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).
This commit is contained in:
atlas 2026-06-17 15:29:52 +02:00
commit 78482cb624
3 changed files with 74 additions and 54 deletions

View file

@ -756,10 +756,13 @@ in
# needed — nginx proxies to forgejo as it does for the operator. # needed — nginx proxies to forgejo as it does for the operator.
# - Shared netns: host loopback is reachable, use direct port. # - Shared netns: host loopback is reachable, use direct port.
# See `docs/gateway.md::HIVE_FORGE_URL`. # See `docs/gateway.md::HIVE_FORGE_URL`.
# Network isolation is now unconditional (the shared-netns mode was
# removed), so agents always reach the forge through the gateway
# vhost rather than host loopback. The `network.enable` guard remains
# only so a deliberately network-less deployment still falls back to
# loopback; in the normal case it is always on.
HIVE_FORGE_URL = HIVE_FORGE_URL =
if if config.services.hyperhive.network.enable then
config.services.hyperhive.network.enable && config.services.hyperhive.network.isolateContainers
then
"http://${config.services.hyperhive.forge.domain}" "http://${config.services.hyperhive.forge.domain}"
else else
"http://127.0.0.1:${toString config.services.hyperhive.forge.httpPort}"; "http://127.0.0.1:${toString config.services.hyperhive.forge.httpPort}";
@ -767,19 +770,17 @@ in
// lib.optionalAttrs config.services.hyperhive.matrix.enable { // lib.optionalAttrs config.services.hyperhive.matrix.enable {
# In-cluster matrix homeserver URL for each agent's # In-cluster matrix homeserver URL for each agent's
# hive-matrix-daemon. Same shape + rationale as HIVE_FORGE_URL: # hive-matrix-daemon. Same shape + rationale as HIVE_FORGE_URL:
# - Isolated (private netns): reach tuwunel via the gateway vhost # network isolation is now unconditional, so agents reach tuwunel
# (`matrix.<domain>`) on plain http:80 — host loopback is dead. # via the gateway vhost (`matrix.<domain>`) on plain http:80 rather
# - Shared netns: direct host loopback on the tuwunel port. # than host loopback. The gatewayHost null-guard falls back to
# gatewayHost null-guard falls back to loopback so a domain-less # loopback so a domain-less config still evals; the `network.enable`
# config doesn't break eval (it just won't work under isolation, # guard covers a deliberately network-less deployment. Forwarded to
# which needs a gateway anyway). Forwarded to agents by meta.rs # agents by meta.rs alongside HIVE_FORGE_URL; shares the same
# alongside HIVE_FORGE_URL; shares the same env-forwarding ordering # env-forwarding ordering caveat (value baked at config-generation
# caveat (value baked at config-generation time). # time).
HIVE_MATRIX_URL = HIVE_MATRIX_URL =
if if
config.services.hyperhive.network.enable config.services.hyperhive.network.enable && config.services.hyperhive.matrix.gatewayHost != null
&& config.services.hyperhive.network.isolateContainers
&& config.services.hyperhive.matrix.gatewayHost != null
then then
"http://${config.services.hyperhive.matrix.gatewayHost}" "http://${config.services.hyperhive.matrix.gatewayHost}"
else else

View file

@ -377,7 +377,7 @@ in
# `environment.etc."resolv.conf".text` is `nameserver <bridgeIp>`. # `environment.etc."resolv.conf".text` is `nameserver <bridgeIp>`.
# This container always shares the host netns # This container always shares the host netns
# (`privateNetwork = false`), so it reaches `bridgeIp` regardless # (`privateNetwork = false`), so it reaches `bridgeIp` regardless
# of `isolateContainers`. Network module off → inherit the host's # of agent-container isolation. Network module off → inherit the host's
# resolv.conf. See `docs/network.md`. # resolv.conf. See `docs/network.md`.
networking = lib.mkMerge [ networking = lib.mkMerge [
(lib.mkIf networkCfg.enable { (lib.mkIf networkCfg.enable {

