version: bump up to v3.1.11

Automated cherry pick of #8813 release 3.1

.semaphore.sh

@@ -0,0 +1,16 @@
+#!/usr/bin/env bash
+
+TEST_SUFFIX=$(date +%s | base64 | head -c 15)
+
+TEST_OPTS="RELEASE_TEST=y INTEGRATION=y PASSES='build unit release integration_e2e functional' MANUAL_VER=v3.1.10"
+if [ "$TEST_ARCH" == "386" ]; then
+	TEST_OPTS="GOARCH=386 PASSES='build unit integration_e2e'"
+fi
+
+docker run \
+	--rm \
+	--volume=`pwd`:/go/src/github.com/coreos/etcd \
+	gcr.io/etcd-development/etcd-test:go1.8.5 \
+	/bin/bash -c "${TEST_OPTS} ./test 2>&1 | tee test-${TEST_SUFFIX}.log"
+
+! egrep "(--- FAIL:|leak)" -A10 -B50 test-${TEST_SUFFIX}.log

.travis.yml

@@ -4,7 +4,7 @@ go_import_path: github.com/coreos/etcd
 sudo: false
 
 go:
-  - 1.8.3
+  - 1.8.5
 
 notifications:
   on_success: never

Dockerfile-test

@@ -0,0 +1,57 @@
+FROM ubuntu:16.10
+
+RUN rm /bin/sh && ln -s /bin/bash /bin/sh
+RUN echo 'debconf debconf/frontend select Noninteractive' | debconf-set-selections
+
+RUN apt-get -y update \
+  && apt-get -y install \
+  build-essential \
+  gcc \
+  apt-utils \
+  pkg-config \
+  software-properties-common \
+  apt-transport-https \
+  libssl-dev \
+  sudo \
+  bash \
+  curl \
+  wget \
+  tar \
+  git \
+  netcat \
+  libaspell-dev \
+  libhunspell-dev \
+  hunspell-en-us \
+  aspell-en \
+  shellcheck \
+  && apt-get -y update \
+  && apt-get -y upgrade \
+  && apt-get -y autoremove \
+  && apt-get -y autoclean
+
+ENV GOROOT /usr/local/go
+ENV GOPATH /go
+ENV PATH ${GOPATH}/bin:${GOROOT}/bin:${PATH}
+ENV GO_VERSION REPLACE_ME_GO_VERSION
+ENV GO_DOWNLOAD_URL https://storage.googleapis.com/golang
+RUN rm -rf ${GOROOT} \
+  && curl -s ${GO_DOWNLOAD_URL}/go${GO_VERSION}.linux-amd64.tar.gz | tar -v -C /usr/local/ -xz \
+  && mkdir -p ${GOPATH}/src ${GOPATH}/bin \
+  && go version
+
+RUN mkdir -p ${GOPATH}/src/github.com/coreos/etcd
+WORKDIR ${GOPATH}/src/github.com/coreos/etcd
+
+ADD ./scripts/install-marker.sh /tmp/install-marker.sh
+
+RUN go get -v -u -tags spell github.com/chzchzchz/goword \
+  && go get -v -u github.com/coreos/license-bill-of-materials \
+  && go get -v -u honnef.co/go/tools/cmd/gosimple \
+  && go get -v -u honnef.co/go/tools/cmd/unused \
+  && go get -v -u honnef.co/go/tools/cmd/staticcheck \
+  && go get -v -u github.com/wadey/gocovmerge \
+  && go get -v -u github.com/gordonklaus/ineffassign \
+  && /tmp/install-marker.sh amd64 \
+  && rm -f /tmp/install-marker.sh \
+  && curl -s https://codecov.io/bash >/codecov \
+  && chmod 700 /codecov

client/integration/client_test.go

@@ -34,7 +34,7 @@ import (
 func TestV2NoRetryEOF(t *testing.T) {
 	defer testutil.AfterTest(t)
 	// generate an EOF response; specify address so appears first in sorted ep list
-	lEOF := integration.NewListenerWithAddr(t, fmt.Sprintf("eof:123.%d.sock", os.Getpid()))
+	lEOF := integration.NewListenerWithAddr(t, fmt.Sprintf("127.0.0.1:%05d", os.Getpid()))
 	defer lEOF.Close()
 	tries := uint32(0)
 	go func() {
@@ -65,8 +65,7 @@ func TestV2NoRetryEOF(t *testing.T) {
 // TestV2NoRetryNoLeader tests destructive api calls won't retry if given an error code.
 func TestV2NoRetryNoLeader(t *testing.T) {
 	defer testutil.AfterTest(t)
-
-	lHttp := integration.NewListenerWithAddr(t, fmt.Sprintf("errHttp:123.%d.sock", os.Getpid()))
+	lHttp := integration.NewListenerWithAddr(t, fmt.Sprintf("127.0.0.1:%05d", os.Getpid()))
 	eh := &errHandler{errCode: http.StatusServiceUnavailable}
 	srv := httptest.NewUnstartedServer(eh)
 	defer lHttp.Close()

cmd/vendor/github.com/boltdb/bolt/bolt_amd64.go

@@ -1,7 +0,0 @@
-package bolt
-
-// maxMapSize represents the largest mmap size supported by Bolt.
-const maxMapSize = 0xFFFFFFFFFFFF // 256TB
-
-// maxAllocSize is the size used when creating array pointers.
-const maxAllocSize = 0x7FFFFFFF

cmd/vendor/github.com/coreos/bbolt/bolt_386.go

@@ -5,3 +5,6 @@ const maxMapSize = 0x7FFFFFFF // 2GB
 
 // maxAllocSize is the size used when creating array pointers.
 const maxAllocSize = 0xFFFFFFF
+
+// Are unaligned load/stores broken on this arch?
+var brokenUnaligned = false

cmd/vendor/github.com/coreos/bbolt/bolt_amd64.go

@@ -0,0 +1,10 @@
+package bolt
+
+// maxMapSize represents the largest mmap size supported by Bolt.
+const maxMapSize = 0xFFFFFFFFFFFF // 256TB
+
+// maxAllocSize is the size used when creating array pointers.
+const maxAllocSize = 0x7FFFFFFF
+
+// Are unaligned load/stores broken on this arch?
+var brokenUnaligned = false

cmd/vendor/github.com/coreos/bbolt/bolt_arm.go

@@ -0,0 +1,28 @@
+package bolt
+
+import "unsafe"
+
+// maxMapSize represents the largest mmap size supported by Bolt.
+const maxMapSize = 0x7FFFFFFF // 2GB
+
+// maxAllocSize is the size used when creating array pointers.
+const maxAllocSize = 0xFFFFFFF
+
+// Are unaligned load/stores broken on this arch?
+var brokenUnaligned bool
+
+func init() {
+	// Simple check to see whether this arch handles unaligned load/stores
+	// correctly.
+
+	// ARM9 and older devices require load/stores to be from/to aligned
+	// addresses. If not, the lower 2 bits are cleared and that address is
+	// read in a jumbled up order.
+
+	// See http://infocenter.arm.com/help/index.jsp?topic=/com.arm.doc.faqs/ka15414.html
+
+	raw := [6]byte{0xfe, 0xef, 0x11, 0x22, 0x22, 0x11}
+	val := *(*uint32)(unsafe.Pointer(uintptr(unsafe.Pointer(&raw)) + 2))
+
+	brokenUnaligned = val != 0x11222211
+}

cmd/vendor/github.com/coreos/bbolt/bolt_arm64.go

@@ -7,3 +7,6 @@ const maxMapSize = 0xFFFFFFFFFFFF // 256TB
 
 // maxAllocSize is the size used when creating array pointers.
 const maxAllocSize = 0x7FFFFFFF
+
+// Are unaligned load/stores broken on this arch?
+var brokenUnaligned = false

cmd/vendor/github.com/coreos/bbolt/bolt_mips64x.go

@@ -0,0 +1,12 @@
+// +build mips64 mips64le
+
+package bolt
+
+// maxMapSize represents the largest mmap size supported by Bolt.
+const maxMapSize = 0x8000000000 // 512GB
+
+// maxAllocSize is the size used when creating array pointers.
+const maxAllocSize = 0x7FFFFFFF
+
+// Are unaligned load/stores broken on this arch?
+var brokenUnaligned = false

cmd/vendor/github.com/coreos/bbolt/bolt_mipsx.go

@@ -1,7 +1,12 @@
+// +build mips mipsle
+
 package bolt
 
 // maxMapSize represents the largest mmap size supported by Bolt.
-const maxMapSize = 0x7FFFFFFF // 2GB
+const maxMapSize = 0x40000000 // 1GB
 
 // maxAllocSize is the size used when creating array pointers.
 const maxAllocSize = 0xFFFFFFF
+
+// Are unaligned load/stores broken on this arch?
+var brokenUnaligned = false

cmd/vendor/github.com/coreos/bbolt/bolt_ppc64.go

@@ -7,3 +7,6 @@ const maxMapSize = 0xFFFFFFFFFFFF // 256TB
 
 // maxAllocSize is the size used when creating array pointers.
 const maxAllocSize = 0x7FFFFFFF
+
+// Are unaligned load/stores broken on this arch?
+var brokenUnaligned = false

cmd/vendor/github.com/coreos/bbolt/bolt_ppc64le.go

@@ -7,3 +7,6 @@ const maxMapSize = 0xFFFFFFFFFFFF // 256TB
 
 // maxAllocSize is the size used when creating array pointers.
 const maxAllocSize = 0x7FFFFFFF
+
+// Are unaligned load/stores broken on this arch?
+var brokenUnaligned = false

cmd/vendor/github.com/coreos/bbolt/bolt_s390x.go

@@ -7,3 +7,6 @@ const maxMapSize = 0xFFFFFFFFFFFF // 256TB
 
 // maxAllocSize is the size used when creating array pointers.
 const maxAllocSize = 0x7FFFFFFF
+
+// Are unaligned load/stores broken on this arch?
+var brokenUnaligned = false

cmd/vendor/github.com/coreos/bbolt/bolt_unix.go

@@ -13,29 +13,32 @@ import (
 // flock acquires an advisory lock on a file descriptor.
 func flock(db *DB, mode os.FileMode, exclusive bool, timeout time.Duration) error {
 	var t time.Time
+	if timeout != 0 {
+		t = time.Now()
+	}
+	fd := db.file.Fd()
+	flag := syscall.LOCK_NB
+	if exclusive {
+		flag |= syscall.LOCK_EX
+	} else {
+		flag |= syscall.LOCK_SH
+	}
 	for {
-		// If we're beyond our timeout then return an error.
-		// This can only occur after we've attempted a flock once.
-		if t.IsZero() {
-			t = time.Now()
-		} else if timeout > 0 && time.Since(t) > timeout {
-			return ErrTimeout
-		}
-		flag := syscall.LOCK_SH
-		if exclusive {
-			flag = syscall.LOCK_EX
-		}
-
-		// Otherwise attempt to obtain an exclusive lock.
-		err := syscall.Flock(int(db.file.Fd()), flag|syscall.LOCK_NB)
+		// Attempt to obtain an exclusive lock.
+		err := syscall.Flock(int(fd), flag)
 		if err == nil {
 			return nil
 		} else if err != syscall.EWOULDBLOCK {
 			return err
 		}
 
