moby--moby/cmd/dockerd/daemon.go

450 lines
13 KiB
Go
Raw Normal View History

package main
import (
"crypto/tls"
"fmt"
"io"
"os"
"path/filepath"
"runtime"
"strings"
"time"
"github.com/Sirupsen/logrus"
"github.com/docker/distribution/uuid"
"github.com/docker/docker/api"
apiserver "github.com/docker/docker/api/server"
"github.com/docker/docker/api/server/middleware"
Remove static errors from errors package. Moving all strings to the errors package wasn't a good idea after all. Our custom implementation of Go errors predates everything that's nice and good about working with errors in Go. Take as an example what we have to do to get an error message: ```go func GetErrorMessage(err error) string { switch err.(type) { case errcode.Error: e, _ := err.(errcode.Error) return e.Message case errcode.ErrorCode: ec, _ := err.(errcode.ErrorCode) return ec.Message() default: return err.Error() } } ``` This goes against every good practice for Go development. The language already provides a simple, intuitive and standard way to get error messages, that is calling the `Error()` method from an error. Reinventing the error interface is a mistake. Our custom implementation also makes very hard to reason about errors, another nice thing about Go. I found several (>10) error declarations that we don't use anywhere. This is a clear sign about how little we know about the errors we return. I also found several error usages where the number of arguments was different than the parameters declared in the error, another clear example of how difficult is to reason about errors. Moreover, our custom implementation didn't really make easier for people to return custom HTTP status code depending on the errors. Again, it's hard to reason about when to set custom codes and how. Take an example what we have to do to extract the message and status code from an error before returning a response from the API: ```go switch err.(type) { case errcode.ErrorCode: daError, _ := err.(errcode.ErrorCode) statusCode = daError.Descriptor().HTTPStatusCode errMsg = daError.Message() case errcode.Error: // For reference, if you're looking for a particular error // then you can do something like : // import ( derr "github.com/docker/docker/errors" ) // if daError.ErrorCode() == derr.ErrorCodeNoSuchContainer { ... } daError, _ := err.(errcode.Error) statusCode = daError.ErrorCode().Descriptor().HTTPStatusCode errMsg = daError.Message default: // This part of will be removed once we've // converted everything over to use the errcode package // FIXME: this is brittle and should not be necessary. // If we need to differentiate between different possible error types, // we should create appropriate error types with clearly defined meaning errStr := strings.ToLower(err.Error()) for keyword, status := range map[string]int{ "not found": http.StatusNotFound, "no such": http.StatusNotFound, "bad parameter": http.StatusBadRequest, "conflict": http.StatusConflict, "impossible": http.StatusNotAcceptable, "wrong login/password": http.StatusUnauthorized, "hasn't been activated": http.StatusForbidden, } { if strings.Contains(errStr, keyword) { statusCode = status break } } } ``` You can notice two things in that code: 1. We have to explain how errors work, because our implementation goes against how easy to use Go errors are. 2. At no moment we arrived to remove that `switch` statement that was the original reason to use our custom implementation. This change removes all our status errors from the errors package and puts them back in their specific contexts. IT puts the messages back with their contexts. That way, we know right away when errors used and how to generate their messages. It uses custom interfaces to reason about errors. Errors that need to response with a custom status code MUST implementent this simple interface: ```go type errorWithStatus interface { HTTPErrorStatusCode() int } ``` This interface is very straightforward to implement. It also preserves Go errors real behavior, getting the message is as simple as using the `Error()` method. I included helper functions to generate errors that use custom status code in `errors/errors.go`. By doing this, we remove the hard dependency we have eeverywhere to our custom errors package. Yes, you can use it as a helper to generate error, but it's still very easy to generate errors without it. Please, read this fantastic blog post about errors in Go: http://dave.cheney.net/2014/12/24/inspecting-errors Signed-off-by: David Calavera <david.calavera@gmail.com>
2016-02-25 15:53:35 +00:00
"github.com/docker/docker/api/server/router"
"github.com/docker/docker/api/server/router/build"
"github.com/docker/docker/api/server/router/container"
"github.com/docker/docker/api/server/router/image"
"github.com/docker/docker/api/server/router/network"
swarmrouter "github.com/docker/docker/api/server/router/swarm"
systemrouter "github.com/docker/docker/api/server/router/system"
"github.com/docker/docker/api/server/router/volume"
"github.com/docker/docker/builder/dockerfile"
cliflags "github.com/docker/docker/cli/flags"
"github.com/docker/docker/cliconfig"
"github.com/docker/docker/daemon"
"github.com/docker/docker/daemon/cluster"
"github.com/docker/docker/daemon/logger"
"github.com/docker/docker/dockerversion"
"github.com/docker/docker/libcontainerd"
"github.com/docker/docker/opts"
"github.com/docker/docker/pkg/authorization"
"github.com/docker/docker/pkg/jsonlog"
"github.com/docker/docker/pkg/listeners"
flag "github.com/docker/docker/pkg/mflag"
"github.com/docker/docker/pkg/pidfile"
"github.com/docker/docker/pkg/signal"
"github.com/docker/docker/pkg/system"
"github.com/docker/docker/registry"
"github.com/docker/docker/runconfig"
"github.com/docker/docker/utils"
"github.com/docker/go-connections/tlsconfig"
)
const (
daemonConfigFileFlag = "-config-file"
)
// DaemonCli represents the daemon CLI.
