cmd/stdiscosrv: New discovery server (fixes #4618)

This is a new revision of the discovery server. Relevant changes and
non-changes:

- Protocol towards clients is unchanged.

- Recommended large scale design is still to be deployed nehind nginx (I
  tested, and it's still a lot faster at terminating TLS).

- Database backend is leveldb again, only. It scales enough, is easy to
  setup, and we don't need any backend to take care of.

- Server supports replication. This is a simple TCP channel - protect it
  with a firewall when deploying over the internet. (We deploy this within
  the same datacenter, and with firewall.) Any incoming client announces
  are sent over the replication channel(s) to other peer discosrvs.
  Incoming replication changes are applied to the database as if they came
  from clients, but without the TLS/certificate overhead.

- Metrics are exposed using the prometheus library, when enabled.

- The database values and replication protocol is protobuf, because JSON
  was quite CPU intensive when I tried that and benchmarked it.

- The "Retry-After" value for failed lookups gets slowly increased from
  a default of 120 seconds, by 5 seconds for each failed lookup,
  independently by each discosrv. This lowers the query load over time for
  clients that are never seen. The Retry-After maxes out at 3600 after a
  couple of weeks of this increase. The number of failed lookups is
  stored in the database, now and then (avoiding making each lookup a
  database put).

All in all this means clients can be pointed towards a cluster using
just multiple A / AAAA records to gain both load sharing and redundancy
(if one is down, clients will talk to the remaining ones).

GitHub-Pull-Request: https://github.com/syncthing/syncthing/pull/4648
This commit is contained in:
Jakob Borg
2018-01-14 08:52:31 +00:00
parent 341b9691a7
commit 916ec63af6
864 changed files with 216825 additions and 64540 deletions

