mirror of
https://github.com/openziti/ziti.git
synced 2026-09-10 00:35:41 +00:00
d562afe1d1
Control channel send back-pressure was showing up as a p99 at the send timeout, with no way to tell which handler was holding the receive goroutine or what it was waiting on. A dump taken after the fact records the goroutine unwinding the diagnostic rather than whatever it was blocked on, so the snapshot has to happen while the handler is still in the handler. That costs a timer per message, which is why this is opt-in rather than always on. - wraps every control channel receive handler on both sides, timing it and snapshotting all goroutines while a slow one is still in the handler - installs nothing when disabled. A nil detector's Wrap returns the handler it was given, so a process that has not asked for this pays no timer, no clock read and no branch per message, rather than paying a check - deduplicates dumps by a normalized signature so a process stuck in one place writes one file and counts the repeats, rather than filling the disk with the same picture - names the subject of each dump, since the goroutine being diagnosed cannot be picked out of the dump and would otherwise be filtered out as a singleton - makes the thresholds and the dump tracking configurable, because what to dump on varies by what is being chased: 500ms finds a wedged handler, lock contention wants tens of milliseconds, and a rare event wants more distinct dumps kept and less time between them - builds one detector per process rather than per channel. Per channel would quietly turn the dump interval and the budget of distinct dumps into per connection limits, and those limits are what bound the disk a dump can cost - refuses settings that would produce nothing, such as a zero threshold or no dumps allowed, but only when enabled, so a config left over from an investigation does not stop a process starting once it is switched off - reports the router's own control channel state It lives in common/diagnostics, shared by the controller and router rather than duplicated per side, and is documented commented-out in the sample configs.
1054 lines
32 KiB
Go
1054 lines
32 KiB
Go
/*
|
|
Copyright NetFoundry Inc.
|
|
|
|
Licensed under the Apache License, Version 2.0 (the "License");
|
|
you may not use this file except in compliance with the License.
|
|
You may obtain a copy of the License at
|
|
|
|
https://www.apache.org/licenses/LICENSE-2.0
|
|
|
|
Unless required by applicable law or agreed to in writing, software
|
|
distributed under the License is distributed on an "AS IS" BASIS,
|
|
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
|
See the License for the specific language governing permissions and
|
|
limitations under the License.
|
|
*/
|
|
|
|
package controller
|
|
|
|
import (
|
|
"bytes"
|
|
"compress/gzip"
|
|
cryptoTls "crypto/tls"
|
|
"crypto/x509"
|
|
"encoding/json"
|
|
"fmt"
|
|
"io"
|
|
"os"
|
|
"strings"
|
|
"sync"
|
|
"sync/atomic"
|
|
|
|
gosundheit "github.com/AppsFlyer/go-sundheit"
|
|
"github.com/michaelquigley/pfxlog"
|
|
"github.com/openziti/channel/v5"
|
|
"github.com/openziti/channel/v5/protobufs"
|
|
"github.com/openziti/foundation/v2/concurrenz"
|
|
nfpem "github.com/openziti/foundation/v2/pem"
|
|
"github.com/openziti/foundation/v2/versions"
|
|
"github.com/openziti/identity"
|
|
"github.com/openziti/metrics"
|
|
"github.com/openziti/transport/v2"
|
|
"github.com/openziti/transport/v2/tls"
|
|
"github.com/openziti/xweb/v3"
|
|
"github.com/openziti/ziti/v2/common/bindpoints"
|
|
"github.com/openziti/ziti/v2/common/capabilities"
|
|
"github.com/openziti/ziti/v2/common/concurrency"
|
|
"github.com/openziti/ziti/v2/common/pb/ctrl_pb"
|
|
"github.com/openziti/ziti/v2/common/profiler"
|
|
"github.com/openziti/ziti/v2/common/servermetrics"
|
|
"github.com/openziti/ziti/v2/controller/command"
|
|
"github.com/openziti/ziti/v2/controller/config"
|
|
"github.com/openziti/ziti/v2/controller/db"
|
|
"github.com/openziti/ziti/v2/controller/env"
|
|
"github.com/openziti/ziti/v2/controller/event"
|
|
"github.com/openziti/ziti/v2/controller/events"
|
|
"github.com/openziti/ziti/v2/controller/gossip"
|
|
"github.com/openziti/ziti/v2/controller/handler_ctrl"
|
|
"github.com/openziti/ziti/v2/controller/handler_peer_ctrl"
|
|
"github.com/openziti/ziti/v2/controller/network"
|
|
"github.com/openziti/ziti/v2/controller/raft"
|
|
"github.com/openziti/ziti/v2/controller/raft/mesh"
|
|
"github.com/openziti/ziti/v2/controller/storage/boltz"
|
|
"github.com/openziti/ziti/v2/controller/webapis"
|
|
"github.com/openziti/ziti/v2/controller/xctrl"
|
|
"github.com/openziti/ziti/v2/controller/xmgmt"
|
|
"github.com/openziti/ziti/v2/controller/xt"
|
|
"github.com/openziti/ziti/v2/controller/xt_random"
|
|
"github.com/openziti/ziti/v2/controller/xt_smartrouting"
|
|
"github.com/openziti/ziti/v2/controller/xt_sticky"
|
|
"github.com/openziti/ziti/v2/controller/xt_weighted"
|
|
"github.com/pkg/errors"
|
|
"github.com/sirupsen/logrus"
|
|
"github.com/teris-io/shortid"
|
|
"go.etcd.io/bbolt"
|
|
)
|
|
|
|
type Controller struct {
|
|
config *config.Config
|
|
env *env.AppEnv
|
|
network *network.Network
|
|
raftController *raft.Controller
|
|
localDispatcher *command.LocalDispatcher
|
|
ctrlConnectHandler *handler_ctrl.ConnectHandler
|
|
xctrls []xctrl.Xctrl
|
|
xmgmts concurrenz.CopyOnWriteSlice[xmgmt.Xmgmt]
|
|
|
|
xwebFactoryRegistry xweb.Registry
|
|
xweb xweb.Instance
|
|
|
|
ctrlListener io.Closer
|
|
mgmtListener channel.UnderlayListener
|
|
|
|
shutdownC chan struct{}
|
|
isShutdown atomic.Bool
|
|
agentBindHandlers []channel.BindHandler
|
|
metricsRegistry metrics.Registry
|
|
versionProvider versions.VersionProvider
|
|
eventDispatcher *events.Dispatcher
|
|
|
|
apiData map[string][]event.ApiAddress
|
|
apiDataBytes []byte
|
|
apiDataOnce sync.Once
|
|
|
|
xwebInitialized concurrency.InitState
|
|
healthChecker gosundheit.Health
|
|
}
|
|
|
|
func init() {
|
|
xweb.BindPointListenerFactoryRegistry = append(xweb.BindPointListenerFactoryRegistry, &bindpoints.BindPointListenerFactory{})
|
|
}
|
|
|
|
func (c *Controller) GetPeerSigners() []*x509.Certificate {
|
|
if c.raftController == nil || c.raftController.Mesh == nil {
|
|
return nil
|
|
}
|
|
|
|
var certs []*x509.Certificate
|
|
|
|
for _, peer := range c.raftController.Mesh.GetPeers() {
|
|
certs = append(certs, peer.SigningCerts...)