+		// If we timed out then return an error.
+		if timeout != 0 && time.Since(t) > timeout-flockRetryTimeout {
+			return ErrTimeout
+		}
+
 		// Wait for a bit and try again.
-		time.Sleep(50 * time.Millisecond)
+		time.Sleep(flockRetryTimeout)
 	}
 }
 

cmd/vendor/github.com/coreos/bbolt/bolt_unix_solaris.go

@@ -13,34 +13,33 @@ import (
 // flock acquires an advisory lock on a file descriptor.
 func flock(db *DB, mode os.FileMode, exclusive bool, timeout time.Duration) error {
 	var t time.Time
+	if timeout != 0 {
+		t = time.Now()
+	}
+	fd := db.file.Fd()
+	var lockType int16
+	if exclusive {
+		lockType = syscall.F_WRLCK
+	} else {
+		lockType = syscall.F_RDLCK
+	}
 	for {
-		// If we're beyond our timeout then return an error.
-		// This can only occur after we've attempted a flock once.
-		if t.IsZero() {
-			t = time.Now()
-		} else if timeout > 0 && time.Since(t) > timeout {
-			return ErrTimeout
-		}
-		var lock syscall.Flock_t
-		lock.Start = 0
-		lock.Len = 0
-		lock.Pid = 0
-		lock.Whence = 0
-		lock.Pid = 0
-		if exclusive {
-			lock.Type = syscall.F_WRLCK
-		} else {
-			lock.Type = syscall.F_RDLCK
-		}
-		err := syscall.FcntlFlock(db.file.Fd(), syscall.F_SETLK, &lock)
+		// Attempt to obtain an exclusive lock.
+		lock := syscall.Flock_t{Type: lockType}
+		err := syscall.FcntlFlock(fd, syscall.F_SETLK, &lock)
 		if err == nil {
 			return nil
 		} else if err != syscall.EAGAIN {
 			return err
 		}
 
+		// If we timed out then return an error.
+		if timeout != 0 && time.Since(t) > timeout-flockRetryTimeout {
+			return ErrTimeout
+		}
+
 		// Wait for a bit and try again.
-		time.Sleep(50 * time.Millisecond)
+		time.Sleep(flockRetryTimeout)
 	}
 }
 

cmd/vendor/github.com/coreos/bbolt/bolt_windows.go

@@ -59,29 +59,30 @@ func flock(db *DB, mode os.FileMode, exclusive bool, timeout time.Duration) erro
 	db.lockfile = f
 
 	var t time.Time
+	if timeout != 0 {
+		t = time.Now()
+	}
+	fd := f.Fd()
+	var flag uint32 = flagLockFailImmediately
+	if exclusive {
+		flag |= flagLockExclusive
+	}
 	for {
-		// If we're beyond our timeout then return an error.
-		// This can only occur after we've attempted a flock once.
-		if t.IsZero() {
-			t = time.Now()
-		} else if timeout > 0 && time.Since(t) > timeout {
-			return ErrTimeout
-		}
-
-		var flag uint32 = flagLockFailImmediately
-		if exclusive {
-			flag |= flagLockExclusive
-		}
-
-		err := lockFileEx(syscall.Handle(db.lockfile.Fd()), flag, 0, 1, 0, &syscall.Overlapped{})
+		// Attempt to obtain an exclusive lock.
+		err := lockFileEx(syscall.Handle(fd), flag, 0, 1, 0, &syscall.Overlapped{})
 		if err == nil {
 			return nil
 		} else if err != errLockViolation {
 			return err
 		}
 
+		// If we timed oumercit then return an error.
+		if timeout != 0 && time.Since(t) > timeout-flockRetryTimeout {
+			return ErrTimeout
+		}
+
 		// Wait for a bit and try again.
-		time.Sleep(50 * time.Millisecond)
+		time.Sleep(flockRetryTimeout)
 	}
 }
 
@@ -89,7 +90,7 @@ func flock(db *DB, mode os.FileMode, exclusive bool, timeout time.Duration) erro
 func funlock(db *DB) error {
 	err := unlockFileEx(syscall.Handle(db.lockfile.Fd()), 0, 1, 0, &syscall.Overlapped{})
 	db.lockfile.Close()
-	os.Remove(db.path+lockExt)
+	os.Remove(db.path + lockExt)
 	return err
 }
 

cmd/vendor/github.com/coreos/bbolt/bucket.go

@@ -14,13 +14,6 @@ const (
 	MaxValueSize = (1 << 31) - 2
 )
 
-const (
-	maxUint = ^uint(0)
-	minUint = 0
-	maxInt  = int(^uint(0) >> 1)
-	minInt  = -maxInt - 1
-)
-
 const bucketHeaderSize = int(unsafe.Sizeof(bucket{}))
 
 const (
@@ -130,9 +123,17 @@ func (b *Bucket) Bucket(name []byte) *Bucket {
 func (b *Bucket) openBucket(value []byte) *Bucket {
 	var child = newBucket(b.tx)
 
+	// If unaligned load/stores are broken on this arch and value is
+	// unaligned simply clone to an aligned byte array.
+	unaligned := brokenUnaligned && uintptr(unsafe.Pointer(&value[0]))&3 != 0
+
+	if unaligned {
+		value = cloneBytes(value)
+	}
+
 	// If this is a writable transaction then we need to copy the bucket entry.
 	// Read-only transactions can point directly at the mmap entry.
-	if b.tx.writable {
+	if b.tx.writable && !unaligned {
 		child.bucket = &bucket{}
 		*child.bucket = *(*bucket)(unsafe.Pointer(&value[0]))
 	} else {
@@ -167,9 +168,8 @@ func (b *Bucket) CreateBucket(key []byte) (*Bucket, error) {
 	if bytes.Equal(key, k) {
 		if (flags & bucketLeafFlag) != 0 {
 			return nil, ErrBucketExists
-		} else {
-			return nil, ErrIncompatibleValue
 		}
+		return nil, ErrIncompatibleValue
 	}
 
 	// Create empty, inline bucket.
@@ -316,7 +316,12 @@ func (b *Bucket) Delete(key []byte) error {
 
 	// Move cursor to correct position.
 	c := b.Cursor()
-	_, _, flags := c.seek(key)
+	k, _, flags := c.seek(key)
+
+	// Return nil if the key doesn't exist.
+	if !bytes.Equal(key, k) {
+		return nil
+	}
 
 	// Return an error if there is already existing bucket value.
 	if (flags & bucketLeafFlag) != 0 {
@@ -329,6 +334,28 @@ func (b *Bucket) Delete(key []byte) error {
 	return nil
 }
 
+// Sequence returns the current integer for the bucket without incrementing it.
+func (b *Bucket) Sequence() uint64 { return b.bucket.sequence }
+
+// SetSequence updates the sequence number for the bucket.
+func (b *Bucket) SetSequence(v uint64) error {
+	if b.tx.db == nil {
+		return ErrTxClosed
+	} else if !b.Writable() {
+		return ErrTxNotWritable
+	}
+
+	// Materialize the root node if it hasn't been already so that the
+	// bucket will be saved during commit.
+	if b.rootNode == nil {
+		_ = b.node(b.root, nil)
+	}
+
+	// Increment and return the sequence.
+	b.bucket.sequence = v
+	return nil
+}
+
 // NextSequence returns an autoincrementing integer for the bucket.
 func (b *Bucket) NextSequence() (uint64, error) {
 	if b.tx.db == nil {

cmd/vendor/github.com/coreos/bbolt/db.go

@@ -7,8 +7,7 @@ import (
 	"log"
 	"os"
 	"runtime"
-	"runtime/debug"
-	"strings"
+	"sort"
 	"sync"
 	"time"
 	"unsafe"
@@ -23,6 +22,8 @@ const version = 2
 // Represents a marker value to indicate that a file is a Bolt DB.
 const magic uint32 = 0xED0CDAED
 
+const pgidNoFreelist pgid = 0xffffffffffffffff
+
 // IgnoreNoSync specifies whether the NoSync field of a DB is ignored when
 // syncing changes to a file.  This is required as some operating systems,
 // such as OpenBSD, do not have a unified buffer cache (UBC) and writes
@@ -39,6 +40,9 @@ const (
 // default page size for db is set to the OS page size.
 var defaultPageSize = os.Getpagesize()
 
+// The time elapsed between consecutive file locking attempts.
+const flockRetryTimeout = 50 * time.Millisecond
+
 // DB represents a collection of buckets persisted to a file on disk.
 // All data access is performed through transactions which can be obtained through the DB.
 // All the functions on DB will return a ErrDatabaseNotOpen if accessed before Open() is called.
@@ -61,6 +65,11 @@ type DB struct {
 	// THIS IS UNSAFE. PLEASE USE WITH CAUTION.
 	NoSync bool
 
+	// When true, skips syncing freelist to disk. This improves the database
+	// write performance under normal operation, but requires a full database
+	// re-sync during recovery.
+	NoFreelistSync bool
+
 	// When true, skips the truncate call when growing the database.
 	// Setting this to true is only safe on non-ext3/ext4 systems.
 	// Skipping truncation avoids preallocation of hard drive space and
@@ -107,9 +116,11 @@ type DB struct {
 	opened   bool
 	rwtx     *Tx
 	txs      []*Tx
-	freelist *freelist
 	stats    Stats
 
+	freelist     *freelist
+	freelistLoad sync.Once
+
 	pagePool sync.Pool
 
 	batchMu sync.Mutex
@@ -148,14 +159,17 @@ func (db *DB) String() string {
 // If the file does not exist then it will be created automatically.
 // Passing in nil options will cause Bolt to open the database with the default options.
 func Open(path string, mode os.FileMode, options *Options) (*DB, error) {
-	var db = &DB{opened: true}
-
+	db := &DB{
+		opened: true,
+	}
 	// Set default options if no options are provided.
 	if options == nil {
 		options = DefaultOptions
 	}
+	db.NoSync = options.NoSync
 	db.NoGrowSync = options.NoGrowSync
 	db.MmapFlags = options.MmapFlags
+	db.NoFreelistSync = options.NoFreelistSync
 
 	// Set default values for later DB operations.
 	db.MaxBatchSize = DefaultMaxBatchSize
@@ -184,6 +198,7 @@ func Open(path string, mode os.FileMode, options *Options) (*DB, error) {
 	// The database file is locked using the shared lock (more than one process may
 	// hold a lock at the same time) otherwise (options.ReadOnly is set).
 	if err := flock(db, mode, !db.readOnly, options.Timeout); err != nil {
+		db.lockfile = nil // make 'unused' happy. TODO: rework locks
 		_ = db.close()
 		return nil, err
 	}
@@ -191,6 +206,11 @@ func Open(path string, mode os.FileMode, options *Options) (*DB, error) {
 	// Default values for test hooks
 	db.ops.writeAt = db.file.WriteAt
 
+	if db.pageSize = options.PageSize; db.pageSize == 0 {
+		// Set the default page size to the OS page size.
+		db.pageSize = defaultPageSize
+	}
+
 	// Initialize the database if it doesn't exist.
 	if info, err := db.file.Stat(); err != nil {
 		return nil, err
@@ -202,20 +222,21 @@ func Open(path string, mode os.FileMode, options *Options) (*DB, error) {
 	} else {
 		// Read the first meta page to determine the page size.
 		var buf [0x1000]byte
-		if _, err := db.file.ReadAt(buf[:], 0); err == nil {
-			m := db.pageInBuffer(buf[:], 0).meta()
-			if err := m.validate(); err != nil {
-				// If we can't read the page size, we can assume it's the same
-				// as the OS -- since that's how the page size was chosen in the
-				// first place.
-				//
-				// If the first page is invalid and this OS uses a different
-				// page size than what the database was created with then we
-				// are out of luck and cannot access the database.
-				db.pageSize = os.Getpagesize()
-			} else {
+		// If we can't read the page size, but can read a page, assume
+		// it's the same as the OS or one given -- since that's how the
+		// page size was chosen in the first place.
+		//
+		// If the first page is invalid and this OS uses a different
+		// page size than what the database was created with then we
+		// are out of luck and cannot access the database.
+		//
+		// TODO: scan for next page
+		if bw, err := db.file.ReadAt(buf[:], 0); err == nil && bw == len(buf) {
+			if m := db.pageInBuffer(buf[:], 0).meta(); m.validate() == nil {
 				db.pageSize = int(m.pageSize)
 			}
+		} else {
+			return nil, ErrInvalid
 		}
 	}
 