603
vendor/github.com/lib/pq/encode.go generated vendored
View File

@@ -1,603 +0,0 @@
package pq
import (
"bytes"
"database/sql/driver"
"encoding/binary"
"encoding/hex"
"errors"
"fmt"
"math"
"strconv"
"strings"
"sync"
"time"
"github.com/lib/pq/oid"
)
func binaryEncode(parameterStatus *parameterStatus, x interface{}) []byte {
switch v := x.(type) {
case []byte:
return v
default:
return encode(parameterStatus, x, oid.T_unknown)
}
}
func encode(parameterStatus *parameterStatus, x interface{}, pgtypOid oid.Oid) []byte {
switch v := x.(type) {
case int64:
return strconv.AppendInt(nil, v, 10)
case float64:
return strconv.AppendFloat(nil, v, 'f', -1, 64)
case []byte:
if pgtypOid == oid.T_bytea {
return encodeBytea(parameterStatus.serverVersion, v)
}
return v
case string:
if pgtypOid == oid.T_bytea {
return encodeBytea(parameterStatus.serverVersion, []byte(v))
}
return []byte(v)
case bool:
return strconv.AppendBool(nil, v)
case time.Time:
return formatTs(v)
default:
errorf("encode: unknown type for %T", v)
}
panic("not reached")
}
func decode(parameterStatus *parameterStatus, s []byte, typ oid.Oid, f format) interface{} {
switch f {
case formatBinary:
return binaryDecode(parameterStatus, s, typ)
case formatText:
return textDecode(parameterStatus, s, typ)
default:
panic("not reached")
}
}
func binaryDecode(parameterStatus *parameterStatus, s []byte, typ oid.Oid) interface{} {
switch typ {
case oid.T_bytea:
return s
case oid.T_int8:
return int64(binary.BigEndian.Uint64(s))
case oid.T_int4:
return int64(int32(binary.BigEndian.Uint32(s)))
case oid.T_int2:
return int64(int16(binary.BigEndian.Uint16(s)))
case oid.T_uuid:
b, err := decodeUUIDBinary(s)
if err != nil {
panic(err)
}
return b
default:
errorf("don't know how to decode binary parameter of type %d", uint32(typ))
}
panic("not reached")
}
func textDecode(parameterStatus *parameterStatus, s []byte, typ oid.Oid) interface{} {
switch typ {
case oid.T_char, oid.T_varchar, oid.T_text:
return string(s)
case oid.T_bytea:
b, err := parseBytea(s)
if err != nil {
errorf("%s", err)
}
return b
case oid.T_timestamptz:
return parseTs(parameterStatus.currentLocation, string(s))
case oid.T_timestamp, oid.T_date:
return parseTs(nil, string(s))
case oid.T_time:
return mustParse("15:04:05", typ, s)
case oid.T_timetz:
return mustParse("15:04:05-07", typ, s)
case oid.T_bool:
return s[0] == 't'
case oid.T_int8, oid.T_int4, oid.T_int2:
i, err := strconv.ParseInt(string(s), 10, 64)
if err != nil {
errorf("%s", err)
}
return i
case oid.T_float4, oid.T_float8:
bits := 64
if typ == oid.T_float4 {
bits = 32
}
f, err := strconv.ParseFloat(string(s), bits)
if err != nil {
errorf("%s", err)
}
return f
}
return s
}
// appendEncodedText encodes item in text format as required by COPY
// and appends to buf
func appendEncodedText(parameterStatus *parameterStatus, buf []byte, x interface{}) []byte {
switch v := x.(type) {
case int64:
return strconv.AppendInt(buf, v, 10)
case float64:
return strconv.AppendFloat(buf, v, 'f', -1, 64)
case []byte:
encodedBytea := encodeBytea(parameterStatus.serverVersion, v)
return appendEscapedText(buf, string(encodedBytea))
case string:
return appendEscapedText(buf, v)
case bool:
return strconv.AppendBool(buf, v)
case time.Time:
return append(buf, formatTs(v)...)
case nil:
return append(buf, "\\N"...)
default:
errorf("encode: unknown type for %T", v)
}
panic("not reached")
}
func appendEscapedText(buf []byte, text string) []byte {
escapeNeeded := false
startPos := 0
var c byte
// check if we need to escape
for i := 0; i < len(text); i++ {
c = text[i]
if c == '\\' || c == '\n' || c == '\r' || c == '\t' {
escapeNeeded = true
startPos = i
break
}
}
if !escapeNeeded {
return append(buf, text...)
}
// copy till first char to escape, iterate the rest
result := append(buf, text[:startPos]...)
for i := startPos; i < len(text); i++ {
c = text[i]
switch c {
case '\\':
result = append(result, '\\', '\\')
case '\n':
result = append(result, '\\', 'n')
case '\r':
result = append(result, '\\', 'r')
case '\t':
result = append(result, '\\', 't')
default:
result = append(result, c)
}
}
return result
}
func mustParse(f string, typ oid.Oid, s []byte) time.Time {
str := string(s)
// check for a 30-minute-offset timezone
if (typ == oid.T_timestamptz || typ == oid.T_timetz) &&
str[len(str)-3] == ':' {
f += ":00"
}
t, err := time.Parse(f, str)
if err != nil {
errorf("decode: %s", err)
}
return t
}
var errInvalidTimestamp = errors.New("invalid timestamp")
type timestampParser struct {