type DaemonCli struct {
*daemon.Config
commonFlags *cliflags.CommonFlags
configFile *string
api *apiserver.Server
d *daemon.Daemon
authzMiddleware *authorization.Middleware // authzMiddleware enables to dynamically reload the authorization plugins
}
func presentInHelp(usage string) string { return usage }
func absentFromHelp(string) string { return "" }
// NewDaemonCli returns a pre-configured daemon CLI
func NewDaemonCli() *DaemonCli {
// TODO(tiborvass): remove InstallFlags?
daemonConfig := new(daemon.Config)
daemonConfig.LogConfig.Config = make(map[string]string)
daemonConfig.ClusterOpts = make(map[string]string)
daemonConfig.InstallFlags(flag.CommandLine, presentInHelp)
configFile := flag.CommandLine.String([]string{daemonConfigFileFlag}, defaultDaemonConfigFile, "Daemon configuration file")
flag.CommandLine.Require(flag.Exact, 0)
if runtime.GOOS != "linux" {
daemonConfig.V2Only = true
}
return &DaemonCli{
Config: daemonConfig,
commonFlags: cliflags.InitCommonFlags(),
configFile: configFile,
}
}
func migrateKey() (err error) {
// Migrate trust key if exists at ~/.docker/key.json and owned by current user
oldPath := filepath.Join(cliconfig.ConfigDir(), cliflags.DefaultTrustKeyFile)
newPath := filepath.Join(getDaemonConfDir(), cliflags.DefaultTrustKeyFile)
if _, statErr := os.Stat(newPath); os.IsNotExist(statErr) && currentUserIsOwner(oldPath) {
defer func() {
// Ensure old path is removed if no error occurred
if err == nil {
err = os.Remove(oldPath)
} else {
logrus.Warnf("Key migration failed, key file not removed at %s", oldPath)
os.Remove(newPath)
}
}()
if err := system.MkdirAll(getDaemonConfDir(), os.FileMode(0644)); err != nil {
return fmt.Errorf("Unable to create daemon configuration directory: %s", err)
}
newFile, err := os.OpenFile(newPath, os.O_RDWR|os.O_CREATE|os.O_TRUNC, 0600)
if err != nil {
return fmt.Errorf("error creating key file %q: %s", newPath, err)
}
defer newFile.Close()
oldFile, err := os.Open(oldPath)
if err != nil {
return fmt.Errorf("error opening key file %q: %s", oldPath, err)
}
defer oldFile.Close()
if _, err := io.Copy(newFile, oldFile); err != nil {
return fmt.Errorf("error copying key: %s", err)
}
logrus.Infof("Migrated key from %s to %s", oldPath, newPath)
}
return nil
}
func (cli *DaemonCli) start() (err error) {
stopc := make(chan bool)
defer close(stopc)
// warn from uuid package when running the daemon
uuid.Loggerf = logrus.Warnf
flags := flag.CommandLine
cli.commonFlags.PostParse()
if cli.commonFlags.TrustKey == "" {
cli.commonFlags.TrustKey = filepath.Join(getDaemonConfDir(), cliflags.DefaultTrustKeyFile)
}
cliConfig, err := loadDaemonCliConfig(cli.Config, flags, cli.commonFlags, *cli.configFile)
if err != nil {