|
|
}
|
|
|
|
return certs
|
|
}
|
|
|
|
func (c *Controller) GetPeerAddresses() []string {
|
|
if c.raftController == nil || c.raftController.Mesh == nil {
|
|
return nil
|
|
}
|
|
|
|
var addresses []string
|
|
|
|
for _, peer := range c.raftController.Mesh.GetPeers() {
|
|
addresses = append(addresses, peer.Address)
|
|
}
|
|
|
|
return addresses
|
|
|
|
}
|
|
|
|
func (c *Controller) GetId() *identity.TokenId {
|
|
return c.config.Id
|
|
}
|
|
|
|
func (c *Controller) GetConfig() *config.Config {
|
|
return c.config
|
|
}
|
|
|
|
func (c *Controller) GetMetricsRegistry() metrics.Registry {
|
|
return c.metricsRegistry
|
|
}
|
|
|
|
func (c *Controller) GetOptions() *config.NetworkConfig {
|
|
return c.config.Network
|
|
}
|
|
|
|
func (c *Controller) GetCommandDispatcher() command.Dispatcher {
|
|
if c.raftController == nil {
|
|
if c.localDispatcher != nil {
|
|
return c.localDispatcher
|
|
}
|
|
devVersion := versions.MustParseSemVer("0.0.0")
|
|
version := versions.MustParseSemVer(c.GetVersionProvider().Version())
|
|
c.localDispatcher = &command.LocalDispatcher{
|
|
EncodeDecodeCommands: devVersion.Equals(version),
|
|
Limiter: command.NewRateLimiter(c.config.Command.RateLimiter, c.metricsRegistry, c.shutdownC),
|
|
}
|
|
return c.localDispatcher
|
|
}
|
|
return c.raftController
|
|
}
|
|
|
|
func (c *Controller) IsRaftEnabled() bool {
|
|
return c.raftController != nil
|
|
}
|
|
|
|
func (c *Controller) IsRaftLeader() bool {
|
|
if c.raftController == nil {
|
|
return false
|
|
}
|
|
return c.raftController.IsLeader()
|
|
}
|
|
|
|
func (c *Controller) GetRaftIndex() uint64 {
|
|
return c.raftController.GetAppliedIndex()
|
|
}
|
|
|
|
func (c *Controller) GetStartRaftIndex() uint64 {
|
|
return c.raftController.Fsm.GetStartIndex()
|
|
}
|
|
|
|
func (c *Controller) GetRaftInfo() (string, string, string) {
|
|
id := c.config.Id.Token
|
|
addr := c.raftController.Mesh.Addr().String()
|
|
|
|
version := c.GetVersionProvider().Version()
|
|
|
|
return addr, id, version
|
|
}
|
|
|
|
func (c *Controller) GetDb() boltz.Db {
|
|
return c.config.Db
|
|
}
|
|
|
|
func (c *Controller) GetVersionProvider() versions.VersionProvider {
|
|
return c.versionProvider
|
|
}
|
|
|
|
func (c *Controller) GetCloseNotify() <-chan struct{} {
|
|
return c.shutdownC
|
|
}
|
|
|
|
// GetGossipPeering tells the network how to reach its peers. The raft controller is created and initialized
|
|
// before the network is, so its mesh is available here and gossip needs nothing handed to it later. Without
|
|
// raft there are no peers, and the zero value puts the network in single-controller mode.
