Run the controller as a Go bundle the host starts as a process (hq issue 213)

The controller is a Go program and was the one piece of the mesh's own Go
code still shipped and run as an image (novox/hq issue 213; ADR 0188 §1:
a module's own code is bundles; §3: a service bundle is a process).

The manifest now builds one Go bundle, `controller`, and runs it as the
process `mesh-controller` (`./mesh-controller serve`) under an account
the module declares. What the container gave it, replaced:

- host network: a process is on the host's network; nothing it reads
  names a container network
- user 65534: the account `mesh-controller`, which owns its secrets and
  its state directory
- the eight mounts: the env names the host paths the mesh already places
  (the store, broker and bus files under the state directory, the
  broker's certificate under /var/lib/mesh-broker-tls); the `broker`
  mount was read by nothing and is gone with the others
- `container-runtime` is no longer required on its machine

Its preparation is the same binary with `prepare`, as a run-once process,
and the process `replaces` the container `server`: the host keeps the
container answering until the process is running (mesh-host). Needs the
previous commit live in the running controller, and the host's
`replaces` on the controller's machine, before it is registered.

No image is built by the mesh any more. The Dockerfile stays for genesis
and the lab (`make image`, its Go base now pinned in the Makefile).
This commit is contained in:
jochen
2026-10-04 01:11:31 +02:00
parent e11caecdad
commit 635363adbd
8 changed files with 196 additions and 59 deletions
@@ -0,0 +1,131 @@
package catalogue
import (
"encoding/json"
"fmt"
"os"
"strings"
"testing"
)
// novox/hq issue 213: the controller is a Go program and was the one piece of the mesh's own Go
// code still shipped as an image (ADR 0188 §1). Its own manifest, composed for the machine that runs
// it, is a Go bundle run by the host as a process — and no container.
func TestTheControllerIsAProcessAndNoContainer(t *testing.T) {
raw, err := os.ReadFile("../../module.json")
if err != nil {
t.Fatal(err)
}
m, err := ParseManifest(raw)
if err != nil {
t.Fatalf("the controller's own manifest does not parse:\n%v", err)
}
if m.Build == nil || len(m.Build.Artifacts) != 1 {
t.Fatalf("the controller builds %+v; it is one bundle", m.Build)
}
a := m.Build.Artifacts[0]
if a.Kind != ArtifactBundle || a.Language != "go" || a.System == "" || BinaryOf(a) != "mesh-controller" {
t.Fatalf("the controller's artifact is %+v, not a Go bundle naming its system and binary", a)
}
for _, c := range m.Capabilities {
if c == "container-runtime" {
t.Error("the controller still requires a container runtime on its machine")
}
}
digest := "sha256:" + strings.Repeat("c", 64)
control, err := m.Resolve([]Built{{Name: a.Name, Kind: ArtifactBundle,
Reference: ArtifactStoreScheme + "mesh-controller/" + a.Name + "@" + digest, Digest: digest}})
if err != nil {
t.Fatal(err)
}
// The node's runtime does not launch it: it serves its seat's verbs itself.
if loads := control.Bundles[0].Loads; len(loads) != 0 {
t.Errorf("the node's runtime would launch the controller as a tools bundle: %v", loads)
}
needed := map[string]map[string]string{"mesh-controller": {}}
for name := range m.OwnSecrets {
needed["mesh-controller"][name] = "sealed-" + name
}
out, err := Resolution{Node: "anchor", Modules: []Manifest{control}}.Declaration(Rendering{
Needed: needed, ArtifactStore: "anchor.internal:5100",
Seats: map[string]map[int]int{"mesh-store": {5432: 6852}},
})
if err != nil {
t.Fatalf("the controller does not compose: %v", err)
}
var process, step map[string]any
account, firstSecret := -1, -1
for i, r := range out {
switch {
case r["type"] == "container":
t.Errorf("the controller's declaration still runs a container: %v", r)
case r["id"] == "mesh-controller.controller":
process = r
case r["id"] == "mesh-controller.controller-prepare":
step = r
if process != nil {
t.Error("the controller's preparation is placed after the process it prepares for")
}
case r["type"] == "user" && r["name"] == "mesh-controller":
account = i
case strings.HasPrefix(fmt.Sprint(r["id"]), "mesh-controller.needs-") && firstSecret < 0:
firstSecret = i
}
}
if process == nil {
t.Fatalf("the controller's process is not in its declaration: %v", out)
}
if run, _ := json.Marshal(process["run"]); string(run) != `["./mesh-controller","serve"]` {
t.Errorf("the controller is run as %s, not its own bundle's binary", run)
}
if process["source"] != "anchor.internal:5100/mesh-controller/"+a.Name+"@"+digest || process["digest"] != digest {
t.Errorf("the controller's bundle is fetched from %v (%v)", process["source"], process["digest"])
}
// The user: an account the host declares, which owns what the process reads.
if process["user"] != "mesh-controller" || account < 0 {