return err
}
cli.Config = cliConfig
if cli.Config.Debug {
utils.EnableDebug()
}
if utils.ExperimentalBuild() {
logrus.Warn("Running experimental build")
}
logrus.SetFormatter(&logrus.TextFormatter{
TimestampFormat: jsonlog.RFC3339NanoFixed,
DisableColors: cli.Config.RawLogs,
})
if err := setDefaultUmask(); err != nil {
return fmt.Errorf("Failed to set umask: %v", err)
}
if len(cli.LogConfig.Config) > 0 {
if err := logger.ValidateLogOpts(cli.LogConfig.Type, cli.LogConfig.Config); err != nil {
return fmt.Errorf("Failed to set log opts: %v", err)
}
}
if cli.Pidfile != "" {
pf, err := pidfile.New(cli.Pidfile)
if err != nil {
return fmt.Errorf("Error starting daemon: %v", err)
}
defer func() {
if err := pf.Remove(); err != nil {
logrus.Error(err)
}
}()
}
serverConfig := &apiserver.Config{
Logging: true,
SocketGroup: cli.Config.SocketGroup,
Version: dockerversion.Version,
EnableCors: cli.Config.EnableCors,
CorsHeaders: cli.Config.CorsHeaders,
}
if cli.Config.TLS {
tlsOptions := tlsconfig.Options{
CAFile: cli.Config.CommonTLSOptions.CAFile,
CertFile: cli.Config.CommonTLSOptions.CertFile,
KeyFile: cli.Config.CommonTLSOptions.KeyFile,
}
if cli.Config.TLSVerify {
// server requires and verifies client's certificate
tlsOptions.ClientAuth = tls.RequireAndVerifyClientCert
}
tlsConfig, err := tlsconfig.Server(tlsOptions)
if err != nil {
return err
}
serverConfig.TLSConfig = tlsConfig
}
if len(cli.Config.Hosts) == 0 {
cli.Config.Hosts = make([]string, 1)
}
api := apiserver.New(serverConfig)
cli.api = api
for i := 0; i < len(cli.Config.Hosts); i++ {
var err error
if cli.Config.Hosts[i], err = opts.ParseHost(cli.Config.TLS, cli.Config.Hosts[i]); err != nil {
return fmt.Errorf("error parsing -H %s : %v", cli.Config.Hosts[i], err)
}
protoAddr := cli.Config.Hosts[i]
protoAddrParts := strings.SplitN(protoAddr, "://", 2)
if len(protoAddrParts) != 2 {
return fmt.Errorf("bad format %s, expected PROTO://ADDR", protoAddr)
}
proto := protoAddrParts[0]
addr := protoAddrParts[1]
// It's a bad idea to bind to TCP without tlsverify.
if proto == "tcp" && (serverConfig.TLSConfig == nil || serverConfig.TLSConfig.ClientAuth != tls.RequireAndVerifyClientCert) {
logrus.Warn("[!] DON'T BIND ON ANY IP ADDRESS WITHOUT setting -tlsverify IF YOU DON'T KNOW WHAT YOU'RE DOING [!]")
}
ls, err := listeners.Init(proto, addr, serverConfig.SocketGroup, serverConfig.TLSConfig)
if err != nil {
return err
}
ls = wrapListeners(proto, ls)
// If we're binding to a TCP port, make sure that a container doesn't try to use it.
if proto == "tcp" {
if err := allocateDaemonPort(addr); err != nil {
return err
}
}
logrus.Debugf("Listener created for HTTP on %s (%s)", proto, addr)
api.Accept(addr, ls...)