|
|
func (c *Controller) GetGossipPeering() network.GossipPeering {
|
|
if c.raftController == nil {
|
|
return network.GossipPeering{}
|
|
}
|
|
return network.GossipPeering{
|
|
Mesh: gossip.NewRaftMeshAdapter(c.raftController.GetMesh()),
|
|
IsLeader: c.raftController.IsLeader,
|
|
}
|
|
}
|
|
|
|
func (c *Controller) GetRaftConfig() *config.RaftConfig {
|
|
return c.config.Raft
|
|
}
|
|
|
|
func (c *Controller) GetRaftRateLimiterConfig() command.AdaptiveRateLimitTrackerConfig {
|
|
return c.config.Raft.RateLimiter
|
|
}
|
|
|
|
func (c *Controller) RenderJsonConfig() (string, error) {
|
|
return c.config.ToJson()
|
|
}
|
|
|
|
func (c *Controller) GetEnv() *env.AppEnv {
|
|
return c.env
|
|
}
|
|
|
|
func NewController(cfg *config.Config, versionProvider versions.VersionProvider) (*Controller, error) {
|
|
metricRegistry := metrics.NewRegistry(cfg.Id.Token, nil)
|
|
|
|
shutdownC := make(chan struct{})
|
|
|
|
tlsHandshakeRateLimiter := command.NewAdaptiveRateLimitTracker(cfg.TlsHandshakeRateLimiter, metricRegistry, shutdownC)
|
|
tls.SetSharedListenerRateLimiter(tlsHandshakeRateLimiter)
|
|
|
|
log := pfxlog.Logger()
|
|
|
|
c := &Controller{
|
|
config: cfg,
|
|
shutdownC: shutdownC,
|
|
xwebFactoryRegistry: xweb.NewRegistryMap(),
|
|
metricsRegistry: metricRegistry,
|
|
versionProvider: versionProvider,
|
|
eventDispatcher: events.NewDispatcher(shutdownC),
|
|
xwebInitialized: concurrency.NewInitState(),
|
|
}
|
|
xwebInstanceOptions := xweb.InstanceOptions{
|
|
DefaultIdentity: c.config.Id,
|
|
DefaultIdentitySection: xweb.DefaultIdentitySection,
|
|
DefaultConfigSection: xweb.DefaultConfigSection,
|
|
InstanceValidators: []xweb.InstanceValidator{
|
|
c.ensureOidcOnClientApiServer,
|
|
},
|
|
ServerMutators: []xweb.ServerMutator{
|
|
func(instance xweb.Instance, serverConfig *xweb.ServerConfig, server *xweb.Server) error {
|
|
for _, httpServer := range server.HttpServers {
|
|
serverTlsConfig := httpServer.TLSConfig.Clone()
|
|
|
|
prev := serverTlsConfig.GetConfigForClient
|
|
httpServer.TLSConfig.GetConfigForClient = func(info *cryptoTls.ClientHelloInfo) (*cryptoTls.Config, error) {
|
|
result := serverTlsConfig
|
|
|
|
//use inner GetConfigForClient if it was defined
|
|
if prev != nil {
|
|
var err error
|
|
result, err = prev(info)
|
|
|
|
if err != nil {
|
|
return nil, err
|
|
}
|
|
}
|
|
|
|
result.ClientCAs = c.env.GetManagers().Ca.GetTrustCache().GetAllPool()
|
|
|
|
return result, nil
|
|
}
|
|
}
|
|
return nil
|
|
},
|
|
},
|
|
}
|
|
|
|
c.xweb = xweb.NewInstance(c.xwebFactoryRegistry, xwebInstanceOptions)
|
|
|
|
if cfg.IsRaftEnabled() {
|
|
c.raftController = raft.NewController(c, c)
|
|
if err := c.raftController.Init(); err != nil {
|
|
log.WithError(err).Panic("error starting raft")
|
|
}
|
|
|
|
cfg.Db = c.raftController.GetDb()
|
|
}
|
|
|
|
c.registerXts()
|
|
|
|
appEnv, err := env.NewAppEnv(c)
|
|
if err != nil {
|
|
return nil, err
|
|
}
|
|
|
|
c.env = appEnv
|
|
|
|
if n, err := network.NewNetwork(c, appEnv); err == nil {
|
|
c.network = n
|
|
n.SetRestartSelfOnSnapshot(c.config.Raft != nil && c.config.Raft.RestartSelf)
|
|
} else {
|
|
return nil, err
|
|
}
|
|
|
|
if c.raftController != nil {
|
|
if err = c.raftController.InitEnv(c.env); err != nil {
|
|
return nil, err
|
|
}
|
|
}
|
|
|
|
c.initWeb() // need to init web before bootstrapping, so we can provide our endpoints to peers
|
|
|
|
if c.raftController != nil {
|
|
_, dbConfigured := c.config.Src["db"]
|
|
if c.raftController.IsBootstrapped() {
|
|
// On a clustered config 'db' only seeds a new cluster on first bootstrap; once
|
|
// initialized it is dead config, so warn rather than let a stale setting sit unnoticed.
|
|
if dbConfigured {
|
|
log.Warn("'db' is set but this clustered controller is already initialized; the 'db' setting is ignored and should be removed from the configuration")
|
|
}
|
|
} else if err = c.TryInitializeRaftFromBoltDb(); err != nil {
|
|
log.WithError(err).Panic("error bootstrapping raft")
|
|
}
|
|
}
|
|
|
|
c.eventDispatcher.InitializeNetworkEvents(c.network)
|
|
|
|
if cfg.Ctrl.Options.NewListener != nil {
|
|
c.network.AddRouterPresenceHandler(&OnConnectSettingsHandler{
|
|
config: cfg,
|
|
settings: map[int32][]byte{
|
|
int32(ctrl_pb.SettingTypes_NewCtrlAddress): []byte((*cfg.Ctrl.Options.NewListener).String()),
|
|
},
|
|
})
|
|
}
|
|
|
|
if c.raftController != nil {
|
|
logrus.Info("Adding router presence handler to send out ctrl addresses")
|
|
c.network.AddRouterPresenceHandler(
|
|
NewOnConnectCtrlAddressesUpdateHandler(c.config.Ctrl.Listener.String(), c.raftController),
|
|
)
|
|
}
|
|
|
|
if err := c.showOptions(); err != nil {
|
|
return nil, err
|
|
}
|
|
|
|
return c, nil
|
|
}
|
|
|
|
func (c *Controller) InitTimelineId(timelineId string) {
|
|
c.env.InitTimelineId(timelineId)
|
|
}
|
|
|
|
func (c *Controller) TimelineId() string {
|
|
return c.env.TimelineId()
|
|
}
|
|
|
|
func (c *Controller) initWeb() {
|
|
healthChecker, err := c.initializeHealthChecks()
|
|
if err != nil {
|
|
logrus.WithError(err).Fatalf("failed to create health checker")
|
|
}
|
|
c.healthChecker = healthChecker
|
|
|
|
if err = c.xweb.GetRegistry().Add(webapis.NewControllerHealthCheckApiFactory(c.env, healthChecker)); err != nil {
|
|
logrus.WithError(err).Fatalf("failed to create health checks api factory")
|
|
}
|
|
|
|
fabricManagementFactory := webapis.NewFabricManagementApiFactory(c.config.Id, c.env, c.network, &c.xmgmts)
|
|
if err = c.xweb.GetRegistry().Add(fabricManagementFactory); err != nil {
|
|
logrus.WithError(err).Fatalf("failed to create management api factory")
|
|
}
|
|
|
|
if err = c.xweb.GetRegistry().Add(webapis.NewMetricsApiFactory(c.config.Id, c.network)); err != nil {
|
|
logrus.WithError(err).Fatalf("failed to create metrics api factory")
|
|
}
|
|
|
|
if err = c.xweb.GetRegistry().Add(webapis.NewSpaFactory()); err != nil {
|
|
logrus.WithError(err).Fatalf("failed to create single page application factory")
|
|
}
|
|
|
|
// Back-compat: preserve the legacy `binding: zac` so existing controller configs keep working.
