Fix dnsmasq reconciliation to use the correct DNS target IP: - tcp-passthrough (k8s instances): DNS → k8s LB IP (LAN direct) - http/static (Central apps): DNS → Central IP (HAProxy terminates) Also handle ExtraDomains properly: - Internal domains (internal.cloud.payne.io) get local=/ entries to prevent upstream DNS forwarding - Extra domains get their own dnsmasq address= entries - HAProxy already propagated ExtraDomains for SNI ACLs Verified: cloud.payne.io → 192.168.8.240 (direct to k8s), cloud2.payne.io → 192.168.8.80, central.payne.io → 192.168.8.151. SNI routing through HAProxy works for both instances. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
303 lines
9.1 KiB
Go
303 lines
9.1 KiB
Go
package v1
|
|
|
|
import (
|
|
"fmt"
|
|
"log/slog"
|
|
"os"
|
|
"path/filepath"
|
|
"strings"
|
|
"time"
|
|
|
|
"github.com/wild-cloud/wild-central/internal/certbot"
|
|
"github.com/wild-cloud/wild-central/internal/config"
|
|
"github.com/wild-cloud/wild-central/internal/haproxy"
|
|
"github.com/wild-cloud/wild-central/internal/network"
|
|
"github.com/wild-cloud/wild-central/internal/services"
|
|
"github.com/wild-cloud/wild-central/internal/sse"
|
|
)
|
|
|
|
// reconcileNetworking reads all registered services and regenerates
|
|
// dnsmasq DNS entries and HAProxy routes to match. Called automatically
|
|
// whenever a service is registered, updated, or deregistered.
|
|
func (api *API) reconcileNetworking() {
|
|
svcs, err := api.services.List()
|
|
if err != nil {
|
|
slog.Error("reconcile: failed to list services", "error", err)
|
|
return
|
|
}
|
|
|
|
globalConfigPath := filepath.Join(api.dataDir, "config.yaml")
|
|
globalCfg, err := config.LoadGlobalConfig(globalConfigPath)
|
|
if err != nil {
|
|
slog.Warn("reconcile: failed to load global config, using empty", "error", err)
|
|
globalCfg = &config.GlobalConfig{}
|
|
}
|
|
|
|
// Build HAProxy routes from registered services
|
|
var instanceRoutes []haproxy.InstanceRoute
|
|
var httpRoutes []haproxy.HTTPRoute
|
|
|
|
for _, svc := range svcs {
|
|
if svc.Reach == services.ReachOff {
|
|
continue
|
|
}
|
|
|
|
switch svc.Backend.Type {
|
|
case services.BackendTCPPassthrough:
|
|
instanceRoutes = append(instanceRoutes, haproxy.InstanceRoute{
|
|
Name: svc.Name,
|
|
Domain: svc.Domain,
|
|
BackendIP: extractHost(svc.Backend.Address),
|
|
ExtraDomains: svc.ExtraDomains,
|
|
})
|
|
case services.BackendHTTP, services.BackendStatic:
|
|
httpRoutes = append(httpRoutes, haproxy.HTTPRoute{
|
|
Name: svc.Name,
|
|
Domain: svc.Domain,
|
|
Backend: svc.Backend.Address,
|
|
HealthPath: svc.Backend.Health,
|
|
})
|
|
}
|
|
}
|
|
|
|
// Determine central domain for HAProxy — only if its cert exists
|
|
centralDomain := ""
|
|
if d := globalCfg.Cloud.Central.Domain; d != "" {
|
|
if _, err := os.Stat(certbot.HAProxyCertPath(d)); err == nil {
|
|
centralDomain = d
|
|
} else {
|
|
slog.Debug("reconcile: skipping central domain in HAProxy (no cert)", "domain", d)
|
|
}
|
|
}
|
|
|
|
// Only include L7 HTTP routes for services that have a cert.
|
|
// HAProxy uses a cert directory — each service needs its own <domain>.pem
|
|
// or be covered by a wildcard cert in the same directory.