}
if err := migrateKey(); err != nil {
return err
}
cli.TrustKeyPath = cli.commonFlags.TrustKey
registryService := registry.NewService(cli.Config.ServiceOptions)
containerdRemote, err := libcontainerd.New(cli.getLibcontainerdRoot(), cli.getPlatformRemoteOptions()...)
if err != nil {
return err
}
signal.Trap(func() {
cli.stop()
<-stopc // wait for daemonCli.start() to return
})
d, err := daemon.NewDaemon(cli.Config, registryService, containerdRemote)
if err != nil {
return fmt.Errorf("Error starting daemon: %v", err)
}
name, _ := os.Hostname()
c, err := cluster.New(cluster.Config{
Split advertised address from listen address There are currently problems with "swarm init" and "swarm join" when an explicit --listen-addr flag is not provided. swarmkit defaults to finding the IP address associated with the default route, and in cloud setups this is often the wrong choice. Introduce a notion of "advertised address", with the client flag --advertise-addr, and the daemon flag --swarm-default-advertise-addr to provide a default. The default listening address is now 0.0.0.0, but a valid advertised address must be detected or specified. If no explicit advertised address is specified, error out if there is more than one usable candidate IP address on the system. This requires a user to explicitly choose instead of letting swarmkit make the wrong choice. For the purposes of this autodetection, we ignore certain interfaces that are unlikely to be relevant (currently docker*). The user is also required to choose a listen address on swarm init if they specify an explicit advertise address that is a hostname or an IP address that's not local to the system. This is a requirement for overlay networking. Also support specifying interface names to --listen-addr, --advertise-addr, and the daemon flag --swarm-default-advertise-addr. This will fail if the interface has multiple IP addresses (unless it has a single IPv4 address and a single IPv6 address - then we resolve the tie in favor of IPv4). This change also exposes the node's externally-reachable address in docker info, as requested by #24017. Make corresponding API and CLI docs changes. Signed-off-by: Aaron Lehmann <aaron.lehmann@docker.com>
2016-07-01 01:07:35 +00:00
Root: cli.Config.Root,
Name: name,
Backend: d,
NetworkSubnetsProvider: d,
DefaultAdvertiseAddr: cli.Config.SwarmDefaultAdvertiseAddr,
})
if err != nil {
logrus.Fatalf("Error creating cluster component: %v", err)
}
logrus.Info("Daemon has completed initialization")
logrus.WithFields(logrus.Fields{
"version": dockerversion.Version,
"commit": dockerversion.GitCommit,
"graphdriver": d.GraphDriverName(),
}).Info("Docker daemon")
cli.initMiddlewares(api, serverConfig)
initRouter(api, d, c)
cli.d = d
cli.setupConfigReloadTrap()
// The serve API routine never exits unless an error occurs
// We need to start it as a goroutine and wait on it so
// daemon doesn't exit
serveAPIWait := make(chan error)
go api.Wait(serveAPIWait)
// after the daemon is done setting up we can notify systemd api
notifySystem()
// Daemon is fully initialized and handling API traffic
// Wait for serve API to complete
errAPI := <-serveAPIWait
c.Cleanup()
shutdownDaemon(d, 15)
containerdRemote.Cleanup()
if errAPI != nil {
return fmt.Errorf("Shutting down due to ServeAPI error: %v", errAPI)
}
return nil
}
func (cli *DaemonCli) reloadConfig() {
reload := func(config *daemon.Config) {
// Reload the authorization plugin
cli.authzMiddleware.SetPlugins(config.AuthorizationPlugins)
if err := cli.d.Reload(config); err != nil {
logrus.Errorf("Error reconfiguring the daemon: %v", err)
return
}
if config.IsValueSet("debug") {
debugEnabled := utils.IsDebugEnabled()
switch {
case debugEnabled && !config.Debug: // disable debug
utils.DisableDebug()
cli.api.DisableProfiler()
case config.Debug && !debugEnabled: // enable debug
utils.EnableDebug()
cli.api.EnableProfiler()
}
}
}
if err := daemon.ReloadConfiguration(*cli.configFile, flag.CommandLine, reload); err != nil {
logrus.Error(err)
}
}
func (cli *DaemonCli) stop() {
cli.api.Close()
}
// shutdownDaemon just wraps daemon.Shutdown() to handle a timeout in case
// d.Shutdown() is waiting too long to kill container or worst it's
// blocked there
func shutdownDaemon(d *daemon.Daemon, timeout time.Duration) {
ch := make(chan struct{})
go func() {
d.Shutdown()
close(ch)
}()
select {
case <-ch:
logrus.Debug("Clean shutdown succeeded")
case <-time.After(timeout * time.Second):
logrus.Error("Force shutdown daemon")
}
}
func loadDaemonCliConfig(config *daemon.Config, flags *flag.FlagSet, commonConfig *cliflags.CommonFlags, configFile string) (*daemon.Config, error) {
config.Debug = commonConfig.Debug
config.Hosts = commonConfig.Hosts
config.LogLevel = commonConfig.LogLevel
