From ea29d0d12e9f2d4e0adf6b43d79e919ab2fca121 Mon Sep 17 00:00:00 2001
From: Audiobuzz <audiobuzz@gmail.com>
Date: Sat, 4 Jul 2026 08:58:38 +0000
Subject: [PATCH] fleet-manager: soft-disable optional planes when their PKI
 isn't staged yet

The compose wires the authorise (broker) and agent planes to fixed cert paths,
so a first bring-up that hasn't run the PKI steps yet crash-looped on the first
missing file (e.g. /pki/broker/clients/fleet-manager.crt) instead of booting.

Treat a configured-but-absent cert as "plane not yet provisioned": log a loud
WARNING and disable that plane, so the manager boots and each plane lights up
as its PKI appears. A malformed/unreadable cert still hard-fails (real
misconfig). The admin plane stays a hard requirement (FLEET_MGR_ADMIN_INSECURE
is its bench escape). Adds a missingFiles() helper; hub push path already
nil-guards a disabled agent plane.
---
 serving/fleet-manager/main.go | 55 ++++++++++++++++++++++++++++-------
 1 file changed, 44 insertions(+), 11 deletions(-)

diff --git a/serving/fleet-manager/main.go b/serving/fleet-manager/main.go
index 4d2f21a..dcacade 100644
--- a/serving/fleet-manager/main.go
+++ b/serving/fleet-manager/main.go
@@ -18,6 +18,7 @@ import (
 	"net/http"
 	"os"
 	"strconv"
+	"strings"
 	"sync"
 	"time"
 
@@ -132,15 +133,24 @@ func main() {
 
 	s := &server{cfg: cfg, store: store}
 
-	// Authorise pillar: the enrol-broker admin client (nil = disabled, so the
-	// manager runs standalone before a broker is wired).
-	broker, err := newBrokerClient(cfg)
-	if err != nil {
-		log.Fatalf("broker client: %v", err)
-	}
-	s.broker = broker
-	if broker == nil {
+	// Authorise pillar: the enrol-broker admin client. Enabled when the broker URL
+	// is set AND its client PKI is present. Missing PKI (not yet provisioned)
+	// DISABLES the pillar with a loud warning rather than crash-looping the whole
+	// manager — so a partial bring-up boots and the pillar lights up once the cert
+	// is staged. A malformed/unreadable cert still hard-fails (a real misconfig).
+	switch {
+	case cfg.brokerAdminURL == "":
 		log.Printf("authorise pillar DISABLED (set FLEET_MGR_BROKER_URL + client cert/key + CA to enable)")
+	default:
+		if miss := missingFiles(cfg.brokerClientCert, cfg.brokerClientKey, cfg.brokerCA); len(miss) > 0 {
+			log.Printf("WARNING: authorise pillar DISABLED — broker PKI not found: %s (run enrol-broker/admin-pki.sh; see DEPLOY.md)", strings.Join(miss, ", "))
+			break
+		}
+		broker, err := newBrokerClient(cfg)
+		if err != nil {
+			log.Fatalf("broker client: %v", err)
+		}
+		s.broker = broker
 	}
 
 	// Metrics gateway: the manager is the only reader/writer of InfluxDB. nil (a
@@ -159,15 +169,22 @@ func main() {
 	// plane can run standalone (e.g. before any host is enrolled). The hub is the
 	// live-stream registry the lifecycle API pushes desired-state through, and it
 	// writes each snapshot to the metrics sink.
-	if cfg.agentCert != "" && cfg.agentKey != "" && cfg.agentClientCA != "" {
+	switch {
+	case cfg.agentCert == "" || cfg.agentKey == "" || cfg.agentClientCA == "":
+		log.Printf("agent plane DISABLED (set FLEET_MGR_AGENT_CERT/_KEY/_CLIENT_CA to enable)")
+	default:
+		if miss := missingFiles(cfg.agentCert, cfg.agentKey, cfg.agentClientCA); len(miss) > 0 {
+			// Same soft-disable as the authorise pillar: boot without the agent
+			// plane until its fleet-CA PKI is staged, rather than crash-looping.
+			log.Printf("WARNING: agent plane DISABLED — PKI not found: %s (run server-pki.sh; see DEPLOY.md)", strings.Join(miss, ", "))
+			break
+		}
 		s.hub = newAgentHub(store, sink)
 		go func() {
 			if err := serveAgent(cfg, s.hub); err != nil {
 				log.Fatalf("agent plane: %v", err)
 			}
 		}()
-	} else {
-		log.Printf("agent plane DISABLED (set FLEET_MGR_AGENT_CERT/_KEY/_CLIENT_CA to enable)")
 	}
 
 	// Admin browser auth = WebAuthn/passkey (secure mode only). Sets up the RP,
@@ -292,6 +309,22 @@ func env(k, def string) string {
 	return def
 }
 
+// missingFiles returns the subset of the given paths that don't exist (or aren't
+// stat-able) on disk. Used to soft-disable an optional plane whose PKI hasn't
+// been provisioned yet, instead of crash-looping the whole manager.
+func missingFiles(paths ...string) []string {
+	var miss []string
+	for _, p := range paths {
+		if p == "" {
+			continue
+		}
+		if _, err := os.Stat(p); err != nil {
+			miss = append(miss, p)
+		}
+	}
+	return miss
+}
+
 func envInt(k string, def int) int {
 	if v := os.Getenv(k); v != "" {
 		if n, err := strconv.Atoi(v); err == nil {