@@ -232,14 +253,50 @@ func Open(path string, mode os.FileMode, options *Options) (*DB, error) {
 		return nil, err
 	}
 
-	// Read in the freelist.
-	db.freelist = newFreelist()
-	db.freelist.read(db.page(db.meta().freelist))
+	if db.readOnly {
+		return db, nil
+	}
+
+	db.loadFreelist()
+
+	// Flush freelist when transitioning from no sync to sync so
+	// NoFreelistSync unaware boltdb can open the db later.
+	if !db.NoFreelistSync && !db.hasSyncedFreelist() {
+		tx, err := db.Begin(true)
+		if tx != nil {
+			err = tx.Commit()
+		}
+		if err != nil {
+			_ = db.close()
+			return nil, err
+		}
+	}
 
 	// Mark the database as opened and return.
 	return db, nil
 }
 
+// loadFreelist reads the freelist if it is synced, or reconstructs it
+// by scanning the DB if it is not synced. It assumes there are no
+// concurrent accesses being made to the freelist.
+func (db *DB) loadFreelist() {
+	db.freelistLoad.Do(func() {
+		db.freelist = newFreelist()
+		if !db.hasSyncedFreelist() {
+			// Reconstruct free list by scanning the DB.
+			db.freelist.readIDs(db.freepages())
+		} else {
+			// Read free list from freelist page.
+			db.freelist.read(db.page(db.meta().freelist))
+		}
+		db.stats.FreePageN = len(db.freelist.ids)
+	})
+}
+
+func (db *DB) hasSyncedFreelist() bool {
+	return db.meta().freelist != pgidNoFreelist
+}
+
 // mmap opens the underlying memory-mapped file and initializes the meta references.
 // minsz is the minimum size that the new mmap can be.
 func (db *DB) mmap(minsz int) error {
@@ -341,9 +398,6 @@ func (db *DB) mmapSize(size int) (int, error) {
 
 // init creates a new database file and initializes its meta pages.
 func (db *DB) init() error {
-	// Set the page size to the OS page size.
-	db.pageSize = os.Getpagesize()
-
 	// Create two meta pages on a buffer.
 	buf := make([]byte, db.pageSize*4)
 	for i := 0; i < 2; i++ {
@@ -526,21 +580,36 @@ func (db *DB) beginRWTx() (*Tx, error) {
 	t := &Tx{writable: true}
 	t.init(db)
 	db.rwtx = t
+	db.freePages()
+	return t, nil
+}
 
-	// Free any pages associated with closed read-only transactions.
-	var minid txid = 0xFFFFFFFFFFFFFFFF
-	for _, t := range db.txs {
-		if t.meta.txid < minid {
-			minid = t.meta.txid
-		}
+// freePages releases any pages associated with closed read-only transactions.
+func (db *DB) freePages() {
+	// Free all pending pages prior to earliest open transaction.
+	sort.Sort(txsById(db.txs))
+	minid := txid(0xFFFFFFFFFFFFFFFF)
+	if len(db.txs) > 0 {
+		minid = db.txs[0].meta.txid
 	}
 	if minid > 0 {
 		db.freelist.release(minid - 1)
 	}
-
-	return t, nil
+	// Release unused txid extents.
+	for _, t := range db.txs {
+		db.freelist.releaseRange(minid, t.meta.txid-1)
+		minid = t.meta.txid + 1
+	}
+	db.freelist.releaseRange(minid, txid(0xFFFFFFFFFFFFFFFF))
+	// Any page both allocated and freed in an extent is safe to release.
 }
 
+type txsById []*Tx
+
+func (t txsById) Len() int           { return len(t) }
+func (t txsById) Swap(i, j int)      { t[i], t[j] = t[j], t[i] }
+func (t txsById) Less(i, j int) bool { return t[i].meta.txid < t[j].meta.txid }
+
 // removeTx removes a transaction from the database.
 func (db *DB) removeTx(tx *Tx) {
 	// Release the read lock on the mmap.
@@ -552,7 +621,10 @@ func (db *DB) removeTx(tx *Tx) {
 	// Remove the transaction.
 	for i, t := range db.txs {
 		if t == tx {
-			db.txs = append(db.txs[:i], db.txs[i+1:]...)
+			last := len(db.txs) - 1
+			db.txs[i] = db.txs[last]
+			db.txs[last] = nil
+			db.txs = db.txs[:last]
 			break
 		}
 	}
@@ -630,11 +702,7 @@ func (db *DB) View(fn func(*Tx) error) error {
 		return err
 	}
 
-	if err := t.Rollback(); err != nil {
-		return err
-	}
-
-	return nil
+	return t.Rollback()
 }
 
 // Batch calls fn as part of a batch. It behaves similar to Update,
@@ -823,7 +891,7 @@ func (db *DB) meta() *meta {
 }
 
 // allocate returns a contiguous block of memory starting at a given page.
-func (db *DB) allocate(count int) (*page, error) {
+func (db *DB) allocate(txid txid, count int) (*page, error) {
 	// Allocate a temporary buffer for the page.
 	var buf []byte
 	if count == 1 {
@@ -835,7 +903,7 @@ func (db *DB) allocate(count int) (*page, error) {
 	p.overflow = uint32(count - 1)
 
 	// Use pages from the freelist if they are available.
-	if p.id = db.freelist.allocate(count); p.id != 0 {
+	if p.id = db.freelist.allocate(txid, count); p.id != 0 {
 		return p, nil
 	}
 
@@ -890,6 +958,38 @@ func (db *DB) IsReadOnly() bool {
 	return db.readOnly
 }
 
+func (db *DB) freepages() []pgid {
+	tx, err := db.beginTx()
+	defer func() {
+		err = tx.Rollback()
+		if err != nil {
+			panic("freepages: failed to rollback tx")
+		}
+	}()
+	if err != nil {
+		panic("freepages: failed to open read only tx")
+	}
+
+	reachable := make(map[pgid]*page)
+	nofreed := make(map[pgid]bool)
+	ech := make(chan error)
+	go func() {
+		for e := range ech {
+			panic(fmt.Sprintf("freepages: failed to get all reachable pages (%v)", e))
+		}
+	}()
+	tx.checkBucket(&tx.root, reachable, nofreed, ech)
+	close(ech)
+
+	var fids []pgid
+	for i := pgid(2); i < db.meta().pgid; i++ {
+		if _, ok := reachable[i]; !ok {
+			fids = append(fids, i)
+		}
+	}
+	return fids
+}
+
 // Options represents the options that can be set when opening a database.
 type Options struct {
 	// Timeout is the amount of time to wait to obtain a file lock.
@@ -900,6 +1000,10 @@ type Options struct {
 	// Sets the DB.NoGrowSync flag before memory mapping the file.
 	NoGrowSync bool
 
+	// Do not sync freelist to disk. This improves the database write performance
+	// under normal operation, but requires a full database re-sync during recovery.
+	NoFreelistSync bool
+
 	// Open database in read-only mode. Uses flock(..., LOCK_SH |LOCK_NB) to
 	// grab a shared lock (UNIX).
 	ReadOnly bool
@@ -916,6 +1020,14 @@ type Options struct {
 	// If initialMmapSize is smaller than the previous database size,
 	// it takes no effect.
 	InitialMmapSize int
+
+	// PageSize overrides the default OS page size.
+	PageSize int
+
+	// NoSync sets the initial value of DB.NoSync. Normally this can just be
+	// set directly on the DB itself when returned from Open(), but this option
+	// is useful in APIs which expose Options but not the underlying DB.
+	NoSync bool
 }
 
 // DefaultOptions represent the options used if nil options are passed into Open().
@@ -952,15 +1064,11 @@ func (s *Stats) Sub(other *Stats) Stats {
 	diff.PendingPageN = s.PendingPageN
 	diff.FreeAlloc = s.FreeAlloc
 	diff.FreelistInuse = s.FreelistInuse
-	diff.TxN = other.TxN - s.TxN
+	diff.TxN = s.TxN - other.TxN
 	diff.TxStats = s.TxStats.Sub(&other.TxStats)
 	return diff
 }
 
-func (s *Stats) add(other *Stats) {
-	s.TxStats.add(&other.TxStats)
-}
-
 type Info struct {
 	Data     uintptr
 	PageSize int
@@ -999,7 +1107,8 @@ func (m *meta) copy(dest *meta) {
 func (m *meta) write(p *page) {
 	if m.root.root >= m.pgid {
 		panic(fmt.Sprintf("root bucket pgid (%d) above high water mark (%d)", m.root.root, m.pgid))
-	} else if m.freelist >= m.pgid {
+	} else if m.freelist >= m.pgid && m.freelist != pgidNoFreelist {
+		// TODO: reject pgidNoFreeList if !NoFreelistSync
 		panic(fmt.Sprintf("freelist pgid (%d) above high water mark (%d)", m.freelist, m.pgid))
 	}
 
@@ -1026,11 +1135,3 @@ func _assert(condition bool, msg string, v ...interface{}) {
 		panic(fmt.Sprintf("assertion failed: "+msg, v...))
 	}
 }
-
-func warn(v ...interface{})              { fmt.Fprintln(os.Stderr, v...) }
-func warnf(msg string, v ...interface{}) { fmt.Fprintf(os.Stderr, msg+"\n", v...) }
-
-func printstack() {
-	stack := strings.Join(strings.Split(string(debug.Stack()), "\n")[2:], "\n")
-	fmt.Fprintln(os.Stderr, stack)
-}

cmd/vendor/github.com/coreos/bbolt/freelist.go

@@ -6,25 +6,40 @@ import (
 	"unsafe"
 )
 
+// txPending holds a list of pgids and corresponding allocation txns
+// that are pending to be freed.
+type txPending struct {
+	ids              []pgid
+	alloctx          []txid // txids allocating the ids
+	lastReleaseBegin txid   // beginning txid of last matching releaseRange
+}
+
 // freelist represents a list of all pages that are available for allocation.
 // It also tracks pages that have been freed but are still in use by open transactions.
 type freelist struct {
-	ids     []pgid          // all free and available free page ids.
-	pending map[txid][]pgid // mapping of soon-to-be free page ids by tx.
-	cache   map[pgid]bool   // fast lookup of all free and pending page ids.
+	ids     []pgid              // all free and available free page ids.
+	allocs  map[pgid]txid       // mapping of txid that allocated a pgid.
+	pending map[txid]*txPending // mapping of soon-to-be free page ids by tx.
+	cache   map[pgid]bool       // fast lookup of all free and pending page ids.
 }
 
 // newFreelist returns an empty, initialized freelist.
 func newFreelist() *freelist {
 	return &freelist{
-		pending: make(map[txid][]pgid),
+		allocs:  make(map[pgid]txid),
+		pending: make(map[txid]*txPending),
 		cache:   make(map[pgid]bool),
 	}
 }
 
 // size returns the size of the page after serialization.
 func (f *freelist) size() int {
-	return pageHeaderSize + (int(unsafe.Sizeof(pgid(0))) * f.count())
+	n := f.count()
+	if n >= 0xFFFF {
+		// The first element will be used to store the count. See freelist.write.
+		n++
+	}
+	return pageHeaderSize + (int(unsafe.Sizeof(pgid(0))) * n)
 }
 
 // count returns count of pages on the freelist
@@ -40,27 +55,26 @@ func (f *freelist) free_count() int {
 // pending_count returns count of pending pages
 func (f *freelist) pending_count() int {
 	var count int
-	for _, list := range f.pending {
-		count += len(list)
+	for _, txp := range f.pending {
+		count += len(txp.ids)
 	}
 	return count
 }
 