|
|
certsDir := "/etc/haproxy/certs/"
|
|
var activeHTTPRoutes []haproxy.HTTPRoute
|
|
for _, route := range httpRoutes {
|
|
if hasCertForDomain(certsDir, route.Domain) {
|
|
activeHTTPRoutes = append(activeHTTPRoutes, route)
|
|
} else {
|
|
slog.Warn("reconcile: skipping L7 route (no cert)", "domain", route.Domain)
|
|
}
|
|
}
|
|
|
|
// Generate and write HAProxy config
|
|
haproxyCfg := api.haproxy.GenerateWithOpts(instanceRoutes, nil, haproxy.GenerateOpts{
|
|
CentralDomain: centralDomain,
|
|
HTTPRoutes: activeHTTPRoutes,
|
|
WildcardCert: certsDir,
|
|
})
|
|
|
|
if err := api.haproxy.WriteConfig(haproxyCfg); err != nil {
|
|
slog.Error("reconcile: failed to write HAProxy config", "error", err)
|
|
} else {
|
|
if err := api.haproxy.ReloadService(); err != nil {
|
|
slog.Warn("reconcile: failed to reload HAProxy", "error", err)
|
|
} else {
|
|
api.broadcastHaproxyEvent("haproxy:config", "HAProxy config regenerated from registered services")
|
|
}
|
|
}
|
|
|
|
// Generate dnsmasq DNS entries from registered services.
|
|
// DNS target IP depends on service type:
|
|
// tcp-passthrough (k8s): DNS → k8s LB IP (LAN traffic goes direct to k8s)
|
|
// http/static: DNS → Central IP (HAProxy terminates TLS)
|
|
centralIP, _ := network.GetWildCentralIP()
|
|
if centralIP == "" {
|
|
centralIP = "127.0.0.1"
|
|
}
|
|
|
|
var instanceConfigs []config.InstanceConfig
|
|
for _, svc := range svcs {
|
|
if svc.Reach == services.ReachOff || svc.Domain == "" {
|
|
continue
|
|
}
|
|
|
|
// Choose the DNS target IP based on service type
|
|
dnsIP := centralIP
|
|
if svc.Backend.Type == services.BackendTCPPassthrough {
|
|
// k8s instances: LAN clients connect directly to the k8s LB
|
|
dnsIP = extractHost(svc.Backend.Address)
|
|
}
|
|
|
|
// Primary domain
|
|
ic := config.InstanceConfig{}
|
|
ic.Cloud.Domain = svc.Domain
|
|
ic.Cluster.LoadBalancerIp = dnsIP
|
|
|
|
// Extra domains (e.g., internal.cloud.payne.io)
|
|
for _, extra := range svc.ExtraDomains {
|
|
if strings.HasPrefix(extra, "internal.") {
|
|
// Internal-only domain: local=/ prevents upstream DNS forwarding
|
|
ic.Cloud.InternalDomain = extra
|
|
}
|
|
}
|
|
|
|
instanceConfigs = append(instanceConfigs, ic)
|
|
}
|
|
|
|
if err := api.dnsmasq.UpdateConfig(globalCfg, instanceConfigs, true); err != nil {
|
|
slog.Error("reconcile: failed to update dnsmasq", "error", err)
|
|
} else {
|
|
api.broadcastDnsmasqEvent("dnsmasq:config", "DNS config regenerated from registered services")
|
|
}
|
|
|
|
// Ensure TLS certs exist for HTTP services (where Central terminates TLS).
|
|
// For services under the gateway domain, a wildcard cert covers them all.
|
|
// For others, provision individual certs.
|
|
if len(httpRoutes) > 0 {
|
|
api.ensureTLSCerts(globalCfg, svcs)
|
|
}
|
|
|
|
slog.Info("reconcile: networking updated",
|
|
"services", len(svcs),
|
|
"l4Routes", len(instanceRoutes),
|
|
"l7Routes", len(httpRoutes),
|
|
)
|
|
}
|
|
|
|
// ensureTLSCerts logs which certificates are missing for registered services.
|
|
// Does NOT auto-provision — cert provisioning is user-initiated via the
|
|
// Certificates page. This just logs warnings so the operator knows what's needed.
|
|
func (api *API) ensureTLSCerts(globalCfg *config.GlobalConfig, svcs []services.Service) {
|
|
gatewayDomain := ""
|
|
if parts := strings.SplitN(globalCfg.Cloud.Central.Domain, ".", 2); len(parts) == 2 {
|
|
gatewayDomain = parts[1]
|
|
}
|
|
|
|
for _, svc := range svcs {
|
|
if svc.TLS != services.TLSTerminate || svc.Domain == "" {
|
|
continue
|
|
}
|
|
|
|
// Check if covered by wildcard
|
|
if gatewayDomain != "" && strings.HasSuffix(svc.Domain, "."+gatewayDomain) {
|
|
wildcardCert := certbot.HAProxyCertPath(gatewayDomain)
|
|
if _, err := os.Stat(wildcardCert); err != nil {
|
|
slog.Warn("reconcile/tls: no wildcard cert for gateway domain — provision via Certificates page",
|
|
"service", svc.Name, "domain", svc.Domain, "needed", "*."+gatewayDomain)
|
|
}
|
|
continue
|
|
}
|
|
|
|
// Check individual cert
|
|
if _, err := os.Stat(certbot.HAProxyCertPath(svc.Domain)); err != nil {
|
|
slog.Warn("reconcile/tls: no cert for service — provision via Certificates page",
|
|
"service", svc.Name, "domain", svc.Domain)
|
|
}
|
|
}
|
|
}
|
|
|
|
// hasCertForDomain checks if a cert exists for a domain — either an individual
|
|
// cert (<domain>.pem) or a wildcard cert that covers it.