config.TLS = commonConfig.TLS
config.TLSVerify = commonConfig.TLSVerify
config.CommonTLSOptions = daemon.CommonTLSOptions{}
if commonConfig.TLSOptions != nil {
config.CommonTLSOptions.CAFile = commonConfig.TLSOptions.CAFile
config.CommonTLSOptions.CertFile = commonConfig.TLSOptions.CertFile
config.CommonTLSOptions.KeyFile = commonConfig.TLSOptions.KeyFile
}
if configFile != "" {
c, err := daemon.MergeDaemonConfigurations(config, flags, configFile)
if err != nil {
if flags.IsSet(daemonConfigFileFlag) || !os.IsNotExist(err) {
return nil, fmt.Errorf("unable to configure the Docker daemon with file %s: %v\n", configFile, err)
}
}
// the merged configuration can be nil if the config file didn't exist.
// leave the current configuration as it is if when that happens.
if c != nil {
config = c
}
}
if err := daemon.ValidateConfiguration(config); err != nil {
return nil, err
}
// Regardless of whether the user sets it to true or false, if they
// specify TLSVerify at all then we need to turn on TLS
if config.IsValueSet(cliflags.TLSVerifyKey) {
config.TLS = true
}
// ensure that the log level is the one set after merging configurations
cliflags.SetDaemonLogLevel(config.LogLevel)
return config, nil
}
func initRouter(s *apiserver.Server, d *daemon.Daemon, c *cluster.Cluster) {
decoder := runconfig.ContainerDecoder{}
Remove static errors from errors package. Moving all strings to the errors package wasn't a good idea after all. Our custom implementation of Go errors predates everything that's nice and good about working with errors in Go. Take as an example what we have to do to get an error message: ```go func GetErrorMessage(err error) string { switch err.(type) { case errcode.Error: e, _ := err.(errcode.Error) return e.Message case errcode.ErrorCode: ec, _ := err.(errcode.ErrorCode) return ec.Message() default: return err.Error() } } ``` This goes against every good practice for Go development. The language already provides a simple, intuitive and standard way to get error messages, that is calling the `Error()` method from an error. Reinventing the error interface is a mistake. Our custom implementation also makes very hard to reason about errors, another nice thing about Go. I found several (>10) error declarations that we don't use anywhere. This is a clear sign about how little we know about the errors we return. I also found several error usages where the number of arguments was different than the parameters declared in the error, another clear example of how difficult is to reason about errors. Moreover, our custom implementation didn't really make easier for people to return custom HTTP status code depending on the errors. Again, it's hard to reason about when to set custom codes and how. Take an example what we have to do to extract the message and status code from an error before returning a response from the API: ```go switch err.(type) { case errcode.ErrorCode: daError, _ := err.(errcode.ErrorCode) statusCode = daError.Descriptor().HTTPStatusCode errMsg = daError.Message() case errcode.Error: // For reference, if you're looking for a particular error // then you can do something like : // import ( derr "github.com/docker/docker/errors" ) // if daError.ErrorCode() == derr.ErrorCodeNoSuchContainer { ... } daError, _ := err.(errcode.Error) statusCode = daError.ErrorCode().Descriptor().HTTPStatusCode errMsg = daError.Message default: // This part of will be removed once we've // converted everything over to use the errcode package // FIXME: this is brittle and should not be necessary. // If we need to differentiate between different possible error types, // we should create appropriate error types with clearly defined meaning errStr := strings.ToLower(err.Error()) for keyword, status := range map[string]int{ "not found": http.StatusNotFound, "no such": http.StatusNotFound, "bad parameter": http.StatusBadRequest, "conflict": http.StatusConflict, "impossible": http.StatusNotAcceptable, "wrong login/password": http.StatusUnauthorized, "hasn't been activated": http.StatusForbidden, } { if strings.Contains(errStr, keyword) { statusCode = status break } } } ``` You can notice two things in that code: 1. We have to explain how errors work, because our implementation goes against how easy to use Go errors are. 2. At no moment we arrived to remove that `switch` statement that was the original reason to use our custom implementation. This change removes all our status errors from the errors package and puts them back in their specific contexts. IT puts the messages back with their contexts. That way, we know right away when errors used and how to generate their messages. It uses custom interfaces to reason about errors. Errors that need to response with a