-// all returns a list of all free ids and all pending ids in one sorted list.
-func (f *freelist) all() []pgid {
-	m := make(pgids, 0)
-
-	for _, list := range f.pending {
-		m = append(m, list...)
+// copyall copies into dst a list of all free ids and all pending ids in one sorted list.
+// f.count returns the minimum length required for dst.
+func (f *freelist) copyall(dst []pgid) {
+	m := make(pgids, 0, f.pending_count())
+	for _, txp := range f.pending {
+		m = append(m, txp.ids...)
 	}
-
 	sort.Sort(m)
-	return pgids(f.ids).merge(m)
+	mergepgids(dst, f.ids, m)
 }
 
 // allocate returns the starting page id of a contiguous list of pages of a given size.
 // If a contiguous block cannot be found then 0 is returned.
-func (f *freelist) allocate(n int) pgid {
+func (f *freelist) allocate(txid txid, n int) pgid {
 	if len(f.ids) == 0 {
 		return 0
 	}
@@ -93,7 +107,7 @@ func (f *freelist) allocate(n int) pgid {
 			for i := pgid(0); i < pgid(n); i++ {
 				delete(f.cache, initial+i)
 			}
-
+			f.allocs[initial] = txid
 			return initial
 		}
 
@@ -110,28 +124,73 @@ func (f *freelist) free(txid txid, p *page) {
 	}
 
 	// Free page and all its overflow pages.
-	var ids = f.pending[txid]
+	txp := f.pending[txid]
+	if txp == nil {
+		txp = &txPending{}
+		f.pending[txid] = txp
+	}
+	allocTxid, ok := f.allocs[p.id]
+	if ok {
+		delete(f.allocs, p.id)
+	} else if (p.flags & freelistPageFlag) != 0 {
+		// Freelist is always allocated by prior tx.
+		allocTxid = txid - 1
+	}
+
 	for id := p.id; id <= p.id+pgid(p.overflow); id++ {
 		// Verify that page is not already free.
 		if f.cache[id] {
 			panic(fmt.Sprintf("page %d already freed", id))
 		}
-
 		// Add to the freelist and cache.
-		ids = append(ids, id)
+		txp.ids = append(txp.ids, id)
+		txp.alloctx = append(txp.alloctx, allocTxid)
 		f.cache[id] = true
 	}
-	f.pending[txid] = ids
 }
 
 // release moves all page ids for a transaction id (or older) to the freelist.
 func (f *freelist) release(txid txid) {
 	m := make(pgids, 0)
-	for tid, ids := range f.pending {
+	for tid, txp := range f.pending {
 		if tid <= txid {
 			// Move transaction's pending pages to the available freelist.
 			// Don't remove from the cache since the page is still free.
-			m = append(m, ids...)
+			m = append(m, txp.ids...)
+			delete(f.pending, tid)
+		}
+	}
+	sort.Sort(m)
+	f.ids = pgids(f.ids).merge(m)
+}
+
+// releaseRange moves pending pages allocated within an extent [begin,end] to the free list.
+func (f *freelist) releaseRange(begin, end txid) {
+	if begin > end {
+		return
+	}
+	var m pgids
+	for tid, txp := range f.pending {
+		if tid < begin || tid > end {
+			continue
+		}
+		// Don't recompute freed pages if ranges haven't updated.
+		if txp.lastReleaseBegin == begin {
+			continue
+		}
+		for i := 0; i < len(txp.ids); i++ {
+			if atx := txp.alloctx[i]; atx < begin || atx > end {
+				continue
+			}
+			m = append(m, txp.ids[i])
+			txp.ids[i] = txp.ids[len(txp.ids)-1]
+			txp.ids = txp.ids[:len(txp.ids)-1]
+			txp.alloctx[i] = txp.alloctx[len(txp.alloctx)-1]
+			txp.alloctx = txp.alloctx[:len(txp.alloctx)-1]
+			i--
+		}
+		txp.lastReleaseBegin = begin
+		if len(txp.ids) == 0 {
 			delete(f.pending, tid)
 		}
 	}
@@ -142,12 +201,29 @@ func (f *freelist) release(txid txid) {
 // rollback removes the pages from a given pending tx.
 func (f *freelist) rollback(txid txid) {
 	// Remove page ids from cache.
-	for _, id := range f.pending[txid] {
-		delete(f.cache, id)
+	txp := f.pending[txid]
+	if txp == nil {
+		return
 	}
-
-	// Remove pages from pending list.
+	var m pgids
+	for i, pgid := range txp.ids {
+		delete(f.cache, pgid)
+		tx := txp.alloctx[i]
+		if tx == 0 {
+			continue
+		}
+		if tx != txid {
+			// Pending free aborted; restore page back to alloc list.
+			f.allocs[pgid] = tx
+		} else {
+			// Freed page was allocated by this txn; OK to throw away.
+			m = append(m, pgid)
+		}
+	}
+	// Remove pages from pending list and mark as free if allocated by txid.
 	delete(f.pending, txid)
+	sort.Sort(m)
+	f.ids = pgids(f.ids).merge(m)
 }
 
 // freed returns whether a given page is in the free list.
@@ -157,6 +233,9 @@ func (f *freelist) freed(pgid pgid) bool {
 
 // read initializes the freelist from a freelist page.
 func (f *freelist) read(p *page) {
+	if (p.flags & freelistPageFlag) == 0 {
+		panic(fmt.Sprintf("invalid freelist page: %d, page type is %s", p.id, p.typ()))
+	}
 	// If the page.count is at the max uint16 value (64k) then it's considered
 	// an overflow and the size of the freelist is stored as the first element.
 	idx, count := 0, int(p.count)
@@ -169,7 +248,7 @@ func (f *freelist) read(p *page) {
 	if count == 0 {
 		f.ids = nil
 	} else {
-		ids := ((*[maxAllocSize]pgid)(unsafe.Pointer(&p.ptr)))[idx:count]
+		ids := ((*[maxAllocSize]pgid)(unsafe.Pointer(&p.ptr)))[idx : idx+count]
 		f.ids = make([]pgid, len(ids))
 		copy(f.ids, ids)
 
@@ -181,27 +260,33 @@ func (f *freelist) read(p *page) {
 	f.reindex()
 }
 
+// read initializes the freelist from a given list of ids.
+func (f *freelist) readIDs(ids []pgid) {
+	f.ids = ids
+	f.reindex()
+}
+
 // write writes the page ids onto a freelist page. All free and pending ids are
 // saved to disk since in the event of a program crash, all pending ids will
 // become free.
 func (f *freelist) write(p *page) error {
 	// Combine the old free pgids and pgids waiting on an open transaction.
-	ids := f.all()
 
 	// Update the header flag.
 	p.flags |= freelistPageFlag
 
 	// The page.count can only hold up to 64k elements so if we overflow that
 	// number then we handle it by putting the size in the first element.
-	if len(ids) == 0 {
-		p.count = uint16(len(ids))
-	} else if len(ids) < 0xFFFF {
-		p.count = uint16(len(ids))
-		copy(((*[maxAllocSize]pgid)(unsafe.Pointer(&p.ptr)))[:], ids)
+	lenids := f.count()
+	if lenids == 0 {
+		p.count = uint16(lenids)
+	} else if lenids < 0xFFFF {
+		p.count = uint16(lenids)
+		f.copyall(((*[maxAllocSize]pgid)(unsafe.Pointer(&p.ptr)))[:])
 	} else {
 		p.count = 0xFFFF
-		((*[maxAllocSize]pgid)(unsafe.Pointer(&p.ptr)))[0] = pgid(len(ids))
-		copy(((*[maxAllocSize]pgid)(unsafe.Pointer(&p.ptr)))[1:], ids)
+		((*[maxAllocSize]pgid)(unsafe.Pointer(&p.ptr)))[0] = pgid(lenids)
+		f.copyall(((*[maxAllocSize]pgid)(unsafe.Pointer(&p.ptr)))[1:])
 	}
 
 	return nil
@@ -213,8 +298,8 @@ func (f *freelist) reload(p *page) {
 
 	// Build a cache of only pending pages.
 	pcache := make(map[pgid]bool)
-	for _, pendingIDs := range f.pending {
-		for _, pendingID := range pendingIDs {
+	for _, txp := range f.pending {
+		for _, pendingID := range txp.ids {
 			pcache[pendingID] = true
 		}
 	}
@@ -236,12 +321,12 @@ func (f *freelist) reload(p *page) {
 
 // reindex rebuilds the free cache based on available and pending free lists.
 func (f *freelist) reindex() {
-	f.cache = make(map[pgid]bool)
+	f.cache = make(map[pgid]bool, len(f.ids))
 	for _, id := range f.ids {
 		f.cache[id] = true
 	}
-	for _, pendingIDs := range f.pending {
-		for _, pendingID := range pendingIDs {
+	for _, txp := range f.pending {
+		for _, pendingID := range txp.ids {
 			f.cache[pendingID] = true
 		}
 	}

cmd/vendor/github.com/coreos/bbolt/node.go

@@ -365,7 +365,7 @@ func (n *node) spill() error {
 		}
 
 		// Allocate contiguous space for the node.
-		p, err := tx.allocate((node.size() / tx.db.pageSize) + 1)
+		p, err := tx.allocate((node.size() + tx.db.pageSize - 1) / tx.db.pageSize)
 		if err != nil {
 			return err
 		}

cmd/vendor/github.com/coreos/bbolt/page.go

@@ -145,12 +145,33 @@ func (a pgids) merge(b pgids) pgids {
 	// Return the opposite slice if one is nil.
 	if len(a) == 0 {
 		return b
-	} else if len(b) == 0 {
+	}
+	if len(b) == 0 {
 		return a
 	}
+	merged := make(pgids, len(a)+len(b))
+	mergepgids(merged, a, b)
+	return merged
+}
 
-	// Create a list to hold all elements from both lists.
-	merged := make(pgids, 0, len(a)+len(b))
+// mergepgids copies the sorted union of a and b into dst.
+// If dst is too small, it panics.
+func mergepgids(dst, a, b pgids) {
+	if len(dst) < len(a)+len(b) {
+		panic(fmt.Errorf("mergepgids bad len %d < %d + %d", len(dst), len(a), len(b)))
+	}
+	// Copy in the opposite slice if one is nil.
+	if len(a) == 0 {
+		copy(dst, b)
+		return
+	}
+	if len(b) == 0 {
+		copy(dst, a)
+		return
+	}
+
+	// Merged will hold all elements from both lists.
+	merged := dst[:0]
 
 	// Assign lead to the slice with a lower starting value, follow to the higher value.
 	lead, follow := a, b
@@ -172,7 +193,5 @@ func (a pgids) merge(b pgids) pgids {
 	}
 
 	// Append what's left in follow.
-	merged = append(merged, follow...)
-
-	return merged
+	_ = append(merged, follow...)
 }

cmd/vendor/github.com/coreos/bbolt/tx.go

@@ -126,10 +126,7 @@ func (tx *Tx) DeleteBucket(name []byte) error {
 // the error is returned to the caller.
 func (tx *Tx) ForEach(fn func(name []byte, b *Bucket) error) error {
 	return tx.root.ForEach(func(k, v []byte) error {
-		if err := fn(k, tx.root.Bucket(k)); err != nil {
-			return err
-		}
-		return nil
+		return fn(k, tx.root.Bucket(k))
 	})
 }
 
@@ -169,28 +166,18 @@ func (tx *Tx) Commit() error {
 	// Free the old root bucket.
 	tx.meta.root.root = tx.root.root
 