|
|
// New configs should use `binding: spa` with an explicit `path`.
|
|
if err = c.xweb.GetRegistry().Add(webapis.NewZitiAdminConsoleFactory()); err != nil {
|
|
logrus.WithError(err).Fatalf("failed to create legacy ZAC factory")
|
|
}
|
|
|
|
if c.IsEdgeEnabled() {
|
|
managementApiFactory := webapis.NewManagementApiFactory(c.env)
|
|
clientApiFactory := webapis.NewClientApiFactory(c.env)
|
|
oidcApiFactory := webapis.NewOidcApiFactory(c.env)
|
|
|
|
if err = c.xweb.GetRegistry().Add(managementApiFactory); err != nil {
|
|
pfxlog.Logger().Fatalf("failed to create Edge Management API factory: %v", err)
|
|
}
|
|
|
|
if err = c.xweb.GetRegistry().Add(clientApiFactory); err != nil {
|
|
pfxlog.Logger().Fatalf("failed to create Edge Client API factory: %v", err)
|
|
}
|
|
|
|
if err = c.xweb.GetRegistry().Add(oidcApiFactory); err != nil {
|
|
pfxlog.Logger().Fatalf("failed to create OIDC API factory: %v", err)
|
|
}
|
|
} else {
|
|
// if no edge we need 1 default API, make the fabric api the default
|
|
fabricManagementFactory.MakeDefault = true
|
|
}
|
|
}
|
|
|
|
func (c *Controller) IsEdgeEnabled() bool {
|
|
return c.config.Edge.Enabled
|
|
}
|
|
|
|
// ensureOidcOnClientApiServer is an xweb InstanceValidator that automatically co-locates the edge OIDC
|
|
// API on the same web listener as the edge client API when the OIDC API is not explicitly configured
|
|
// anywhere. Selection prefers the web listener whose bind point address matches edge.api.address. If
|
|
// no bind point matches and there is only one edge-client web listener, OIDC is added there with a
|
|
// warning. If multiple edge-client web listeners exist and none match, OIDC is added to all of them
|
|
// with a warning.
|
|
func (c *Controller) ensureOidcOnClientApiServer(instanceConfig *xweb.InstanceConfig) error {
|
|
if !c.IsEdgeEnabled() {
|
|
return nil
|
|
}
|
|
|
|
if c.config.Edge.Api.DisableOidcAutoBinding {
|
|
return nil
|
|
}
|
|
|
|
for _, serverConfig := range instanceConfig.ServerConfigs {
|
|
for _, api := range serverConfig.APIs {
|
|
if api.Binding() == webapis.OidcApiBinding {
|
|
return nil
|
|
}
|
|
}
|
|
}
|
|
|
|
edgeApiAddress := c.config.Edge.Api.Address
|
|
|
|
addOidc := func(serverConfig *xweb.ServerConfig) error {
|
|
oidcApiConfig := &xweb.ApiConfig{}
|
|
if err := oidcApiConfig.Parse(map[interface{}]interface{}{
|
|
"binding": webapis.OidcApiBinding,
|
|
}); err != nil {
|
|
return fmt.Errorf("could not auto-configure edge OIDC API: %w", err)
|
|
}
|
|
serverConfig.APIs = append(serverConfig.APIs, oidcApiConfig)
|
|
return nil
|
|
}
|
|
|
|
var addressMatchedServers []*xweb.ServerConfig
|
|
var clientApiServers []*xweb.ServerConfig
|
|
|
|
for _, serverConfig := range instanceConfig.ServerConfigs {
|
|
hasClientApi := false
|
|
for _, api := range serverConfig.APIs {
|
|
if api.Binding() == webapis.ClientApiBinding {
|
|
hasClientApi = true
|
|
break
|
|
}
|
|
}
|
|
if !hasClientApi {
|
|
continue
|
|
}
|
|
clientApiServers = append(clientApiServers, serverConfig)
|
|
for _, bp := range serverConfig.BindPoints {
|
|
if bp.ServerAddress() == edgeApiAddress {
|
|
addressMatchedServers = append(addressMatchedServers, serverConfig)
|
|
break
|
|
}
|
|
}
|
|
}
|
|
|
|
if len(addressMatchedServers) > 0 {
|
|
for _, serverConfig := range addressMatchedServers {
|
|
if err := addOidc(serverConfig); err != nil {
|
|
return err
|
|
}
|
|
pfxlog.Logger().Infof("edge OIDC API not explicitly configured; automatically added to '%s' web listener", serverConfig.Name)
|
|
}
|
|
return nil
|
|
}
|
|
|
|
switch len(clientApiServers) {
|
|
case 0:
|
|
return nil
|
|
case 1:
|
|
if err := addOidc(clientApiServers[0]); err != nil {
|
|
return err
|
|
}
|
|
pfxlog.Logger().Warnf("edge OIDC API not explicitly configured; no edge-client bind point matched [edge.api.address] value [%s]; automatically added to '%s' web listener", edgeApiAddress, clientApiServers[0].Name)
|
|
default:
|
|
for _, serverConfig := range clientApiServers {
|
|
if err := addOidc(serverConfig); err != nil {
|
|
return err
|
|
}
|
|
}
|
|
var names []string
|
|
for _, s := range clientApiServers {
|
|
names = append(names, "'"+s.Name+"'")
|
|
}
|
|
pfxlog.Logger().Warnf("edge OIDC API not explicitly configured; no edge-client bind point matched [edge.api.address] value [%s]; automatically added to all edge-client web listeners: %s", edgeApiAddress, strings.Join(names, ", "))
|
|
}
|
|
|
|
return nil
|
|
}
|
|
|
|
func (c *Controller) Run() error {
|
|
c.startProfiling()
|
|
|
|
if err := c.registerComponents(); err != nil {
|
|
return fmt.Errorf("error registering component: %s", err)
|
|
}
|
|
versionInfo := c.network.VersionProvider.AsVersionInfo()
|
|
versionHeader, err := c.network.VersionProvider.EncoderDecoder().Encode(versionInfo)
|
|
|
|
if err != nil {
|
|
pfxlog.Logger().Panicf("could not prepare version headers: %v", err)
|
|
}
|
|
|
|
capabilityMask := capabilities.GetControllerCapabilitiesMask()
|
|
|
|
headers := map[int32][]byte{
|
|
channel.HelloVersionHeader: versionHeader,
|
|
int32(ctrl_pb.ControlHeaders_CapabilitiesHeader): capabilityMask.Bytes(),
|
|
ctrl_pb.LegacyCapabilitiesHeader: capabilityMask.Bytes(), // for pre-2.0 routers
|
|
}
|
|
|
|
/**
|
|
* ctrl listener/accepter.