custom status code MUST implementent this simple interface: ```go type errorWithStatus interface { HTTPErrorStatusCode() int } ``` This interface is very straightforward to implement. It also preserves Go errors real behavior, getting the message is as simple as using the `Error()` method. I included helper functions to generate errors that use custom status code in `errors/errors.go`. By doing this, we remove the hard dependency we have eeverywhere to our custom errors package. Yes, you can use it as a helper to generate error, but it's still very easy to generate errors without it. Please, read this fantastic blog post about errors in Go: http://dave.cheney.net/2014/12/24/inspecting-errors Signed-off-by: David Calavera <david.calavera@gmail.com>
2016-02-25 15:53:35 +00:00
routers := []router.Router{
container.NewRouter(d, decoder),
image.NewRouter(d, decoder),
systemrouter.NewRouter(d, c),
volume.NewRouter(d),
build.NewRouter(dockerfile.NewBuildManager(d)),
swarmrouter.NewRouter(c),
Remove static errors from errors package. Moving all strings to the errors package wasn't a good idea after all. Our custom implementation of Go errors predates everything that's nice and good about working with errors in Go. Take as an example what we have to do to get an error message: ```go func GetErrorMessage(err error) string { switch err.(type) { case errcode.Error: e, _ := err.(errcode.Error) return e.Message case errcode.ErrorCode: ec, _ := err.(errcode.ErrorCode) return ec.Message() default: return err.Error() } } ``` This goes against every good practice for Go development. The language already provides a simple, intuitive and standard way to get error messages, that is calling the `Error()` method from an error. Reinventing the error interface is a mistake. Our custom implementation also makes very hard to reason about errors, another nice thing about Go. I found several (>10) error declarations that we don't use anywhere. This is a clear sign about how little we know about the errors we return. I also found several error usages where the number of arguments was different than the parameters declared in the error, another clear example of how difficult is to reason about errors. Moreover, our custom implementation didn't really make easier for people to return custom HTTP status code depending on the errors. Again, it's hard to reason about when to set custom codes and how. Take an example what we have to do to extract the message and status code from an error before returning a response from the API: ```go switch err.(type) { case errcode.ErrorCode: daError, _ := err.(errcode.ErrorCode) statusCode = daError.Descriptor().HTTPStatusCode errMsg = daError.Message() case errcode.Error: // For reference, if you're looking for a particular error // then you can do something like : // import ( derr "github.com/docker/docker/errors" ) // if daError.ErrorCode() == derr.ErrorCodeNoSuchContainer { ... } daError, _ := err.(errcode.Error) statusCode = daError.ErrorCode().Descriptor().HTTPStatusCode errMsg = daError.Message default: // This part of will be removed once we've // converted everything over to use the errcode package // FIXME: this is brittle and should not be necessary. // If we need to differentiate between different possible error types, // we should create appropriate error types with clearly defined meaning errStr := strings.ToLower(err.Error()) for keyword, status := range map[string]int{ "not found": http.StatusNotFound, "no such": http.StatusNotFound, "bad parameter": http.StatusBadRequest, "conflict": http.StatusConflict, "impossible": http.StatusNotAcceptable, "wrong login/password": http.StatusUnauthorized, "hasn't been activated": http.StatusForbidden, } { if strings.Contains(errStr, keyword) { statusCode = status break } } } ``` You can notice two things in that code: 1. We have to explain how errors work, because our implementation goes against how easy to use Go errors are. 2. At no moment we arrived to remove that `switch` statement that was the original reason to use our custom implementation. This change removes all our status errors from the errors package and puts them back in their specific contexts. IT puts the messages back with their contexts. That way, we know right away when errors used and how to generate their messages. It uses custom interfaces to reason about errors. Errors that need to response with a custom status code MUST implementent this simple interface: ```go type errorWithStatus interface { HTTPErrorStatusCode() int } ``` This interface is very straightforward to implement. It also preserves Go errors real behavior, getting the message is as simple as using the `Error()` method. I included helper functions to generate errors that use custom status code in `errors/errors.go`. By doing this, we remove the hard dependency we have eeverywhere to our custom errors package. Yes, you can use it as a helper to generate error, but it's still very easy to generate errors without it. Please, read this fantastic blog post about errors in Go: http://dave.cheney.net/2014/12/24/inspecting-errors Signed-off-by: David Calavera <david.calavera@gmail.com>
