server: add diagnostics service (#2024)

Signed-off-by: Lonng <heng@lonng.org>

conf/config.toml

@@ -45,8 +45,6 @@ level = "info"
 # max-days = 28
 ## maximum number of old log files to retain
 # max-backups = 7
-## rotate log by day
-# log-rotate = true
 
 [metric]
 ## prometheus client push interval, set "0s" to disable prometheus.

go.mod

@@ -16,27 +16,26 @@ require (
 	github.com/ghodss/yaml v1.0.1-0.20190212211648-25d852aebe32
 	github.com/gogo/protobuf v1.0.0
 	github.com/golang/groupcache v0.0.0-20181024230925-c65c006176ff // indirect
-	github.com/golang/protobuf v1.2.0
+	github.com/golang/protobuf v1.3.2
 	github.com/golang/snappy v0.0.0-20180518054509-2e65f85255db // indirect
 	github.com/google/btree v0.0.0-20180813153112-4030bb1f1f0c
 	github.com/gorilla/context v0.0.0-20160226214623-1ea25387ff6f // indirect
 	github.com/gorilla/mux v1.6.1
 	github.com/gorilla/websocket v1.2.0 // indirect
 	github.com/grpc-ecosystem/go-grpc-prometheus v1.2.0
 	github.com/juju/ratelimit v1.0.1
-	github.com/kr/pretty v0.1.0 // indirect
 	github.com/mattn/go-shellwords v1.0.3
 	github.com/matttproud/golang_protobuf_extensions v1.0.1 // indirect
 	github.com/montanaflynn/stats v0.0.0-20151014174947-eeaced052adb
 	github.com/opentracing/opentracing-go v1.0.2
-	github.com/pingcap/check v0.0.0-20190102082844-67f458068fc8
+	github.com/pingcap/check v0.0.0-20191107115940-caf2b9e6ccf4
 	github.com/pingcap/errcode v0.0.0-20180921232412-a1a7271709d9
-	github.com/pingcap/failpoint v0.0.0-20190512135322-30cc7431d99c
-	github.com/pingcap/kvproto v0.0.0-20191030021250-51b332bcb20b
-	github.com/pingcap/log v0.0.0-20190715063458-479153f07ebd
+	github.com/pingcap/failpoint v0.0.0-20191029060244-12f4ac2fd11d
+	github.com/pingcap/kvproto v0.0.0-20191213111810-93cb7c623c8b
+	github.com/pingcap/log v0.0.0-20191012051959-b742a5d432e9
+	github.com/pingcap/sysutil v0.0.0-20191216090214-5f9620d22b3b
 	github.com/pkg/errors v0.8.1
 	github.com/prometheus/client_golang v0.8.0
-	github.com/prometheus/client_model v0.0.0-20171117100541-99fa1f4be8e5 // indirect
 	github.com/sirupsen/logrus v1.0.5
 	github.com/spf13/cobra v0.0.3
 	github.com/spf13/pflag v1.0.1
@@ -46,12 +45,10 @@ require (
 	github.com/urfave/negroni v0.3.0
 	go.etcd.io/etcd v0.0.0-20190320044326-77d4b742cdbf
 	go.uber.org/goleak v0.10.0
-	go.uber.org/zap v1.9.1
+	go.uber.org/zap v1.12.0
 	golang.org/x/crypto v0.0.0-20190909091759-094676da4a83 // indirect
 	golang.org/x/net v0.0.0-20190909003024-a7b16738d86b // indirect
-	golang.org/x/sync v0.0.0-20190423024810-112230192c58 // indirect
 	golang.org/x/sys v0.0.0-20190909082730-f460065e899a // indirect
-	google.golang.org/grpc v1.14.0
-	gopkg.in/check.v1 v1.0.0-20180628173108-788fd7840127 // indirect
+	google.golang.org/grpc v1.25.1
 	gopkg.in/natefinch/lumberjack.v2 v2.0.0
 )

pkg/logutil/log.go

@@ -44,8 +44,6 @@ const (
 type FileLogConfig struct {
 	// Log filename, leave empty to disable file log.
 	Filename string `toml:"filename" json:"filename"`
-	// Is log rotate enabled. TODO.
-	LogRotate bool `toml:"log-rotate" json:"log-rotate"`
 	// Max size for a single file, in MB.
 	MaxSize int `toml:"max-size" json:"max-size"`
 	// Max log keep days, default is never deleting.