|
|
*/
|
|
ctrlChannelListenerConfig := channel.ListenerConfig{
|
|
ConnectOptions: c.config.Ctrl.Options.ConnectOptions,
|
|
PoolConfigurator: servermetrics.GoroutinesPoolMetricsConfigF(c.network.GetMetricsRegistry(), "pool.listener.ctrl"),
|
|
Headers: headers,
|
|
TransportConfig: transport.Configuration{"protocol": "ziti-ctrl"},
|
|
}
|
|
|
|
if c.raftController != nil {
|
|
ctrlChannelListenerConfig.HeadersF = c.raftController.GetListenerHeaders
|
|
}
|
|
|
|
pfxlog.Logger().Infof("staring control channel listener on %s", c.config.Ctrl.Listener.String())
|
|
|
|
ctrlAccepter := handler_ctrl.NewCtrlAccepter(c.network, c.xctrls, c.config.Ctrl.Options.Options,
|
|
c.config.Ctrl.Options.RouterHeartbeatOptions, c.config.Trace.Handler, c.config.SlowHandlers)
|
|
|
|
ctrlAcceptors := map[string]channel.HelloAcceptor{}
|
|
if c.raftController != nil {
|
|
raftMesh := c.raftController.GetMesh()
|
|
c.eventDispatcher.AddClusterEventHandler(event.ClusterEventHandlerF(func(evt *event.ClusterEvent) {
|
|
for _, peer := range evt.Peers {
|
|
switch evt.EventType {
|
|
case event.ClusterPeerConnected:
|
|
c.network.GossipStore.PeerConnected(peer.Id)
|
|
case event.ClusterPeerDisconnected:
|
|
c.network.GossipStore.PeerDisconnected(peer.Id)
|
|
}
|
|
}
|
|
}))
|
|
c.raftController.ConfigureMeshHandlers(handler_peer_ctrl.NewBindHandler(c.network, c.raftController, c.network.GossipStore, c.config.Ctrl.Options.PeerHeartbeatOptions))
|
|
ctrlAcceptors[mesh.ChannelTypeMesh] = channel.AsHelloAcceptor(raftMesh)
|
|
}
|
|
|
|
// Channel types with a dedicated acceptor below validate their own peers; the connect handler
|
|
// must skip exactly those and validate everything else (which the routing acceptor sends to the
|
|
// default router control acceptor, including unrecognized types). Set this before the listener
|
|
// starts accepting connections.
|
|
separatelyValidatedTypes := map[string]struct{}{}
|
|
for chType := range ctrlAcceptors {
|
|
separatelyValidatedTypes[chType] = struct{}{}
|
|
}
|
|
c.ctrlConnectHandler.SetSeparatelyValidatedChannelTypes(separatelyValidatedTypes)
|
|
|
|
acceptor := &channel.TypeRoutingAcceptor{
|
|
Acceptors: ctrlAcceptors,
|
|
DefaultAcceptor: ctrlAccepter.NewMultiListener(),
|
|
}
|
|
|
|
ctrlChannelListenerConfig.ConnectionHandlers = append(ctrlChannelListenerConfig.ConnectionHandlers, c.ctrlConnectHandler)
|
|
|
|
ctrlListener, err := channel.NewClassicListenerWithAcceptor(c.config.Id, c.config.Ctrl.Listener, ctrlChannelListenerConfig, acceptor)
|
|
if err != nil {
|
|
panic(err)
|
|
}
|
|
c.ctrlListener = ctrlListener
|
|
|
|
if err = c.config.Configure(c.xweb); err != nil {
|
|
panic(err)
|
|
}
|
|
|
|
// Signal that xweb is fully configured. GetApiAddresses waits on this before reading the
|
|
// xweb config, so that mesh hellos triggered by early raft connections can't cache an empty
|
|
// address set (populated only once Configure has run) for the lifetime of the process.