2016-02-25 15:53:35 +00:00
}
if d.NetworkControllerEnabled() {
routers = append(routers, network.NewRouter(d, c))
Remove static errors from errors package. Moving all strings to the errors package wasn't a good idea after all. Our custom implementation of Go errors predates everything that's nice and good about working with errors in Go. Take as an example what we have to do to get an error message: ```go func GetErrorMessage(err error) string { switch err.(type) { case errcode.Error: e, _ := err.(errcode.Error) return e.Message case errcode.ErrorCode: ec, _ := err.(errcode.ErrorCode) return ec.Message() default: return err.Error() } } ``` This goes against every good practice for Go development. The language already provides a simple, intuitive and standard way to get error messages, that is calling the `Error()` method from an error. Reinventing the error interface is a mistake. Our custom implementation also makes very hard to reason about errors, another nice thing about Go. I found several (>10) error declarations that we don't use anywhere. This is a clear sign about how little we know about the errors we return. I also found several error usages where the number of arguments was different than the parameters declared in the error, another clear example of how difficult is to reason about errors. Moreover, our custom implementation didn't really make easier for people to return custom HTTP status code depending on the errors. Again, it's hard to reason about when to set custom codes and how. Take an example what we have to do to extract the message and status code from an error before returning a response from the API: ```go switch err.(type) { case errcode.ErrorCode: daError, _ := err.(errcode.ErrorCode) statusCode = daError.Descriptor().HTTPStatusCode errMsg = daError.Message() case errcode.Error: // For reference, if you're looking for a particular error // then you can do something like : // import ( derr "github.com/docker/docker/errors" ) // if daError.ErrorCode() == derr.ErrorCodeNoSuchContainer { ... } daError, _ := err.(errcode.Error) statusCode = daError.ErrorCode().Descriptor().HTTPStatusCode errMsg = daError.Message default: // This part of will be removed once we've // converted everything over to use the errcode package // FIXME: this is brittle and should not be necessary. // If we need to differentiate between different possible error types, // we should create appropriate error types with clearly defined meaning errStr := strings.ToLower(err.Error()) for keyword, status := range map[string]int{ "not found": http.StatusNotFound, "no such": http.StatusNotFound, "bad parameter": http.StatusBadRequest, "conflict": http.StatusConflict, "impossible": http.StatusNotAcceptable, "wrong login/password": http.StatusUnauthorized, "hasn't been activated": http.StatusForbidden, } { if strings.Contains(errStr, keyword) { statusCode = status break } } } ``` You can notice two things in that code: 1. We have to explain how errors work, because our implementation goes against how easy to use Go errors are. 2. At no moment we arrived to remove that `switch` statement that was the original reason to use our custom implementation. This change removes all our status errors from the errors package and puts them back in their specific contexts. IT puts the messages back with their contexts. That way, we know right away when errors used and how to generate their messages. It uses custom interfaces to reason about errors. Errors that need to response with a custom status code MUST implementent this simple interface: ```go type errorWithStatus interface { HTTPErrorStatusCode() int } ``` This interface is very straightforward to implement. It also preserves Go errors real behavior, getting the message is as simple as using the `Error()` method. I included helper functions to generate errors that use custom status code in `errors/errors.go`. By doing this, we remove the hard dependency we have eeverywhere to our custom errors package. Yes, you can use it as a helper to generate error, but it's still very easy to generate errors without it. Please, read this fantastic blog post about errors in Go: http://dave.cheney.net/2014/12/24/inspecting-errors Signed-off-by: David Calavera <david.calavera@gmail.com>
2016-02-25 15:53:35 +00:00
}