pkg/testutil/leak.go

@@ -25,6 +25,7 @@ var LeakOptions = []goleak.Option{
 	goleak.IgnoreTopFunction("google.golang.org/grpc.(*ccBalancerWrapper).watcher"),
 	goleak.IgnoreTopFunction("google.golang.org/grpc.(*ccResolverWrapper).watcher"),
 	goleak.IgnoreTopFunction("google.golang.org/grpc.(*addrConn).createTransport"),
+	goleak.IgnoreTopFunction("google.golang.org/grpc.(*addrConn).resetTransport"),
 	goleak.IgnoreTopFunction("go.etcd.io/etcd/pkg/logutil.(*MergeLogger).outputLoop"),
 	// TODO: remove the below options once we fixed the http connection leak problems
 	goleak.IgnoreTopFunction("internal/poll.runtime_pollWait"),

server/config/config.go

@@ -156,7 +156,6 @@ func NewConfig() *Config {
 
 	fs.StringVar(&cfg.Log.Level, "L", "", "log level: debug, info, warn, error, fatal (default 'info')")
 	fs.StringVar(&cfg.Log.File.Filename, "log-file", "", "log file path")
-	fs.BoolVar(&cfg.Log.File.LogRotate, "log-rotate", true, "rotate log")
 
 	fs.StringVar(&cfg.Security.CAPath, "cacert", "", "Path of file that contains list of trusted TLS CAs")
 	fs.StringVar(&cfg.Security.CertPath, "cert", "", "Path of file that contains X509 certificate in PEM format")

server/server.go

@@ -30,6 +30,7 @@ import (
 	"github.com/golang/protobuf/proto"
 	"github.com/gorilla/mux"
 	"github.com/pingcap/failpoint"
+	"github.com/pingcap/kvproto/pkg/diagnosticspb"
 	"github.com/pingcap/kvproto/pkg/metapb"
 	"github.com/pingcap/kvproto/pkg/pdpb"
 	"github.com/pingcap/log"
@@ -46,6 +47,7 @@ import (
 	syncer "github.com/pingcap/pd/server/region_syncer"
 	"github.com/pingcap/pd/server/schedule/opt"
 	"github.com/pingcap/pd/server/tso"
+	"github.com/pingcap/sysutil"
 	"github.com/pkg/errors"
 	"github.com/urfave/negroni"
 	"go.etcd.io/etcd/clientv3"
@@ -72,6 +74,8 @@ var EnableZap = false
 
 // Server is the pd server.
 type Server struct {
+	diagnosticspb.DiagnosticsServer
+
 	// Server state.
 	isServing int64
 
@@ -169,9 +173,10 @@ func CreateServer(cfg *config.Config, apiBuilders ...HandlerBuilder) (*Server, e
 	rand.Seed(time.Now().UnixNano())
 
 	s := &Server{
-		cfg:         cfg,
-		scheduleOpt: config.NewScheduleOption(cfg),
-		member:      &member.Member{},
+		DiagnosticsServer: sysutil.NewDiagnosticsServer(cfg.Log.File.Filename),
+		cfg:               cfg,
+		scheduleOpt:       config.NewScheduleOption(cfg),
+		member:            &member.Member{},
 	}
 	s.handler = newHandler(s)
 
@@ -191,7 +196,10 @@ func CreateServer(cfg *config.Config, apiBuilders ...HandlerBuilder) (*Server, e
 			webPath:     http.StripPrefix(webPath, ui.Handler()),
 		}
 	}
-	etcdCfg.ServiceRegister = func(gs *grpc.Server) { pdpb.RegisterPDServer(gs, s) }
+	etcdCfg.ServiceRegister = func(gs *grpc.Server) {
+		pdpb.RegisterPDServer(gs, s)
+		diagnosticspb.RegisterDiagnosticsServer(gs, s)
+	}
 	s.etcdCfg = etcdCfg
 	if EnableZap {
 		// The etcd master version has removed embed.Config.SetupLogging.