain = "example.com";`).
'';
}
];
# Virtual bridge — each agent container attaches a veth pair (isolation
# is unconditional now).
networking.bridges.${cfg.bridgeName}.interfaces = [ ];
# Bridge IP — dnsmasq (in the gateway container) binds here.
networking.interfaces.${cfg.bridgeName}.ipv4.addresses = [
{
address = cfg.bridgeIp;
prefixLength = cfg.bridgePrefixLength;
}
];
# DNS only on the bridge interface — no external amplification surface.
networking.firewall.interfaces.${cfg.bridgeName} = {
allowedUDPPorts = [ 53 ];
allowedTCPPorts = [ 53 ];
};
})
# Container isolation overlay — now unconditional (the shared-netns
# mode was removed). See docs/network.md#container-isolation.
(lib.mkIf config.services.hyperhive.enable {
# Agents route internet traffic via the bridge; NAT masquerades their RFC-1918 IPs.
boot.kernel.sysctl."net.ipv4.ip_forward" = 1;
networking.nat = {
enable = true;
internalInterfaces = [ cfg.bridgeName ];
};
# Defence-in-depth: DROP bridge→loopback so compromised agents can't
# reach host-loopback services even via routing table leaks.
networking.firewall.extraInputRules = ''
ip saddr ${cfg.bridgeIp}/${toString cfg.bridgePrefixLength} ip daddr 127.0.0.0/8 drop
'';
# Allow isolated agents to reach the gateway (nginx on the host, shared
# netns). Port 80 covers `http://forge.<domain>`, per-agent UI proxies,
# and any other HTTP services the gateway fronts. Port 443 for HTTPS.
networking.firewall.interfaces.${cfg.bridgeName}.allowedTCPPorts = [
80
443
];
# Tells hive-c0re to pass PRIVATE_NETWORK + bridge settings to each
# container. HIVE_NETWORK_SUBNET is host-bridge IP/prefix, not canonical
# network address — the Rust side normalises before subnet arithmetic.
systemd.services.hive-c0re.environment = {
HIVE_NETWORK_ISOLATION = "1";
HIVE_NETWORK_BRIDGE = cfg.bridgeName;
HIVE_NETWORK_SUBNET = "${cfg.bridgeIp}/${toString cfg.bridgePrefixLength}";
};
})
# Deprecation surface for the removed toggles. Both options are kept so
# existing configs that set them to `true` still eval cleanly; setting
# either to `false` no longer does anything (network + isolation are
# unconditional now), so warn rather than silently ignore.
{
# Only warn when hyperhive itself is enabled — otherwise `cfg.enable`
# defaults to `false` (tracking `hyperhive.enable`) and we'd fire a
# spurious deprecation warning on a host that doesn't run hyperhive.
warnings = lib.optionals config.services.hyperhive.enable (
lib.optional (!cfg.enable) ''
services.hyperhive.network.enable = false is deprecated and ignored
the hive network is now always on (private-netns isolation is the
only mode). Remove the setting.
''
++ lib.optional (!cfg.isolateContainers) ''
services.hyperhive.network.isolateContainers = false is deprecated
and ignored network isolation is now the only mode and is always
on. Remove the setting.
''
);
}
];
}