-	opgid := tx.meta.pgid
-
-	// Free the freelist and allocate new pages for it. This will overestimate
-	// the size of the freelist but not underestimate the size (which would be bad).
-	tx.db.freelist.free(tx.meta.txid, tx.db.page(tx.meta.freelist))
-	p, err := tx.allocate((tx.db.freelist.size() / tx.db.pageSize) + 1)
-	if err != nil {
-		tx.rollback()
-		return err
+	// Free the old freelist because commit writes out a fresh freelist.
+	if tx.meta.freelist != pgidNoFreelist {
+		tx.db.freelist.free(tx.meta.txid, tx.db.page(tx.meta.freelist))
 	}
-	if err := tx.db.freelist.write(p); err != nil {
-		tx.rollback()
-		return err
-	}
-	tx.meta.freelist = p.id
 
-	// If the high water mark has moved up then attempt to grow the database.
-	if tx.meta.pgid > opgid {
-		if err := tx.db.grow(int(tx.meta.pgid+1) * tx.db.pageSize); err != nil {
-			tx.rollback()
+	if !tx.db.NoFreelistSync {
+		err := tx.commitFreelist()
+		if err != nil {
 			return err
 		}
+	} else {
+		tx.meta.freelist = pgidNoFreelist
 	}
 
 	// Write dirty pages to disk.
@@ -235,6 +222,31 @@ func (tx *Tx) Commit() error {
 	return nil
 }
 
+func (tx *Tx) commitFreelist() error {
+	// Allocate new pages for the new free list. This will overestimate
+	// the size of the freelist but not underestimate the size (which would be bad).
+	opgid := tx.meta.pgid
+	p, err := tx.allocate((tx.db.freelist.size() / tx.db.pageSize) + 1)
+	if err != nil {
+		tx.rollback()
+		return err
+	}
+	if err := tx.db.freelist.write(p); err != nil {
+		tx.rollback()
+		return err
+	}
+	tx.meta.freelist = p.id
+	// If the high water mark has moved up then attempt to grow the database.
+	if tx.meta.pgid > opgid {
+		if err := tx.db.grow(int(tx.meta.pgid+1) * tx.db.pageSize); err != nil {
+			tx.rollback()
+			return err
+		}
+	}
+
+	return nil
+}
+
 // Rollback closes the transaction and ignores all previous updates. Read-only
 // transactions must be rolled back and not committed.
 func (tx *Tx) Rollback() error {
@@ -305,7 +317,11 @@ func (tx *Tx) WriteTo(w io.Writer) (n int64, err error) {
 	if err != nil {
 		return 0, err
 	}
-	defer func() { _ = f.Close() }()
+	defer func() {
+		if cerr := f.Close(); err == nil {
+			err = cerr
+		}
+	}()
 
 	// Generate a meta page. We use the same page data for both meta pages.
 	buf := make([]byte, tx.db.pageSize)
@@ -333,7 +349,7 @@ func (tx *Tx) WriteTo(w io.Writer) (n int64, err error) {
 	}
 
 	// Move past the meta pages in the file.
-	if _, err := f.Seek(int64(tx.db.pageSize*2), os.SEEK_SET); err != nil {
+	if _, err := f.Seek(int64(tx.db.pageSize*2), io.SeekStart); err != nil {
 		return n, fmt.Errorf("seek: %s", err)
 	}
 
@@ -344,7 +360,7 @@ func (tx *Tx) WriteTo(w io.Writer) (n int64, err error) {
 		return n, err
 	}
 
-	return n, f.Close()
+	return n, nil
 }
 
 // CopyFile copies the entire database to file at the given path.
@@ -379,9 +395,14 @@ func (tx *Tx) Check() <-chan error {
 }
 
 func (tx *Tx) check(ch chan error) {
+	// Force loading free list if opened in ReadOnly mode.
+	tx.db.loadFreelist()
+
 	// Check if any pages are double freed.
 	freed := make(map[pgid]bool)
-	for _, id := range tx.db.freelist.all() {
+	all := make([]pgid, tx.db.freelist.count())
+	tx.db.freelist.copyall(all)
+	for _, id := range all {
 		if freed[id] {
 			ch <- fmt.Errorf("page %d: already freed", id)
 		}
@@ -392,8 +413,10 @@ func (tx *Tx) check(ch chan error) {
 	reachable := make(map[pgid]*page)
 	reachable[0] = tx.page(0) // meta0
 	reachable[1] = tx.page(1) // meta1
-	for i := uint32(0); i <= tx.page(tx.meta.freelist).overflow; i++ {
-		reachable[tx.meta.freelist+pgid(i)] = tx.page(tx.meta.freelist)
+	if tx.meta.freelist != pgidNoFreelist {
+		for i := uint32(0); i <= tx.page(tx.meta.freelist).overflow; i++ {
+			reachable[tx.meta.freelist+pgid(i)] = tx.page(tx.meta.freelist)
+		}
 	}
 
 	// Recursively check buckets.
@@ -451,7 +474,7 @@ func (tx *Tx) checkBucket(b *Bucket, reachable map[pgid]*page, freed map[pgid]bo
 
 // allocate returns a contiguous block of memory starting at a given page.
 func (tx *Tx) allocate(count int) (*page, error) {
-	p, err := tx.db.allocate(count)
+	p, err := tx.db.allocate(tx.meta.txid, count)
 	if err != nil {
 		return nil, err
 	}

e2e/etcd_release_upgrade_test.go

@@ -23,6 +23,7 @@ import (
 
 	"github.com/coreos/etcd/pkg/fileutil"
 	"github.com/coreos/etcd/pkg/testutil"
+	"github.com/coreos/etcd/version"
 )
 
 // TestReleaseUpgrade ensures that changes to master branch does not affect
@@ -53,7 +54,7 @@ func TestReleaseUpgrade(t *testing.T) {
 	// so there's a window at boot time where it doesn't have V3rpcCapability enabled
 	// poll /version until etcdcluster is >2.3.x before making v3 requests
 	for i := 0; i < 7; i++ {
-		if err = cURLGet(epc, cURLReq{endpoint: "/version", expected: `"etcdcluster":"3.0`}); err != nil {
+		if err = cURLGet(epc, cURLReq{endpoint: "/version", expected: `"etcdcluster":"` + version.Cluster(version.Version)}); err != nil {
 			t.Logf("#%d: v3 is not ready yet (%v)", i, err)
 			time.Sleep(time.Second)
 			continue

etcdctl/ctlv3/command/snapshot_command.go

@@ -27,7 +27,7 @@ import (
 	"reflect"
 	"strings"
 
-	"github.com/boltdb/bolt"
+	bolt "github.com/coreos/bbolt"
 	"github.com/coreos/etcd/etcdserver"
 	"github.com/coreos/etcd/etcdserver/etcdserverpb"
 	"github.com/coreos/etcd/etcdserver/membership"

etcdserver/server.go

@@ -608,7 +608,7 @@ type raftReadyHandler struct {
 }
 
 func (s *EtcdServer) run() {
-	snap, err := s.r.raftStorage.Snapshot()
+	snapshot, err := s.r.raftStorage.Snapshot()
 	if err != nil {
 		plog.Panicf("get snapshot from raft storage error: %v", err)
 	}
@@ -666,10 +666,10 @@ func (s *EtcdServer) run() {
 	// asynchronously accept apply packets, dispatch progress in-order
 	sched := schedule.NewFIFOScheduler()
 	ep := etcdProgress{
-		confState: snap.Metadata.ConfState,
-		snapi:     snap.Metadata.Index,
-		appliedt:  snap.Metadata.Term,
-		appliedi:  snap.Metadata.Index,
+		confState: snapshot.Metadata.ConfState,
+		snapi:     snapshot.Metadata.Index,
+		appliedt:  snapshot.Metadata.Term,
+		appliedi:  snapshot.Metadata.Index,
 	}
 
 	defer func() {

glide.lock

@@ -1,5 +1,5 @@
-hash: ca3c895fa60c9ca9f53408202fb7643705f9960212d342967ed0da8e93606cc4
-updated: 2017-01-18T10:26:48.990115455-08:00
+hash: 8a2307db167c2f9bc511956fbec0a4400bc38db103603fc6a87bb179ac853bf5
+updated: 2017-11-21T11:36:36.822426751-08:00
 imports:
 - name: github.com/beorn7/perks
   version: 4c0e84591b9aa9e6dcfdf3e020114cd81f89d5f9
@@ -7,10 +7,10 @@ imports:
   - quantile
 - name: github.com/bgentry/speakeasy
   version: 36e9cfdd690967f4f690c6edcc9ffacd006014a0
-- name: github.com/boltdb/bolt
-  version: 583e8937c61f1af6513608ccc75c97b6abdf4ff9
 - name: github.com/cockroachdb/cmux
   version: 112f0506e7743d64a6eb8fedbcff13d9979bbf92
+- name: github.com/coreos/bbolt
+  version: 32c383e75ce054674c53b5a07e55de85332aee14
 - name: github.com/coreos/go-semver
   version: 568e959cd89871e61434c1143528d9162da89ef2
   subpackages:
@@ -74,7 +74,7 @@ imports:
   subpackages:
   - prometheus
 - name: github.com/prometheus/client_model
-  version: fa8ad6fec33561be4280a8f0514318c79d7f6cb6
+  version: 6f3806018612930941127f2a7c6c453ba2c527d2
   subpackages:
   - go
 - name: github.com/prometheus/common

glide.yaml

@@ -2,8 +2,8 @@ package: github.com/coreos/etcd
 import:
 - package: github.com/bgentry/speakeasy
   version: 36e9cfdd690967f4f690c6edcc9ffacd006014a0
-- package: github.com/boltdb/bolt
-  version: v1.3.0
+- package: github.com/coreos/bbolt
+  version: v1.3.1-coreos.5
 - package: github.com/cockroachdb/cmux
   version: 112f0506e7743d64a6eb8fedbcff13d9979bbf92
 - package: github.com/coreos/go-semver
@@ -48,9 +48,21 @@ import:
 - package: github.com/olekukonko/tablewriter
   version: cca8bbc0798408af109aaaa239cbd2634846b340
 - package: github.com/prometheus/client_golang
-  version: v0.8.0
+  version: c5b7fccd204277076155f10851dad72b76a49317
   subpackages:
   - prometheus
+- package: github.com/prometheus/client_model
+  version: 6f3806018612930941127f2a7c6c453ba2c527d2
+  subpackages:
+  - go
+- package: github.com/prometheus/common
+  version: 195bde7883f7c39ea62b0d92ab7359b5327065cb
+  subpackages:
+  - expfmt
+  - internal/bitbucket.org/ww/goautoneg
+  - model
+- package: github.com/prometheus/procfs
+  version: fcdb11ccb4389efb1b210b7ffb623ab71c5fdd60
 - package: github.com/spf13/cobra
   version: 1c44ec8d3f1552cac48999f9306da23c4d8a288b
 - package: github.com/spf13/pflag
@@ -102,4 +114,22 @@ import:
   subpackages:
   - assert
 - package: github.com/karlseguin/ccache
-  version: v2.0.2
+  version: a2d62155777b39595c825ed3824279e642a5db3c
+- package: github.com/beorn7/perks
+  version: 4c0e84591b9aa9e6dcfdf3e020114cd81f89d5f9
+- package: github.com/cpuguy83/go-md2man
+  version: a65d4d2de4d5f7c74868dfa9b202a3c8be315aaa
+- package: github.com/matttproud/golang_protobuf_extensions
+  version: c12348ce28de40eed0136aa2b644d0ee0650e56c
+  subpackages:
+  - pbutil
+- package: github.com/russross/blackfriday
+  version: 5f33e7b7878355cd2b7e6b8eefc48a5472c69f70
+- package: github.com/mattn/go-runewidth
+  version: 737072b4e32b7a5018b4a7125da8d12de90e8045
+- package: github.com/shurcooL/sanitized_anchor_name
+  version: 1dba4b3954bc059efc3991ec364f9f9a35f597d2
+- package: golang.org/x/sys
+  version: 478fcf54317e52ab69f40bb4c7a1520288d7f7ea
+  subpackages:
+  - unix