|
|
c.xwebInitialized.MarkInitialized()
|
|
|
|
c.xweb.Run()
|
|
|
|
for _, helloHeader := range c.GetHelloHeaderProviders() {
|
|
helloHeader.Apply(headers)
|
|
}
|
|
|
|
// event handlers
|
|
if err := c.eventDispatcher.WireEventHandlers(c.getEventHandlerConfigs()); err != nil {
|
|
panic(err)
|
|
}
|
|
|
|
if c.raftController != nil {
|
|
c.raftController.StartEventGeneration()
|
|
}
|
|
|
|
if c.config.Ctrl.Dialer.Enabled {
|
|
ctrlDialer := handler_ctrl.NewCtrlDialer(
|
|
&c.config.Ctrl.Dialer,
|
|
c.network,
|
|
ctrlAccepter,
|
|
c.config.Id,
|
|
headers,
|
|
c.shutdownC,
|
|
c.metricsRegistry,
|
|
)
|
|
c.network.AddRouterPresenceHandler(ctrlDialer)
|
|
c.network.Router.Store.AddEntityIdListener(ctrlDialer.RouterUpdated, boltz.EntityUpdated)
|
|
c.network.Router.Store.AddEntityIdListener(ctrlDialer.RouterCreated, boltz.EntityCreated)
|
|
c.network.Router.Store.AddEntityIdListener(ctrlDialer.RouterDeleted, boltz.EntityDeleted)
|
|
c.network.AddInspectTarget(ctrlDialer.Inspect)
|
|
c.network.SetCtrlDialerValidator(ctrlDialer.Validate)
|
|
go ctrlDialer.Run()
|
|
}
|
|
|
|
c.network.Run()
|
|
|
|
return nil
|
|
}
|
|
|
|
func (c *Controller) getEventHandlerConfigs() []*events.EventHandlerConfig {
|
|
var result []*events.EventHandlerConfig
|
|
|
|
if e, ok := c.config.Src["events"]; ok {
|
|
if em, ok := e.(map[interface{}]interface{}); ok {
|
|
for id, v := range em {
|
|
if eventHandlerConfig, ok := v.(map[interface{}]interface{}); ok {
|
|
result = append(result, &events.EventHandlerConfig{
|
|
Id: id,
|
|
Config: eventHandlerConfig,
|
|
})
|
|
}
|
|
}
|
|
}
|
|
}
|
|
return result
|
|
}
|
|
|
|
func (c *Controller) GetCloseNotifyChannel() <-chan struct{} {
|
|
return c.shutdownC
|
|
}
|
|
|
|
func (c *Controller) Shutdown() {
|
|
if c.isShutdown.CompareAndSwap(false, true) {
|
|
close(c.shutdownC)
|
|
|
|
if c.ctrlListener != nil {
|
|
if err := c.ctrlListener.Close(); err != nil {
|
|
pfxlog.Logger().WithError(err).Error("failed to close ctrl channel listener")
|
|
}
|
|
}
|
|
|
|
if c.mgmtListener != nil {
|
|
if err := c.mgmtListener.Close(); err != nil {
|
|
pfxlog.Logger().WithError(err).Error("failed to close mgmt channel listener")
|
|
}
|
|
}
|
|
|
|
if c.config.Db != nil {
|
|
if err := c.config.Db.Close(); err != nil {
|
|
pfxlog.Logger().WithError(err).Error("failed to close db")
|
|
}
|
|
}
|
|
|
|
go c.xweb.Shutdown()
|
|
|
|
if c.raftController != nil {
|
|
if err := c.raftController.Shutdown(); err != nil {
|
|
pfxlog.Logger().WithError(err).Error("failed to shutdown raft")
|
|
}
|
|
}
|
|
|
|
c.config.Id.StopWatchingFiles()
|
|
if c.config.Edge.Enrollment.SigningCert != nil {
|
|
c.config.Edge.Enrollment.SigningCert.StopWatchingFiles()
|
|
}
|
|
|
|
if c.healthChecker != nil {
|
|
c.healthChecker.DeregisterAll()
|
|
}
|
|
}
|
|
}
|
|
|
|
func (c *Controller) showOptions() error {
|
|
if ctrl, err := json.MarshalIndent(c.config.Ctrl.Options, "", " "); err == nil {
|
|
pfxlog.Logger().Infof("ctrl = %s", string(ctrl))
|
|
} else {
|
|
return err
|
|
}
|
|
return nil
|
|
}
|
|
|
|
func (c *Controller) startProfiling() {
|
|
if c.config.Profile.Memory.Path != "" {
|
|
go profiler.NewMemoryWithShutdown(c.config.Profile.Memory.Path, c.config.Profile.Memory.Interval, c.shutdownC).Run()
|
|
}
|
|
if c.config.Profile.CPU.Path != "" {
|
|
if cpu, err := profiler.NewCPUWithShutdown(c.config.Profile.CPU.Path, c.shutdownC); err == nil {
|
|
go cpu.Run()
|
|
} else {
|
|
logrus.Errorf("unexpected error launching cpu profiling (%v)", err)
|
|
}
|
|
}
|
|
}
|
|
|
|
func (c *Controller) registerXts() {
|
|
xt.GlobalRegistry().RegisterFactory(xt_smartrouting.NewFactory())
|
|
xt.GlobalRegistry().RegisterFactory(xt_random.NewFactory())
|
|
xt.GlobalRegistry().RegisterFactory(xt_weighted.NewFactory())
|
|
xt.GlobalRegistry().RegisterFactory(xt_sticky.NewFactory())
|
|
}
|
|
|
|
func (c *Controller) registerComponents() error {
|
|
c.ctrlConnectHandler = handler_ctrl.NewConnectHandler(c.config.Id, c.network, c.signingCertRoots())
|
|
c.eventDispatcher.AddClusterEventHandler(event.ClusterEventHandlerF(c.routerDispatchCallback))
|
|
return nil
|
|
}
|
|
|
|
// signingCertRoots returns the trust anchors from the edge enrollment signing CA bundle, which issues the
|
|
// certificates routers present on the control channel. It is empty when no signing CA bundle is
|
|
// configured, in which case a router is trusted only through the controller's own CA bundle.