View file

@ -19,18 +19,18 @@ in
defaultText = lib.literalExpression "config.services.hyperhive.enable"; defaultText = lib.literalExpression "config.services.hyperhive.enable";
example = false; example = false;
description = '' description = ''
Stand up the hive-internal bridge + dnsmasq resolver. **DEPRECATED ignored.** The hive network (bridge + dnsmasq
Defaults to `config.services.hyperhive.enable` so it comes resolver + private-netns isolation) is now always on whenever
on automatically with the rest of hyperhive. Requires hyperhive is enabled; setting this to `false` warns and has no
`services.hyperhive.domain` to be set the dnsmasq resolver effect. Retained as a no-op so existing configs eval; will be
is authoritative for `<hive-domain>` and its sub-domains. removed in a future release.
When enabled: a bridge interface (`bridgeName`) appears on the
host with `bridgeIp` assigned, and the hive-gateway container The network requires `services.hyperhive.domain` to be set the
runs a dnsmasq listening on that IP for `<hive-domain>` + dnsmasq resolver is authoritative for `<hive-domain>` and its
sub-domains. Agent containers still default to shared host sub-domains. A bridge interface (`bridgeName`) appears on the host
netns the endpoint is up but only used once with `bridgeIp` assigned, the hive-gateway container runs a dnsmasq
`isolateContainers = true` flips containers to private netns on that IP, and each agent container runs in a private netns with a
+ veth peers on this bridge. veth pair on the bridge.
''; '';
}; };
@ -94,11 +94,16 @@ in
isolateContainers = lib.mkOption { isolateContainers = lib.mkOption {
type = lib.types.bool; type = lib.types.bool;
default = false; default = true;
example = true; example = true;
description = '' description = ''
Flip agent containers from shared host netns to private netns. **DEPRECATED ignored.** Network isolation is now the only mode and
When true, each agent container gets a dedicated veth 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 pair attached to `bridgeName` and a deterministic IP from
the bridge subnet. The bridge (already up when `enable = true`) the bridge subnet. The bridge (already up when `enable = true`)
becomes the sole routed path between the host and agent becomes the sole routed path between the host and agent
@ -142,21 +147,27 @@ in
}; };
config = lib.mkMerge [ config = lib.mkMerge [
(lib.mkIf cfg.enable { # 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 = [ assertions = [
{ {
assertion = config.services.hyperhive.domain != null; assertion = config.services.hyperhive.domain != null;
message = '' message = ''
services.hyperhive.network.enable = true requires hyperhive requires services.hyperhive.domain to be set the
services.hyperhive.domain to be set the resolver needs a hive resolver is authoritative for `<hive-domain>` and its
domain to be authoritative for. Either pin a hostname sub-domains, and agents reach the forge/matrix through the
(`services.hyperhive.domain = "example.com";`) or set gateway by that domain. Pin a hostname
`services.hyperhive.network.enable = false` explicitly. (`services.hyperhive.domain = "example.com";`).
''; '';
} }
]; ];
# Virtual bridge — veth pairs attach when isolateContainers flips on. # Virtual bridge — each agent container attaches a veth pair (isolation
# is unconditional now).
networking.bridges.${cfg.bridgeName}.interfaces = [ ]; networking.bridges.${cfg.bridgeName}.interfaces = [ ];
# Bridge IP — dnsmasq (in the gateway container) binds here. # Bridge IP — dnsmasq (in the gateway container) binds here.
@ -174,23 +185,9 @@ in
}; };
}) })
# Guard: fires unconditionally on isolateContainers so the assertion # Container isolation overlay — now unconditional (the shared-netns
# is not silently swallowed when enable=false. # mode was removed). See docs/network.md#container-isolation.
(lib.mkIf cfg.isolateContainers { (lib.mkIf config.services.hyperhive.enable {
assertions = [
{
assertion = cfg.enable;
message = ''
services.hyperhive.network.isolateContainers = true requires
services.hyperhive.network.enable = true (the bridge and
resolver must be running before isolation is flipped on).
'';
}
];
})
# Container isolation overlay — see docs/network.md#container-isolation.
(lib.mkIf (cfg.enable && cfg.isolateContainers) {
# Agents route internet traffic via the bridge; NAT masquerades their RFC-1918 IPs. # Agents route internet traffic via the bridge; NAT masquerades their RFC-1918 IPs.
boot.kernel.sysctl."net.ipv4.ip_forward" = 1; boot.kernel.sysctl."net.ipv4.ip_forward" = 1;
@ -222,5 +219,27 @@ in
HIVE_NETWORK_SUBNET = "${cfg.bridgeIp}/${toString cfg.bridgePrefixLength}"; 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.
''
);
}
]; ];
} }