hack/scripts-dev/Makefile

@@ -0,0 +1,243 @@
+# run from repository root
+#
+# Example:
+#   make clean -f ./hack/scripts-dev/Makefile
+#   make build -f ./hack/scripts-dev/Makefile
+
+.PHONY: build
+build:
+	GO_BUILD_FLAGS="-v" ./build
+	./bin/etcd --version
+	ETCDCTL_API=3 ./bin/etcdctl version
+
+clean:
+	rm -f ./codecov
+	rm -rf ./covdir
+	rm -f ./*.log
+	rm -f ./bin/Dockerfile-release
+	rm -rf ./bin/*.etcd
+	rm -rf ./gopath
+	rm -rf ./release
+	rm -f ./integration/127.0.0.1:* ./integration/localhost:*
+	rm -f ./clientv3/integration/127.0.0.1:* ./clientv3/integration/localhost:*
+	rm -f ./clientv3/ordering/127.0.0.1:* ./clientv3/ordering/localhost:*
+
+_GO_VERSION = 1.9.2
+ifdef GO_VERSION
+	_GO_VERSION = $(GO_VERSION)
+endif
+
+# Example:
+#   GO_VERSION=1.8.5 make build-docker-test -f ./hack/scripts-dev/Makefile
+#   make build-docker-test -f ./hack/scripts-dev/Makefile
+#   gcloud docker -- login -u _json_key -p "$(cat /etc/gcp-key-etcd.json)" https://gcr.io
+#   GO_VERSION=1.8.5 make push-docker-test -f ./hack/scripts-dev/Makefile
+#   make push-docker-test -f ./hack/scripts-dev/Makefile
+#   gsutil -m acl ch -u allUsers:R -r gs://artifacts.etcd-development.appspot.com
+#   GO_VERSION=1.8.5 make pull-docker-test -f ./hack/scripts-dev/Makefile
+#   make pull-docker-test -f ./hack/scripts-dev/Makefile
+
+build-docker-test:
+	$(info GO_VERSION: $(_GO_VERSION))
+	@cat ./Dockerfile-test | sed s/REPLACE_ME_GO_VERSION/$(_GO_VERSION)/ \
+	  > ./.Dockerfile-test
+	docker build \
+	  --tag gcr.io/etcd-development/etcd-test:go$(_GO_VERSION) \
+	  --file ./.Dockerfile-test .
+
+push-docker-test:
+	$(info GO_VERSION: $(_GO_VERSION))
+	gcloud docker -- push gcr.io/etcd-development/etcd-test:go$(_GO_VERSION)
+
+pull-docker-test:
+	$(info GO_VERSION: $(_GO_VERSION))
+	docker pull gcr.io/etcd-development/etcd-test:go$(_GO_VERSION)
+
+compile-with-docker-test:
+	$(info GO_VERSION: $(_GO_VERSION))
+	docker run \
+	  --rm \
+	  --volume=`pwd`/:/etcd \
+	  gcr.io/etcd-development/etcd-test:go$(_GO_VERSION) \
+	  /bin/bash -c "cd /etcd && GO_BUILD_FLAGS=-v ./build && ./bin/etcd --version"
+
+# Local machine:
+#   TEST_OPTS="PASSES='fmt'" make test -f ./hack/scripts-dev/Makefile
+#   TEST_OPTS="PASSES='fmt bom dep compile build unit'" make test -f ./hack/scripts-dev/Makefile
+#   TEST_OPTS="RELEASE_TEST=y INTEGRATION=y PASSES='build unit release integration_e2e functional'" make test -f ./hack/scripts-dev/Makefile
+#   TEST_OPTS="PASSES='build grpcproxy'" make test -f ./hack/scripts-dev/Makefile
+#
+# Example (test with docker):
+#   make pull-docker-test -f ./hack/scripts-dev/Makefile
+#   TEST_OPTS="PASSES='fmt'" make docker-test -f ./hack/scripts-dev/Makefile
+#   TEST_OPTS="VERBOSE=2 PASSES='unit'" make docker-test -f ./hack/scripts-dev/Makefile
+#
+# Travis CI (test with docker):
+#   TEST_OPTS="PASSES='fmt bom dep compile build unit'" make docker-test -f ./hack/scripts-dev/Makefile
+#
+# Semaphore CI (test with docker):
+#   TEST_OPTS="RELEASE_TEST=y INTEGRATION=y PASSES='build unit release integration_e2e functional'" make docker-test -f ./hack/scripts-dev/Makefile
+#   TEST_OPTS="GOARCH=386 PASSES='build unit integration_e2e'" make docker-test -f ./hack/scripts-dev/Makefile
+#
+# grpc-proxy tests (test with docker):
+#   TEST_OPTS="PASSES='build grpcproxy'" make docker-test -f ./hack/scripts-dev/Makefile
+
+TEST_SUFFIX = $(shell date +%s | base64 | head -c 15)
+
+_TEST_OPTS = "PASSES='unit'"
+ifdef TEST_OPTS
+	_TEST_OPTS = $(TEST_OPTS)
+endif
+
+.PHONY: test
+test:
+	$(info TEST_OPTS: $(_TEST_OPTS))
+	$(info log-file: test-$(TEST_SUFFIX).log)
+	$(_TEST_OPTS) ./test 2>&1 | tee test-$(TEST_SUFFIX).log
+	! egrep "(--- FAIL:|leak)" -A10 -B50 test-$(TEST_SUFFIX).log
+
+docker-test:
+	$(info GO_VERSION: $(_GO_VERSION))
+	$(info TEST_OPTS: $(_TEST_OPTS))
+	$(info log-file: test-$(TEST_SUFFIX).log)
+	docker run \
+	  --rm \
+	  --volume=`pwd`:/go/src/github.com/coreos/etcd \
+	  gcr.io/etcd-development/etcd-test:go$(_GO_VERSION) \
+	  /bin/bash -c "$(_TEST_OPTS) ./test 2>&1 | tee test-$(TEST_SUFFIX).log"
+	! egrep "(--- FAIL:|leak)" -A10 -B50 test-$(TEST_SUFFIX).log
+
+docker-test-coverage:
+	$(info GO_VERSION: $(_GO_VERSION))
+	$(info log-file: docker-test-coverage-$(TEST_SUFFIX).log)
+	docker run \
+	  --rm \
+	  --volume=`pwd`:/go/src/github.com/coreos/etcd \
+	  gcr.io/etcd-development/etcd-test:go$(_GO_VERSION) \
+	  /bin/bash -c "COVERDIR=covdir PASSES='build build_cov cov' ./test 2>&1 | tee docker-test-coverage-$(TEST_SUFFIX).log && /codecov -t 6040de41-c073-4d6f-bbf8-d89256ef31e1"
+	! egrep "(--- FAIL:|leak)" -A10 -B50 docker-test-coverage-$(TEST_SUFFIX).log
+
+# build release container image with Linux
+_ETCD_VERSION ?= $(shell git rev-parse --short HEAD || echo "GitNotFound")
+ifdef ETCD_VERSION
+	_ETCD_VERSION = $(ETCD_VERSION)
+endif
+
+# Example:
+#   ETCD_VERSION=v3.3.0-test.0 make build-docker-release-master -f ./hack/scripts-dev/Makefile
+#   ETCD_VERSION=v3.3.0-test.0 make push-docker-release-master -f ./hack/scripts-dev/Makefile
+#   gsutil -m acl ch -u allUsers:R -r gs://artifacts.etcd-development.appspot.com
+
+build-docker-release-master: compile-with-docker-test
+	$(info ETCD_VERSION: $(_ETCD_VERSION))
+	cp ./Dockerfile-release ./bin/Dockerfile-release
+	docker build \
+	  --tag gcr.io/etcd-development/etcd:$(_ETCD_VERSION) \
+	  --file ./bin/Dockerfile-release \
+	  ./bin
+	rm -f ./bin/Dockerfile-release
+
+	docker run \
+	  --rm \
+	  gcr.io/etcd-development/etcd:$(_ETCD_VERSION) \
+	  /bin/sh -c "/usr/local/bin/etcd --version && ETCDCTL_API=3 /usr/local/bin/etcdctl version"
+
+push-docker-release-master:
+	$(info ETCD_VERSION: $(_ETCD_VERSION))
+	gcloud docker -- push gcr.io/etcd-development/etcd:$(_ETCD_VERSION)
+
+# Example:
+#   make build-docker-test -f ./hack/scripts-dev/Makefile
+#   make compile-with-docker-test -f ./hack/scripts-dev/Makefile
+#   make build-docker-dns-test -f ./hack/scripts-dev/Makefile
+#   gcloud docker -- login -u _json_key -p "$(cat /etc/gcp-key-etcd.json)" https://gcr.io
+#   make push-docker-dns-test -f ./hack/scripts-dev/Makefile
+#   gsutil -m acl ch -u allUsers:R -r gs://artifacts.etcd-development.appspot.com
+#   make pull-docker-dns-test -f ./hack/scripts-dev/Makefile
+#   make docker-dns-test-run -f ./hack/scripts-dev/Makefile
+
+# build base container image for DNS testing
+build-docker-dns-test:
+	$(info GO_VERSION: $(_GO_VERSION))
+	@cat ./hack/scripts-dev/docker-dns/Dockerfile | sed s/REPLACE_ME_GO_VERSION/$(_GO_VERSION)/ \
+	  > ./hack/scripts-dev/docker-dns/.Dockerfile
+
+	docker build \
+	  --tag gcr.io/etcd-development/etcd-dns-test:go$(_GO_VERSION) \
+	  --file ./hack/scripts-dev/docker-dns/.Dockerfile \
+	  ./hack/scripts-dev/docker-dns
+
+	docker run \
+	  --rm \
+	  --dns 127.0.0.1 \
+	  gcr.io/etcd-development/etcd-dns-test:go$(_GO_VERSION) \
+	  /bin/bash -c "/etc/init.d/bind9 start && cat /dev/null >/etc/hosts && dig etcd.local"
+
+push-docker-dns-test:
+	$(info GO_VERSION: $(_GO_VERSION))
+	gcloud docker -- push gcr.io/etcd-development/etcd-dns-test:go$(_GO_VERSION)
+
+pull-docker-dns-test:
+	$(info GO_VERSION: $(_GO_VERSION))
+	docker pull gcr.io/etcd-development/etcd-dns-test:go$(_GO_VERSION)
+
+# run DNS tests inside container
+docker-dns-test-run:
+	$(info GO_VERSION: $(_GO_VERSION))
+	docker run \
+	  --rm \
+	  --tty \
+	  --dns 127.0.0.1 \
+	  --volume=`pwd`/bin:/etcd \
+	  --volume=`pwd`/integration/fixtures:/certs \
+	  gcr.io/etcd-development/etcd-dns-test:go$(_GO_VERSION) \
+	  /bin/bash -c "cd /etcd && /run.sh && rm -rf m*.etcd"
+
+# Example:
+#   make build-docker-test -f ./hack/scripts-dev/Makefile
+#   make compile-with-docker-test -f ./hack/scripts-dev/Makefile
+#   make build-docker-dns-srv-test -f ./hack/scripts-dev/Makefile
+#   gcloud docker -- login -u _json_key -p "$(cat /etc/gcp-key-etcd.json)" https://gcr.io
+#   make push-docker-dns-srv-test -f ./hack/scripts-dev/Makefile
+#   gsutil -m acl ch -u allUsers:R -r gs://artifacts.etcd-development.appspot.com
+#   make pull-docker-dns-srv-test -f ./hack/scripts-dev/Makefile
+#   make docker-dns-srv-test-run -f ./hack/scripts-dev/Makefile
+
+# build base container image for DNS/SRV testing
+build-docker-dns-srv-test:
+	$(info GO_VERSION: $(_GO_VERSION))
+	@cat ./hack/scripts-dev/docker-dns-srv/Dockerfile | sed s/REPLACE_ME_GO_VERSION/$(_GO_VERSION)/ \
+	  > ./hack/scripts-dev/docker-dns-srv/.Dockerfile
+
+	docker build \
+	  --tag gcr.io/etcd-development/etcd-dns-srv-test:go$(_GO_VERSION) \
+	  --file ./hack/scripts-dev/docker-dns-srv/.Dockerfile \
+	  ./hack/scripts-dev/docker-dns-srv
+
+	docker run \
+	  --rm \
+	  --dns 127.0.0.1 \
+	  gcr.io/etcd-development/etcd-dns-srv-test:go$(_GO_VERSION) \
+	  /bin/bash -c "/etc/init.d/bind9 start && cat /dev/null >/etc/hosts && dig +noall +answer SRV _etcd-client-ssl._tcp.etcd.local && dig +noall +answer SRV _etcd-server-ssl._tcp.etcd.local && dig +noall +answer m1.etcd.local m2.etcd.local m3.etcd.local"
+
+push-docker-dns-srv-test:
+	$(info GO_VERSION: $(_GO_VERSION))
+	gcloud docker -- push gcr.io/etcd-development/etcd-dns-srv-test:go$(_GO_VERSION)
+
+pull-docker-dns-srv-test:
+	$(info GO_VERSION: $(_GO_VERSION))
+	docker pull gcr.io/etcd-development/etcd-dns-srv-test:go$(_GO_VERSION)
+
+# run DNS/SRV tests inside container
+docker-dns-srv-test-run:
+	$(info GO_VERSION: $(_GO_VERSION))
+	docker run \
+	  --rm \
+	  --tty \
+	  --dns 127.0.0.1 \
+	  --volume=`pwd`/bin:/etcd \
+	  --volume=`pwd`/integration/fixtures:/certs \
+	  gcr.io/etcd-development/etcd-dns-srv-test:go$(_GO_VERSION) \
+	  /bin/bash -c "cd /etcd && /run.sh && rm -rf m*.etcd"
+
+# TODO: add DNS integration tests