|
|
func (c *Controller) signingCertRoots() []*x509.Certificate {
|
|
if c.config.Edge == nil {
|
|
return nil
|
|
}
|
|
return nfpem.PemBytesToCertificates(c.config.Edge.Enrollment.SigningCertCaPem)
|
|
}
|
|
|
|
func (c *Controller) RegisterXctrl(x xctrl.Xctrl) error {
|
|
if err := c.config.Configure(x); err != nil {
|
|
return err
|
|
}
|
|
if x.Enabled() {
|
|
c.xctrls = append(c.xctrls, x)
|
|
if c.config.Trace.Handler != nil {
|
|
for _, decoder := range x.GetTraceDecoders() {
|
|
c.config.Trace.Handler.AddDecoder(decoder)
|
|
}
|
|
}
|
|
}
|
|
return nil
|
|
}
|
|
|
|
func (c *Controller) RegisterXmgmt(x xmgmt.Xmgmt) error {
|
|
if err := c.config.Configure(x); err != nil {
|
|
return err
|
|
}
|
|
pfxlog.Logger().Infof("adding xmgmt %T, enabled? %v", x, x.Enabled())
|
|
if x.Enabled() {
|
|
c.xmgmts.Append(x)
|
|
}
|
|
return nil
|
|
}
|
|
|
|
func (c *Controller) GetXWebInstance() xweb.Instance {
|
|
return c.xweb
|
|
}
|
|
|
|
func (c *Controller) GetNetwork() *network.Network {
|
|
return c.network
|
|
}
|
|
|
|
func (c *Controller) Identity() identity.Identity {
|
|
return c.config.Id
|
|
}
|
|
|
|
func (c *Controller) GetEventDispatcher() event.Dispatcher {
|
|
return c.eventDispatcher
|
|
}
|
|
|
|
func (c *Controller) routerDispatchCallback(evt *event.ClusterEvent) {
|
|
if evt.EventType == event.ClusterLeadershipGained {
|
|
req := &ctrl_pb.UpdateClusterLeader{
|
|
Index: evt.Index,
|
|
}
|
|
|
|
for _, r := range c.network.AllConnectedRouters() {
|
|
log := pfxlog.Logger().WithFields(map[string]interface{}{
|
|
"index": evt.Index,
|
|
})
|
|
|
|
if err := protobufs.MarshalTyped(req).Send(r.Control.GetDefaultSender()); err != nil {
|
|
pfxlog.Logger().WithError(err).WithField("routerId", r.Id).Error("unable to update cluster leader on router")
|
|
} else {
|
|
log.WithField("routerId", r.Id).WithField("routerName", r.Name).Info("router updated with info on new leader")
|
|
}
|
|
}
|
|
}
|
|
|
|
if evt.EventType == event.ClusterMembersChanged {
|
|
var endpoints []string
|
|
var controllers []*ctrl_pb.CtrlDetail
|
|
for _, peer := range evt.Peers {
|
|
endpoints = append(endpoints, peer.Addr)
|
|
controllers = append(controllers, &ctrl_pb.CtrlDetail{
|
|
Id: peer.Id,
|
|
Endpoints: []*ctrl_pb.CtrlEndpoint{{Address: peer.Addr}},
|
|
})
|
|
}
|
|
|
|
updMsg := &ctrl_pb.UpdateCtrlAddresses{
|
|
Addresses: endpoints,
|
|
IsLeader: c.raftController.IsLeader(),
|
|
Index: evt.Index,
|
|
Controllers: controllers,
|
|
}
|
|
|
|
log := pfxlog.Logger().WithFields(map[string]interface{}{
|
|
"addresses": endpoints,
|
|
"index": evt.Index,
|
|
})
|
|
|
|
log.Info("syncing updated ctrl addresses to connected routers")
|
|
|
|
for _, r := range c.network.AllConnectedRouters() {
|
|
if err := protobufs.MarshalTyped(updMsg).Send(r.Control.GetDefaultSender()); err != nil {
|
|
pfxlog.Logger().WithError(err).WithField("routerId", r.Id).Error("unable to update controller endpoints on router")
|
|
} else {
|
|
log.WithField("routerId", r.Id).WithField("routerName", r.Name).Info("router updated with latest ctrl addresses")
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
func (c *Controller) getMigrationDb() (*string, error) {
|
|
val, found := c.config.Src["db"]
|
|
if !found {
|
|
return nil, nil
|
|
}
|
|
|
|
path := fmt.Sprintf("%v", val)
|
|
if _, err := os.Stat(path); err != nil {
|
|
if os.IsNotExist(err) {
|
|
return nil, errors.Wrapf(err, "source db not found at [%v], either remove 'db' config setting or fix path ", path)
|
|
}
|
|
return nil, errors.Wrapf(err, "invalid db path [%v]", path)
|
|
}
|
|
|
|
return &path, nil
|
|
}
|
|
|
|
func (c *Controller) ValidateMigrationEnvironment() error {
|
|
_, err := c.getMigrationDb()
|
|
return err
|
|
}
|
|
|
|
func (c *Controller) TryInitializeRaftFromBoltDb() error {
|
|
path, err := c.getMigrationDb()
|
|
if err != nil || path == nil {
|
|
return err
|
|
}
|
|
return c.InitializeRaftFromBoltDb(*path)
|
|
}
|
|
|
|
func (c *Controller) InitializeRaftFromBoltDb(sourceDbPath string) error {
|
|
if c.raftController == nil {
|
|
return errors.New("can't initialize non-raft controller using initialize from db")
|
|
}
|
|
|
|
if c.raftController.IsBootstrapped() {
|
|
return errors.New("raft is already bootstrapped, must start with a uninitialized controller")
|
|
}
|
|
|
|
return c.RaftRestoreFromBoltDb(sourceDbPath)
|
|
}
|
|
|
|
// validateMigrationSourceDb rejects a migration source that is not an initialized controller db.
|
|
// db.Open creates the root bucket on any file, so existence is not enough; it checks for a default
|
|
// admin identity, which an initialized controller always has. The check is read-only.