routers = addExperimentalRouters(routers)
Remove static errors from errors package. Moving all strings to the errors package wasn't a good idea after all. Our custom implementation of Go errors predates everything that's nice and good about working with errors in Go. Take as an example what we have to do to get an error message: ```go func GetErrorMessage(err error) string { switch err.(type) { case errcode.Error: e, _ := err.(errcode.Error) return e.Message case errcode.ErrorCode: ec, _ := err.(errcode.ErrorCode) return ec.Message() default: return err.Error() } } ``` This goes against every good practice for Go development. The language already provides a simple, intuitive and standard way to get error messages, that is calling the `Error()` method from an error. Reinventing the error interface is a mistake. Our custom implementation also makes very hard to reason about errors, another nice thing about Go. I found several (>10) error declarations that we don't use anywhere. This is a clear sign about how little we know about the errors we return. I also found several error usages where the number of arguments was different than the parameters declared in the error, another clear example of how difficult is to reason about errors. Moreover, our custom implementation didn't really make easier for people to return custom HTTP status code depending on the errors. Again, it's hard to reason about when to set custom codes and how. Take an example what we have to do to extract the message and status code from an error before returning a response from the API: ```go switch err.(type) { case errcode.ErrorCode: daError, _ := err.(errcode.ErrorCode) statusCode = daError.Descriptor().HTTPStatusCode errMsg = daError.Message() case errcode.Error: // For reference, if you're looking for a particular error // then you can do something like : // import ( derr "github.com/docker/docker/errors" ) // if daError.ErrorCode() == derr.ErrorCodeNoSuchContainer { ... } daError, _ := err.(errcode.Error) statusCode = daError.ErrorCode().Descriptor().HTTPStatusCode errMsg = daError.Message default: // This part of will be removed once we've // converted everything over to use the errcode package // FIXME: this is brittle and should not be necessary. // If we need to differentiate between different possible error types, // we should create appropriate error types with clearly defined meaning errStr := strings.ToLower(err.Error()) for keyword, status := range map[string]int{ "not found": http.StatusNotFound, "no such": http.StatusNotFound, "bad parameter": http.StatusBadRequest, "conflict": http.StatusConflict, "impossible": http.StatusNotAcceptable, "wrong login/password": http.StatusUnauthorized, "hasn't been activated": http.StatusForbidden, } { if strings.Contains(errStr, keyword) { statusCode = status break } } } ``` You can notice two things in that code: 1. We have to explain how errors work, because our implementation goes against how easy to use Go errors are. 2. At no moment we arrived to remove that `switch` statement that was the original reason to use our custom implementation. This change removes all our status errors from the errors package and puts them back in their specific contexts. IT puts the messages back with their contexts. That way, we know right away when errors used and how to generate their messages. It uses custom interfaces to reason about errors. Errors that need to response with a custom status code MUST implementent this simple interface: ```go type errorWithStatus interface { HTTPErrorStatusCode() int } ``` This interface is very straightforward to implement. It also preserves Go errors real behavior, getting the message is as simple as using the `Error()` method. I included helper functions to generate errors that use custom status code in `errors/errors.go`. By doing this, we remove the hard dependency we have eeverywhere to our custom errors package. Yes, you can use it as a helper to generate error, but it's still very easy to generate errors without it. Please, read this fantastic blog post about errors in Go: http://dave.cheney.net/2014/12/24/inspecting-errors Signed-off-by: David Calavera <david.calavera@gmail.com>
2016-02-25 15:53:35 +00:00
s.InitRouter(utils.IsDebugEnabled(), routers...)
}
func (cli *DaemonCli) initMiddlewares(s *apiserver.Server, cfg *apiserver.Config) {
v := cfg.Version
vm := middleware.NewVersionMiddleware(v, api.DefaultVersion, api.MinVersion)
s.UseMiddleware(vm)
if cfg.EnableCors {
c := middleware.NewCORSMiddleware(cfg.CorsHeaders)
s.UseMiddleware(c)
}
u := middleware.NewUserAgentMiddleware(v)
s.UseMiddleware(u)
cli.authzMiddleware = authorization.NewMiddleware(cli.Config.AuthorizationPlugins)
s.UseMiddleware(cli.authzMiddleware)
}