hack/scripts-dev/README

@@ -0,0 +1,2 @@
+
+scripts for etcd development

hack/scripts-dev/docker-dns-srv/.Dockerfile

@@ -0,0 +1,48 @@
+FROM ubuntu:16.10
+
+RUN rm /bin/sh && ln -s /bin/bash /bin/sh
+RUN echo 'debconf debconf/frontend select Noninteractive' | debconf-set-selections
+
+RUN apt-get -y update \
+  && apt-get -y install \
+  build-essential \
+  gcc \
+  apt-utils \
+  pkg-config \
+  software-properties-common \
+  apt-transport-https \
+  libssl-dev \
+  sudo \
+  bash \
+  curl \
+  tar \
+  git \
+  netcat \
+  bind9 \
+  dnsutils \
+  && apt-get -y update \
+  && apt-get -y upgrade \
+  && apt-get -y autoremove \
+  && apt-get -y autoclean
+
+ENV GOROOT /usr/local/go
+ENV GOPATH /go
+ENV PATH ${GOPATH}/bin:${GOROOT}/bin:${PATH}
+ENV GO_VERSION 1.9.2
+ENV GO_DOWNLOAD_URL https://storage.googleapis.com/golang
+RUN rm -rf ${GOROOT} \
+  && curl -s ${GO_DOWNLOAD_URL}/go${GO_VERSION}.linux-amd64.tar.gz | tar -v -C /usr/local/ -xz \
+  && mkdir -p ${GOPATH}/src ${GOPATH}/bin \
+  && go version
+
+RUN mkdir -p /var/bind /etc/bind
+RUN chown root:bind /var/bind /etc/bind
+ADD Procfile /Procfile
+ADD run.sh /run.sh
+
+ADD named.conf etcd.zone rdns.zone /etc/bind/
+RUN chown root:bind /etc/bind/named.conf /etc/bind/etcd.zone /etc/bind/rdns.zone
+ADD resolv.conf /etc/resolv.conf
+
+RUN go get github.com/mattn/goreman
+CMD ["/run.sh"]

hack/scripts-dev/docker-dns-srv/Dockerfile

@@ -0,0 +1,48 @@
+FROM ubuntu:16.10
+
+RUN rm /bin/sh && ln -s /bin/bash /bin/sh
+RUN echo 'debconf debconf/frontend select Noninteractive' | debconf-set-selections
+
+RUN apt-get -y update \
+  && apt-get -y install \
+  build-essential \
+  gcc \
+  apt-utils \
+  pkg-config \
+  software-properties-common \
+  apt-transport-https \
+  libssl-dev \
+  sudo \
+  bash \
+  curl \
+  tar \
+  git \
+  netcat \
+  bind9 \
+  dnsutils \
+  && apt-get -y update \
+  && apt-get -y upgrade \
+  && apt-get -y autoremove \
+  && apt-get -y autoclean
+
+ENV GOROOT /usr/local/go
+ENV GOPATH /go
+ENV PATH ${GOPATH}/bin:${GOROOT}/bin:${PATH}
+ENV GO_VERSION REPLACE_ME_GO_VERSION
+ENV GO_DOWNLOAD_URL https://storage.googleapis.com/golang
+RUN rm -rf ${GOROOT} \
+  && curl -s ${GO_DOWNLOAD_URL}/go${GO_VERSION}.linux-amd64.tar.gz | tar -v -C /usr/local/ -xz \
+  && mkdir -p ${GOPATH}/src ${GOPATH}/bin \
+  && go version
+
+RUN mkdir -p /var/bind /etc/bind
+RUN chown root:bind /var/bind /etc/bind
+ADD Procfile /Procfile
+ADD run.sh /run.sh
+
+ADD named.conf etcd.zone rdns.zone /etc/bind/
+RUN chown root:bind /etc/bind/named.conf /etc/bind/etcd.zone /etc/bind/rdns.zone
+ADD resolv.conf /etc/resolv.conf
+
+RUN go get github.com/mattn/goreman
+CMD ["/run.sh"]

hack/scripts-dev/docker-dns-srv/Procfile

@@ -0,0 +1,5 @@
+etcd1: ./etcd --name m1 --listen-client-urls https://127.0.0.1:2379 --advertise-client-urls https://m1.etcd.local:2379 --listen-peer-urls https://127.0.0.1:2380 --initial-advertise-peer-urls=https://m1.etcd.local:2380 --initial-cluster-token tkn --discovery-srv=etcd.local --initial-cluster-state new --peer-cert-file=/certs/server-wildcard.crt --peer-key-file=/certs/server-wildcard.key.insecure --peer-trusted-ca-file=/certs/ca.crt --peer-client-cert-auth --cert-file=/certs/server-wildcard.crt --key-file=/certs/server-wildcard.key.insecure --trusted-ca-file=/certs/ca.crt --client-cert-auth
+
+etcd2: ./etcd --name m2 --listen-client-urls https://127.0.0.1:22379 --advertise-client-urls https://m2.etcd.local:22379 --listen-peer-urls https://127.0.0.1:22380 --initial-advertise-peer-urls=https://m2.etcd.local:22380 --initial-cluster-token tkn --discovery-srv=etcd.local --initial-cluster-state new --peer-cert-file=/certs/server-wildcard.crt --peer-key-file=/certs/server-wildcard.key.insecure --peer-trusted-ca-file=/certs/ca.crt --peer-client-cert-auth --cert-file=/certs/server-wildcard.crt --key-file=/certs/server-wildcard.key.insecure --trusted-ca-file=/certs/ca.crt --client-cert-auth
+
+etcd3: ./etcd --name m3 --listen-client-urls https://127.0.0.1:32379 --advertise-client-urls https://m3.etcd.local:32379 --listen-peer-urls https://127.0.0.1:32380 --initial-advertise-peer-urls=https://m3.etcd.local:32380 --initial-cluster-token tkn --discovery-srv=etcd.local --initial-cluster-state new --peer-cert-file=/certs/server-wildcard.crt --peer-key-file=/certs/server-wildcard.key.insecure --peer-trusted-ca-file=/certs/ca.crt --peer-client-cert-auth --cert-file=/certs/server-wildcard.crt --key-file=/certs/server-wildcard.key.insecure --trusted-ca-file=/certs/ca.crt --client-cert-auth

hack/scripts-dev/docker-dns-srv/etcd.zone

@@ -0,0 +1,16 @@
+etcd.local.	IN	SOA	bindhostname. admin.etcd.local. (
+1452607488
+10800
+3600
+604800
+38400 )
+etcd.local.	IN	NS	bindhostname.
+m1.etcd.local.	300	IN	A	127.0.0.1
+m2.etcd.local.	300	IN	A	127.0.0.1
+m3.etcd.local.	300	IN	A	127.0.0.1
+_etcd-client-ssl._tcp	300	IN	SRV	0 0 2379 m1.etcd.local.
+_etcd-client-ssl._tcp	300	IN	SRV	0 0 22379 m2.etcd.local.
+_etcd-client-ssl._tcp	300	IN	SRV	0 0 32379 m3.etcd.local.
+_etcd-server-ssl._tcp	300	IN	SRV	0 0 2380 m1.etcd.local.
+_etcd-server-ssl._tcp	300	IN	SRV	0 0 22380 m2.etcd.local.
+_etcd-server-ssl._tcp	300	IN	SRV	0 0 32380 m3.etcd.local.

hack/scripts-dev/docker-dns-srv/named.conf

@@ -0,0 +1,23 @@
+options {
+        directory "/var/bind";
+        listen-on { 127.0.0.1; };
+        listen-on-v6 { none; };
+        allow-transfer {
+                none;
+        };
+        // If you have problems and are behind a firewall:
+        query-source address * port 53;
+        pid-file "/var/run/named/named.pid";
+        allow-recursion { none; };
+        recursion no;
+};
+
+zone "etcd.local" IN {
+      type master;
+      file "/etc/bind/etcd.zone";
+};
+
+zone "0.0.127.in-addr.arpa" {
+      type master;
+      file "/etc/bind/rdns.zone";
+};

hack/scripts-dev/docker-dns-srv/rdns.zone

@@ -0,0 +1,13 @@
+$TTL    86400
+@   IN  SOA  etcdns.local. root.etcdns.local. (
+             100500     ; Serial
+             604800     ; Refresh
+              86400     ; Retry
+            2419200     ; Expire
+              86400 )   ; Negative Cache TTL
+    IN  NS  ns.etcdns.local.
+    IN  A   127.0.0.1
+
+1 IN PTR m1.etcd.local.
+1 IN PTR m2.etcd.local.
+1 IN PTR m3.etcd.local.