|
|
func validateMigrationSourceDb(sourceDb boltz.Db) error {
|
|
hasDefaultAdmin := false
|
|
err := sourceDb.View(func(tx *bbolt.Tx) error {
|
|
identities := boltz.Path(tx, db.RootBucket, db.EntityTypeIdentities)
|
|
if identities == nil {
|
|
return nil
|
|
}
|
|
return identities.ForEachTypedBucket(func(_ string, identity *boltz.TypedBucket) error {
|
|
if identity.GetBoolWithDefault(db.FieldIdentityIsDefaultAdmin, false) {
|
|
hasDefaultAdmin = true
|
|
}
|
|
return nil
|
|
})
|
|
})
|
|
if err != nil {
|
|
return errors.Wrap(err, "unable to read identities from source db")
|
|
}
|
|
if !hasDefaultAdmin {
|
|
return errors.New("source db has no default admin identity; it is empty or was never a fully initialized controller")
|
|
}
|
|
return nil
|
|
}
|
|
|
|
func (c *Controller) RaftRestoreFromBoltDb(sourceDbPath string) error {
|
|
log := pfxlog.Logger()
|
|
|
|
if c.raftController == nil {
|
|
return errors.New("can't initialize non-raft controller using initialize from db")
|
|
}
|
|
|
|
if _, err := os.Stat(sourceDbPath); err != nil {
|
|
if os.IsNotExist(err) {
|
|
return errors.Wrapf(err, "source db not found at [%v]", sourceDbPath)
|
|
}
|
|
return errors.Wrapf(err, "invalid db path [%v]", sourceDbPath)
|
|
}
|
|
|
|
sourceDb, err := db.Open(sourceDbPath)
|
|
if err != nil {
|
|
return err
|
|
}
|
|
defer func() {
|
|
if err = sourceDb.Close(); err != nil {
|
|
log.WithError(err).Error("error closing migration source bolt db")
|
|
}
|
|
}()
|
|
|
|
if err = validateMigrationSourceDb(sourceDb); err != nil {
|
|
return errors.Wrapf(err, "migration source db [%v] is not a valid initialized controller database", sourceDbPath)
|
|
}
|
|
|
|
timelineId, err := sourceDb.GetTimelineId(boltz.TimelineModeForceReset, shortid.Generate)
|
|
if err != nil {
|
|
return err
|
|
}
|
|
log.WithField("timelineId", timelineId).WithField("path", sourceDbPath).Info("restoring from bolt db")
|
|
|
|
buf := &bytes.Buffer{}
|
|
gzWriter := gzip.NewWriter(buf)
|
|
if err = sourceDb.StreamToWriter(gzWriter); err != nil {
|
|
if closeErr := gzWriter.Close(); closeErr != nil {
|
|
log.WithError(closeErr).Error("error closing db snapshot buffer")
|
|
}
|
|
return err
|
|
}
|
|
|
|
if err = gzWriter.Close(); err != nil {
|
|
return errors.Wrap(err, "error finishing gz compression of migration snapshot")
|
|
}
|
|
|
|
cmd := &command.SyncSnapshotCommand{
|
|
TimelineId: timelineId,
|
|
Snapshot: buf.Bytes(),
|
|
SnapshotSink: c.network.RestoreSnapshot,
|
|
}
|
|
|
|
if err = c.raftController.Bootstrap(); err != nil {
|
|
return fmt.Errorf("unable to bootstrap cluster (%w)", err)
|
|
}
|
|
|
|
// Carry the cluster id Bootstrap established so RestoreSnapshot can write it back after the
|
|
// restore (the migration source has none). Blank here means a bug in Bootstrap, so fail.
|
|
cmd.ClusterId = c.raftController.GetClusterId()
|
|
if cmd.ClusterId == "" {
|
|
return errors.New("cluster id is blank after bootstrap; refusing to restore without a durable cluster id")
|
|
}
|
|
|
|
return c.raftController.Dispatch(cmd)
|
|
}
|
|
|
|
// TODO: this functions is a temporary hack and should be provided by xweb
|
|
func getApiPath(binding string) string {
|
|
switch binding {
|
|
case "edge-client":
|
|
return "/edge/client/v1"
|
|
case "edge-management":
|
|
return "/edge/management/v1"
|
|
case "fabric":
|
|
return "/fabric/v1"
|
|
case "health-checks":
|
|
return "/health-checks"
|
|
case "edge-oidc":
|
|
return "/oidc"
|
|
}
|
|
|
|
return ""
|
|
}
|
|
|
|
func (c *Controller) GetApiAddresses() (map[string][]event.ApiAddress, []byte) {
|
|
c.apiDataOnce.Do(func() {
|
|
c.xwebInitialized.WaitTillInitialized()
|
|
xwebConfig := c.xweb.GetConfig()
|
|
|
|
apiData := map[string][]event.ApiAddress{}
|
|
for _, serverConfig := range xwebConfig.ServerConfigs {
|
|
for _, bindPoint := range serverConfig.BindPoints {
|
|
if bindPoint.Type() != bindpoints.BindPointTypeUnderlay {
|
|
continue
|
|
}
|
|
for _, api := range serverConfig.APIs {
|
|
apiData[api.Binding()] = append(apiData[api.Binding()], event.ApiAddress{
|
|
Url: "https://" + bindPoint.ServerAddress() + getApiPath(api.Binding()), //TODO: temp till xweb support reporting API paths
|
|
Version: "v1", //TODO: temp till xweb supports reporting versions via api.Version()
|
|
})
|
|
}
|
|
}
|
|
}
|
|
|
|
c.apiData = apiData
|
|
c.apiDataBytes, _ = json.Marshal(apiData)
|
|
})
|
|
|
|
return c.apiData, c.apiDataBytes
|
|
}
|
|
|
|
func (c *Controller) GetHelloHeaderProviders() []mesh.HeaderProvider {
|
|
apiAddrProvider := mesh.HeaderProviderFunc(func(headers map[int32][]byte) {
|
|
_, apiDataBytes := c.GetApiAddresses()
|
|
headers[mesh.ApiAddressesHeader] = apiDataBytes
|
|
headers[mesh.LegacyApiAddressesHeader] = apiDataBytes
|
|
})
|
|
|
|
preferredLeaderProvider := mesh.HeaderProviderFunc(func(headers map[int32][]byte) {
|
|
if c.config.Raft != nil && c.config.Raft.PreferredLeader {
|
|
headers[mesh.PreferredLeaderHeader] = []byte{1}
|
|
}
|
|
})
|
|
|
|
return []mesh.HeaderProvider{apiAddrProvider, preferredLeaderProvider}
|
|
}
|