err error
}
func (p *timestampParser) expect(str string, char byte, pos int) {
if p.err != nil {
return
}
if pos+1 > len(str) {
p.err = errInvalidTimestamp
return
}
if c := str[pos]; c != char && p.err == nil {
p.err = fmt.Errorf("expected '%v' at position %v; got '%v'", char, pos, c)
}
}
func (p *timestampParser) mustAtoi(str string, begin int, end int) int {
if p.err != nil {
return 0
}
if begin < 0 || end < 0 || begin > end || end > len(str) {
p.err = errInvalidTimestamp
return 0
}
result, err := strconv.Atoi(str[begin:end])
if err != nil {
if p.err == nil {
p.err = fmt.Errorf("expected number; got '%v'", str)
}
return 0
}
return result
}
// The location cache caches the time zones typically used by the client.
type locationCache struct {
cache map[int]*time.Location
lock sync.Mutex
}
// All connections share the same list of timezones. Benchmarking shows that
// about 5% speed could be gained by putting the cache in the connection and
// losing the mutex, at the cost of a small amount of memory and a somewhat
// significant increase in code complexity.
var globalLocationCache = newLocationCache()
func newLocationCache() *locationCache {
return &locationCache{cache: make(map[int]*time.Location)}
}
// Returns the cached timezone for the specified offset, creating and caching
// it if necessary.
func (c *locationCache) getLocation(offset int) *time.Location {
c.lock.Lock()
defer c.lock.Unlock()
location, ok := c.cache[offset]
if !ok {
location = time.FixedZone("", offset)
c.cache[offset] = location
}
return location
}
var infinityTsEnabled = false
var infinityTsNegative time.Time
var infinityTsPositive time.Time
const (
infinityTsEnabledAlready = "pq: infinity timestamp enabled already"
infinityTsNegativeMustBeSmaller = "pq: infinity timestamp: negative value must be smaller (before) than positive"
)
// EnableInfinityTs controls the handling of Postgres' "-infinity" and
// "infinity" "timestamp"s.
//
// If EnableInfinityTs is not called, "-infinity" and "infinity" will return
// []byte("-infinity") and []byte("infinity") respectively, and potentially
// cause error "sql: Scan error on column index 0: unsupported driver -> Scan
// pair: []uint8 -> *time.Time", when scanning into a time.Time value.
//
// Once EnableInfinityTs has been called, all connections created using this
// driver will decode Postgres' "-infinity" and "infinity" for "timestamp",
// "timestamp with time zone" and "date" types to the predefined minimum and
// maximum times, respectively. When encoding time.Time values, any time which
// equals or precedes the predefined minimum time will be encoded to
// "-infinity". Any values at or past the maximum time will similarly be
// encoded to "infinity".
//
// If EnableInfinityTs is called with negative >= positive, it will panic.
// Calling EnableInfinityTs after a connection has been established results in
// undefined behavior. If EnableInfinityTs is called more than once, it will
// panic.
func EnableInfinityTs(negative time.Time, positive time.Time) {
if infinityTsEnabled {
panic(infinityTsEnabledAlready)
}
if !negative.Before(positive) {
panic(infinityTsNegativeMustBeSmaller)
}
infinityTsEnabled = true
infinityTsNegative = negative
infinityTsPositive = positive
}
/*
* Testing might want to toggle infinityTsEnabled
*/
func disableInfinityTs() {
infinityTsEnabled = false
}
// This is a time function specific to the Postgres default DateStyle
// setting ("ISO, MDY"), the only one we currently support. This
// accounts for the discrepancies between the parsing available with
// time.Parse and the Postgres date formatting quirks.
func parseTs(currentLocation *time.Location, str string) interface{} {
switch str {
case "-infinity":
if infinityTsEnabled {
return infinityTsNegative
}
return []byte(str)
case "infinity":
if infinityTsEnabled {
return infinityTsPositive
}
return []byte(str)
}
t, err := ParseTimestamp(currentLocation, str)
if err != nil {
panic(err)
}
return t
}
// ParseTimestamp parses Postgres' text format. It returns a time.Time in
// currentLocation iff that time's offset agrees with the offset sent from the
// Postgres server. Otherwise, ParseTimestamp returns a time.Time with the
// fixed offset offset provided by the Postgres server.
func ParseTimestamp(currentLocation *time.Location, str string) (time.Time, error) {
p := timestampParser{}
monSep := strings.IndexRune(str, '-')
// this is Gregorian year, not ISO Year
// In Gregorian system, the year 1 BC is followed by AD 1