hack/scripts-dev/docker-dns-srv/resolv.conf

@@ -0,0 +1 @@
+nameserver 127.0.0.1

hack/scripts-dev/docker-dns-srv/run.sh

@@ -0,0 +1,16 @@
+#!/bin/sh
+
+/etc/init.d/bind9 start
+
+# get rid of hosts so go lookup won't resolve 127.0.0.1 to localhost
+cat /dev/null >/etc/hosts
+
+goreman -f /Procfile start &
+sleep 7s
+
+ETCDCTL_API=3 ./etcdctl \
+  --cacert=/certs/ca.crt \
+  --cert=/certs/server-wildcard.crt \
+  --key=/certs//server-wildcard.key.insecure \
+  --discovery-srv etcd.local \
+  put foo bar

hack/scripts-dev/docker-dns/Dockerfile

@@ -0,0 +1,48 @@
+FROM ubuntu:16.10
+
+RUN rm /bin/sh && ln -s /bin/bash /bin/sh
+RUN echo 'debconf debconf/frontend select Noninteractive' | debconf-set-selections
+
+RUN apt-get -y update \
+  && apt-get -y install \
+  build-essential \
+  gcc \
+  apt-utils \
+  pkg-config \
+  software-properties-common \
+  apt-transport-https \
+  libssl-dev \
+  sudo \
+  bash \
+  curl \
+  tar \
+  git \
+  netcat \
+  bind9 \
+  dnsutils \
+  && apt-get -y update \
+  && apt-get -y upgrade \
+  && apt-get -y autoremove \
+  && apt-get -y autoclean
+
+ENV GOROOT /usr/local/go
+ENV GOPATH /go
+ENV PATH ${GOPATH}/bin:${GOROOT}/bin:${PATH}
+ENV GO_VERSION REPLACE_ME_GO_VERSION
+ENV GO_DOWNLOAD_URL https://storage.googleapis.com/golang
+RUN rm -rf ${GOROOT} \
+  && curl -s ${GO_DOWNLOAD_URL}/go${GO_VERSION}.linux-amd64.tar.gz | tar -v -C /usr/local/ -xz \
+  && mkdir -p ${GOPATH}/src ${GOPATH}/bin \
+  && go version
+
+RUN mkdir -p /var/bind /etc/bind
+RUN chown root:bind /var/bind /etc/bind
+ADD Procfile.tls /Procfile.tls
+ADD run.sh /run.sh
+
+ADD named.conf etcd.zone rdns.zone /etc/bind/
+RUN chown root:bind /etc/bind/named.conf /etc/bind/etcd.zone /etc/bind/rdns.zone
+ADD resolv.conf /etc/resolv.conf
+
+RUN go get github.com/mattn/goreman
+CMD ["/run.sh"]

hack/scripts-dev/docker-dns/Procfile.tls

@@ -0,0 +1,6 @@
+# Use goreman to run `go get github.com/mattn/goreman`
+etcd1: ./etcd --name m1 --listen-client-urls https://127.0.0.1:2379 --advertise-client-urls https://m1.etcd.local:2379 --listen-peer-urls https://127.0.0.1:2380 --initial-advertise-peer-urls=https://m1.etcd.local:2380 --initial-cluster-token tkn --initial-cluster=m1=https://m1.etcd.local:2380,m2=https://m2.etcd.local:22380,m3=https://m3.etcd.local:32380 --initial-cluster-state new --peer-cert-file=/certs/server-wildcard.crt --peer-key-file=/certs/server-wildcard.key.insecure --peer-trusted-ca-file=/certs/ca.crt --peer-client-cert-auth --cert-file=/certs/server-wildcard.crt --key-file=/certs/server-wildcard.key.insecure --trusted-ca-file=/certs/ca.crt --client-cert-auth
+
+etcd2: ./etcd --name m2 --listen-client-urls https://127.0.0.1:22379 --advertise-client-urls https://m2.etcd.local:22379 --listen-peer-urls https://127.0.0.1:22380 --initial-advertise-peer-urls=https://m2.etcd.local:22380 --initial-cluster-token tkn --initial-cluster=m1=https://m1.etcd.local:2380,m2=https://m2.etcd.local:22380,m3=https://m3.etcd.local:32380 --initial-cluster-state new --peer-cert-file=/certs/server-wildcard.crt --peer-key-file=/certs/server-wildcard.key.insecure --peer-trusted-ca-file=/certs/ca.crt --peer-client-cert-auth --cert-file=/certs/server-wildcard.crt --key-file=/certs/server-wildcard.key.insecure --trusted-ca-file=/certs/ca.crt --client-cert-auth
+
+etcd3: ./etcd --name m3 --listen-client-urls https://127.0.0.1:32379 --advertise-client-urls https://m3.etcd.local:32379 --listen-peer-urls https://127.0.0.1:32380 --initial-advertise-peer-urls=https://m3.etcd.local:32380 --initial-cluster-token tkn --initial-cluster=m1=https://m1.etcd.local:2380,m2=https://m2.etcd.local:22380,m3=https://m3.etcd.local:32380 --initial-cluster-state new --peer-cert-file=/certs/server-wildcard.crt --peer-key-file=/certs/server-wildcard.key.insecure --peer-trusted-ca-file=/certs/ca.crt --peer-client-cert-auth --cert-file=/certs/server-wildcard.crt --key-file=/certs/server-wildcard.key.insecure --trusted-ca-file=/certs/ca.crt --client-cert-auth

hack/scripts-dev/docker-dns/etcd.zone

@@ -0,0 +1,14 @@
+$TTL    86400
+@   IN  SOA  etcdns.local. root.etcdns.local. (
+             100500     ; Serial
+             604800     ; Refresh
+              86400     ; Retry
+            2419200     ; Expire
+              86400 )   ; Negative Cache TTL
+    IN  NS  ns.etcdns.local.
+    IN  A   127.0.0.1
+
+ns IN A 127.0.0.1
+m1 IN A 127.0.0.1
+m2 IN A 127.0.0.1
+m3 IN A 127.0.0.1

hack/scripts-dev/docker-dns/named.conf

@@ -0,0 +1,23 @@
+options {
+        directory "/var/bind";
+        listen-on { 127.0.0.1; };
+        listen-on-v6 { none; };
+        allow-transfer {
+                none;
+        };
+        // If you have problems and are behind a firewall:
+        query-source address * port 53;
+        pid-file "/var/run/named/named.pid";
+        allow-recursion { none; };
+        recursion no;
+};
+
+zone "etcd.local" IN {
+      type master;
+      file "/etc/bind/etcd.zone";
+};
+
+zone "0.0.127.in-addr.arpa" {
+      type master;
+      file "/etc/bind/rdns.zone";
+};

hack/scripts-dev/docker-dns/rdns.zone

@@ -0,0 +1,13 @@
+$TTL    86400
+@   IN  SOA  etcdns.local. root.etcdns.local. (
+             100500     ; Serial
+             604800     ; Refresh
+              86400     ; Retry
+            2419200     ; Expire
+              86400 )   ; Negative Cache TTL
+    IN  NS  ns.etcdns.local.
+    IN  A   127.0.0.1
+
+1 IN PTR m1.etcd.local.
+1 IN PTR m2.etcd.local.
+1 IN PTR m3.etcd.local.

hack/scripts-dev/docker-dns/resolv.conf

@@ -0,0 +1 @@
+nameserver 127.0.0.1

hack/scripts-dev/docker-dns/run.sh

@@ -0,0 +1,16 @@
+#!/bin/sh
+
+/etc/init.d/bind9 start
+
+# get rid of hosts so go lookup won't resolve 127.0.0.1 to localhost
+cat /dev/null >/etc/hosts
+
+goreman -f /Procfile.tls start &
+sleep 7s
+
+ETCDCTL_API=3 ./etcdctl \
+  --cacert=/certs/ca.crt \
+  --cert=/certs/server-wildcard.crt \
+  --key=/certs//server-wildcard.key.insecure \
+  --endpoints=https://m1.etcd.local:2379,https://m2.etcd.local:22379,https://m3.etcd.local:32379 \
+  put abc def

mvcc/backend/backend.go

@@ -25,7 +25,7 @@ import (
 	"sync/atomic"
 	"time"
 
-	"github.com/boltdb/bolt"
+	bolt "github.com/coreos/bbolt"
 	"github.com/coreos/pkg/capnslog"
 )
 

mvcc/backend/backend_test.go

@@ -21,7 +21,7 @@ import (
 	"testing"
 	"time"
 
-	"github.com/boltdb/bolt"
+	bolt "github.com/coreos/bbolt"
 )
 
 func TestBackendClose(t *testing.T) {

mvcc/backend/batch_tx.go

@@ -20,7 +20,7 @@ import (
 	"sync/atomic"
 	"time"
 
-	"github.com/boltdb/bolt"
+	bolt "github.com/coreos/bbolt"
 )
 
 type BatchTx interface {

mvcc/backend/batch_tx_test.go

@@ -19,7 +19,7 @@ import (
 	"testing"
 	"time"
 
-	"github.com/boltdb/bolt"
+	bolt "github.com/coreos/bbolt"
 )
 
 func TestBatchTxPut(t *testing.T) {

mvcc/backend/boltoption_default.go

@@ -16,6 +16,6 @@
 
 package backend
 
-import "github.com/boltdb/bolt"
+import bolt "github.com/coreos/bbolt"
 
 var boltOpenOptions *bolt.Options = nil

mvcc/backend/boltoption_linux.go

@@ -17,7 +17,7 @@ package backend
 import (
 	"syscall"
 
-	"github.com/boltdb/bolt"
+	bolt "github.com/coreos/bbolt"
 )
 
 // syscall.MAP_POPULATE on linux 2.6.23+ does sequential read-ahead

mvcc/watchable_store.go

@@ -258,6 +258,21 @@ func (s *watchableStore) cancelWatcher(wa *watcher) {
 	s.mu.Unlock()
 }
 
+func (s *watchableStore) Restore(b backend.Backend) error {
+	s.mu.Lock()
+	defer s.mu.Unlock()
+	err := s.store.Restore(b)
+	if err != nil {
+		return err
+	}
+
+	for wa := range s.synced.watchers {
+		s.unsynced.watchers.add(wa)
+	}
+	s.synced = newWatcherGroup()
+	return nil
+}
+
 // syncWatchersLoop syncs the watcher in the unsynced map every 100ms.
 func (s *watchableStore) syncWatchersLoop() {
 	defer s.wg.Done()

mvcc/watchable_store_test.go

@@ -294,6 +294,39 @@ func TestWatchFutureRev(t *testing.T) {
 	}
 }
 
+func TestWatchRestore(t *testing.T) {
+	b, tmpPath := backend.NewDefaultTmpBackend()
+	s := newWatchableStore(b, &lease.FakeLessor{}, nil)
+	defer cleanup(s, b, tmpPath)
+
+	testKey := []byte("foo")
+	testValue := []byte("bar")
+	rev := s.Put(testKey, testValue, lease.NoLease)
+
+	newBackend, newPath := backend.NewDefaultTmpBackend()
+	newStore := newWatchableStore(newBackend, &lease.FakeLessor{}, nil)
+	defer cleanup(newStore, newBackend, newPath)
+
+	w := newStore.NewWatchStream()
+	w.Watch(testKey, nil, rev-1)
+
+	newStore.Restore(b)
+	select {
+	case resp := <-w.Chan():
+		if resp.Revision != rev {
+			t.Fatalf("rev = %d, want %d", resp.Revision, rev)
+		}
+		if len(resp.Events) != 1 {
+			t.Fatalf("failed to get events from the response")
+		}
+		if resp.Events[0].Kv.ModRevision != rev {
+			t.Fatalf("kv.rev = %d, want %d", resp.Events[0].Kv.ModRevision, rev)
+		}
+	case <-time.After(time.Second):
+		t.Fatal("failed to receive event in 1 second.")
+	}
+}
+
 // TestWatchBatchUnsynced tests batching on unsynced watchers
 func TestWatchBatchUnsynced(t *testing.T) {
 	b, tmpPath := backend.NewDefaultTmpBackend()

tools/etcd-dump-db/backend.go

@@ -18,7 +18,7 @@ import (
 	"fmt"
 	"path/filepath"
 
-	"github.com/boltdb/bolt"
+	bolt "github.com/coreos/bbolt"
 	"github.com/coreos/etcd/mvcc"
 	"github.com/coreos/etcd/mvcc/backend"
 )

version/version.go

@@ -26,7 +26,7 @@ import (
 var (
 	// MinClusterVersion is the min cluster version this etcd binary is compatible with.
 	MinClusterVersion = "3.0.0"
-	Version           = "3.1.10"
+	Version           = "3.1.11"
 	APIVersion        = "unknown"
 
 	// Git SHA Value will be set during build