t.Errorf("the controller runs as %v, and the account declared is at %d", process["user"], account)
}
if firstSecret >= 0 && account > firstSecret {
t.Error("the controller's secrets are written before the account they belong to exists")
}
for _, r := range out {
if strings.HasPrefix(fmt.Sprint(r["id"]), "mesh-controller.needs-") && r["owner"] != "mesh-controller" {
t.Errorf("%v belongs to %v, which the controller's process cannot read", r["id"], r["owner"])
}
}
if dir := fileNamed(out, "mesh-controller.mesh-state"); dir == nil || dir["owner"] != "mesh-controller" {
t.Errorf("the controller's state directory is not its account's to enter: %v", dir)
}
// Each mount became a path the process reads: nothing it is told is a path inside a container.
state := fmt.Sprint(fileNamed(out, "mesh-controller.mesh-state")["path"])
env, _ := process["env"].(map[string]any)
for key, value := range env {
v := fmt.Sprint(value)
if strings.HasPrefix(v, "/run/secrets") || strings.HasPrefix(v, "/broker-tls") {
t.Errorf("%s=%s is a path inside the container the controller no longer runs in", key, v)
}
if strings.HasSuffix(key, "_FILE") && !strings.HasPrefix(v, state+"/") {
t.Errorf("%s=%s is not one of the files the mesh places for it", key, v)
}
}
if env["MESH_BROKER_CERTIFICATE"] != "/var/lib/mesh-broker-tls/tls.crt" {
t.Errorf("the controller reads the broker's certificate from %v", env["MESH_BROKER_CERTIFICATE"])
}
if env["MESH_STORE_INVENTORY_PORT"] != "6852" {
t.Errorf("the controller is told the store is on %v; the node put it on 6852", env["MESH_STORE_INVENTORY_PORT"])
}
// The handover: the container it ran as goes only once this is running.
if got, _ := json.Marshal(process["replaces"]); string(got) != `["mesh-controller.server"]` {
t.Errorf("the controller's process replaces %s, not the container it ran as", got)
}
// And its state is prepared first, by the same program as the same account.
if step == nil || step["run-once"] != true || step["user"] != "mesh-controller" {
t.Fatalf("the controller's preparation is %v", step)
}
if run, _ := json.Marshal(step["run"]); string(run) != `["./mesh-controller","prepare"]` {
t.Errorf("the controller's preparation runs %s", run)
}
}
+10 -9
View File
@@ -101,7 +101,7 @@ func TestTheControlPlanesOwnAddressesFollowTheNodesPorts(t *testing.T) {
}
// The manifest itself names them now; withSeatPorts is a no-op on it, and this holds it so.
for _, r := range m.Resources {
if r["type"] != "container" {
if r["type"] != "process" {
continue
}
env, _ := r["env"].(map[string]any)
@@ -113,8 +113,9 @@ func TestTheControlPlanesOwnAddressesFollowTheNodesPorts(t *testing.T) {
}
m = withSeatPorts(m)
control, err := m.Resolve([]Built{{
Name: "server", Kind: ArtifactImage,
Reference: ArtifactStoreScheme + "mesh-controller/server@sha256:" + strings.Repeat("c", 64),
Name: "controller", Kind: ArtifactBundle,
Reference: ArtifactStoreScheme + "mesh-controller/controller@sha256:" + strings.Repeat("c", 64),
Digest: "sha256:" + strings.Repeat("c", 64),
}})
if err != nil {
t.Fatal(err)
@@ -135,9 +136,9 @@ func TestTheControlPlanesOwnAddressesFollowTheNodesPorts(t *testing.T) {
if err != nil {
t.Fatalf("the control plane does not compose: %v", err)
}
server := fileNamed(out, "mesh-controller.server")
server := fileNamed(out, "mesh-controller.controller")
if server == nil {
t.Fatalf("the control plane's container is not in the declaration: %v", out)
t.Fatalf("the control plane's process is not in the declaration: %v", out)
}
env, _ := server["env"].(map[string]any)
for key, want := range map[string]string{
@@ -151,8 +152,8 @@ func TestTheControlPlanesOwnAddressesFollowTheNodesPorts(t *testing.T) {
t.Errorf("the control plane is told %s=%v; the node put it on %s", key, env[key], want)
}
}
if got := server["image"]; got != "anchor.internal:5100/mesh-controller/server@sha256:"+strings.Repeat("c", 64) {
t.Errorf("the control plane's own image is %v, not routed through the store", got)
if got := server["source"]; got != "anchor.internal:5100/mesh-controller/controller@sha256:"+strings.Repeat("c", 64) {
t.Errorf("the control plane's own bundle is fetched from %v, not routed through the store", got)
}
// And on a mesh where the foundation is where genesis raised it, nothing is added.
@@ -160,7 +161,7 @@ func TestTheControlPlanesOwnAddressesFollowTheNodesPorts(t *testing.T) {
if err != nil {
t.Fatal(err)
}
env, _ = fileNamed(out, "mesh-controller.server")["env"].(map[string]any)
env, _ = fileNamed(out, "mesh-controller.controller")["env"].(map[string]any)
if env["MESH_STORE_INVENTORY_PORT"] != "" {
t.Errorf("with no settings, the control plane is told %v", env)
}
@@ -186,7 +187,7 @@ func withSeatPorts(m Manifest) Manifest {
out := m
out.Resources = nil
for _, r := range m.Resources {
if r["type"] != "container" {
if r["type"] != "container" && r["type"] != "process" {
out.Resources = append(out.Resources, r)
continue
}