|
|
func hasCertForDomain(certsDir, domain string) bool {
|
|
// Check individual cert
|
|
if _, err := os.Stat(certbot.HAProxyCertPath(domain)); err == nil {
|
|
return true
|
|
}
|
|
// Check wildcard certs — a *.example.com cert covers foo.example.com
|
|
parts := strings.SplitN(domain, ".", 2)
|
|
if len(parts) == 2 {
|
|
// Wildcard cert might be stored as the base domain PEM
|
|
wildcardBase := parts[1]
|
|
if _, err := os.Stat(certbot.HAProxyCertPath(wildcardBase)); err == nil {
|
|
return true
|
|
}
|
|
}
|
|
// Check if the certs directory has ANY .pem files (HAProxy needs at least one)
|
|
entries, err := os.ReadDir(certsDir)
|
|
if err != nil {
|
|
return false
|
|
}
|
|
for _, e := range entries {
|
|
if strings.HasSuffix(e.Name(), ".pem") {
|
|
// There's at least one cert — HAProxy can start with the dir bind.
|
|
// The specific domain may not have its own cert but HAProxy won't crash.
|
|
// It will serve whichever cert matches best (or the first one as default).
|
|
return true
|
|
}
|
|
}
|
|
return false
|
|
}
|
|
|
|
// extractHost gets the host part from a host:port string
|
|
func extractHost(addr string) string {
|
|
for i := len(addr) - 1; i >= 0; i-- {
|
|
if addr[i] == ':' {
|
|
return addr[:i]
|
|
}
|
|
}
|
|
return addr
|
|
}
|
|
|
|
// broadcastDnsmasqEvent broadcasts SSE events for dnsmasq status changes
|
|
func (api *API) broadcastDnsmasqEvent(eventType string, message string) {
|
|
if api.sseManager == nil {
|
|
return
|
|
}
|
|
|
|
// Get current dnsmasq status
|
|
status, err := api.dnsmasq.GetStatus()
|
|
if err != nil {
|
|
status = nil
|
|
}
|
|
|
|
event := &sse.Event{
|
|
ID: fmt.Sprintf("dnsmasq-%d", time.Now().UnixNano()),
|
|
Type: eventType,
|
|
InstanceName: "global", // dnsmasq is global/central-level, not instance-specific
|
|
Timestamp: time.Now(),
|
|
Data: map[string]any{
|
|
"message": message,
|
|
"status": status,
|
|
},
|
|
}
|
|
|
|
api.sseManager.Broadcast(event)
|
|
}
|
|
|
|
// broadcastHaproxyEvent broadcasts SSE events for HAProxy status changes
|
|
func (api *API) broadcastHaproxyEvent(eventType string, message string) {
|
|
if api.sseManager == nil {
|
|
return
|
|
}
|
|
|
|
event := &sse.Event{
|
|
ID: fmt.Sprintf("haproxy-%d", time.Now().UnixNano()),
|
|
Type: eventType,
|
|
InstanceName: "global",
|
|
Timestamp: time.Now(),
|
|
Data: map[string]any{
|
|
"message": message,
|
|
},
|
|
}
|
|
|
|
api.sseManager.Broadcast(event)
|
|
}
|
|
|
|
// broadcastCentralStatusEvent broadcasts SSE events for central status changes
|
|
func (api *API) broadcastCentralStatusEvent(startTime time.Time) {
|
|
if api.sseManager == nil {
|
|
return
|
|
}
|
|
|
|
uptime := time.Since(startTime)
|
|
|
|
event := &sse.Event{
|
|
ID: fmt.Sprintf("central-%d", time.Now().UnixNano()),
|
|
Type: "central:status",
|
|
InstanceName: "global",
|
|
Timestamp: time.Now(),
|
|
Data: map[string]any{
|
|
"status": "running",
|
|
"version": api.version,
|
|
"uptime": uptime.String(),
|
|
"uptimeSeconds": int(uptime.Seconds()),
|
|
},
|
|
}
|
|
|
|
api.sseManager.Broadcast(event)
|
|
}
|