year := p.mustAtoi(str, 0, monSep)
daySep := monSep + 3
month := p.mustAtoi(str, monSep+1, daySep)
p.expect(str, '-', daySep)
timeSep := daySep + 3
day := p.mustAtoi(str, daySep+1, timeSep)
minLen := monSep + len("01-01") + 1
isBC := strings.HasSuffix(str, " BC")
if isBC {
minLen += 3
}
var hour, minute, second int
if len(str) > minLen {
p.expect(str, ' ', timeSep)
minSep := timeSep + 3
p.expect(str, ':', minSep)
hour = p.mustAtoi(str, timeSep+1, minSep)
secSep := minSep + 3
p.expect(str, ':', secSep)
minute = p.mustAtoi(str, minSep+1, secSep)
secEnd := secSep + 3
second = p.mustAtoi(str, secSep+1, secEnd)
}
remainderIdx := monSep + len("01-01 00:00:00") + 1
// Three optional (but ordered) sections follow: the
// fractional seconds, the time zone offset, and the BC
// designation. We set them up here and adjust the other
// offsets if the preceding sections exist.
nanoSec := 0
tzOff := 0
if remainderIdx < len(str) && str[remainderIdx] == '.' {
fracStart := remainderIdx + 1
fracOff := strings.IndexAny(str[fracStart:], "-+ ")
if fracOff < 0 {
fracOff = len(str) - fracStart
}
fracSec := p.mustAtoi(str, fracStart, fracStart+fracOff)
nanoSec = fracSec * (1000000000 / int(math.Pow(10, float64(fracOff))))
remainderIdx += fracOff + 1
}
if tzStart := remainderIdx; tzStart < len(str) && (str[tzStart] == '-' || str[tzStart] == '+') {
// time zone separator is always '-' or '+' (UTC is +00)
var tzSign int
switch c := str[tzStart]; c {
case '-':
tzSign = -1
case '+':
tzSign = +1
default:
return time.Time{}, fmt.Errorf("expected '-' or '+' at position %v; got %v", tzStart, c)
}
tzHours := p.mustAtoi(str, tzStart+1, tzStart+3)
remainderIdx += 3
var tzMin, tzSec int
if remainderIdx < len(str) && str[remainderIdx] == ':' {
tzMin = p.mustAtoi(str, remainderIdx+1, remainderIdx+3)
remainderIdx += 3
}
if remainderIdx < len(str) && str[remainderIdx] == ':' {
tzSec = p.mustAtoi(str, remainderIdx+1, remainderIdx+3)
remainderIdx += 3
}
tzOff = tzSign * ((tzHours * 60 * 60) + (tzMin * 60) + tzSec)
}
var isoYear int
if isBC {
isoYear = 1 - year
remainderIdx += 3
} else {
isoYear = year
}
if remainderIdx < len(str) {
return time.Time{}, fmt.Errorf("expected end of input, got %v", str[remainderIdx:])
}
t := time.Date(isoYear, time.Month(month), day,
hour, minute, second, nanoSec,
globalLocationCache.getLocation(tzOff))
if currentLocation != nil {
// Set the location of the returned Time based on the session's
// TimeZone value, but only if the local time zone database agrees with
// the remote database on the offset.
lt := t.In(currentLocation)
_, newOff := lt.Zone()
if newOff == tzOff {
t = lt
}
}
return t, p.err
}
// formatTs formats t into a format postgres understands.
func formatTs(t time.Time) []byte {
if infinityTsEnabled {
// t <= -infinity : ! (t > -infinity)
if !t.After(infinityTsNegative) {
return []byte("-infinity")
}
// t >= infinity : ! (!t < infinity)
if !t.Before(infinityTsPositive) {
return []byte("infinity")
}
}
return FormatTimestamp(t)
}
// FormatTimestamp formats t into Postgres' text format for timestamps.
func FormatTimestamp(t time.Time) []byte {
// Need to send dates before 0001 A.D. with " BC" suffix, instead of the
// minus sign preferred by Go.
// Beware, "0000" in ISO is "1 BC", "-0001" is "2 BC" and so on
bc := false
if t.Year() <= 0 {
// flip year sign, and add 1, e.g: "0" will be "1", and "-10" will be "11"
t = t.AddDate((-t.Year())*2+1, 0, 0)
bc = true
}
b := []byte(t.Format("2006-01-02 15:04:05.999999999Z07:00"))
_, offset := t.Zone()
offset = offset % 60
if offset != 0 {
// RFC3339Nano already printed the minus sign
if offset < 0 {
offset = -offset
}
b = append(b, ':')
if offset < 10 {
b = append(b, '0')
}
b = strconv.AppendInt(b, int64(offset), 10)
}
if bc {
b = append(b, " BC"...)
}
return b
}
// Parse a bytea value received from the server. Both "hex" and the legacy
// "escape" format are supported.
func parseBytea(s []byte) (result []byte, err error) {
if len(s) >= 2 && bytes.Equal(s[:2], []byte("\\x")) {
// bytea_output = hex
s = s[2:] // trim off leading "\\x"
result = make([]byte, hex.DecodedLen(len(s)))
_, err := hex.Decode(result, s)
if err != nil {
return nil, err
}
} else {
// bytea_output = escape
for len(s) > 0 {
if s[0] == '\\' {
// escaped '\\'
if len(s) >= 2 && s[1] == '\\' {
result = append(result, '\\')
s = s[2:]
continue
}
// '\\' followed by an octal number
if len(s) < 4 {
return nil, fmt.Errorf("invalid bytea sequence %v", s)
}
r, err := strconv.ParseInt(string(s[1:4]), 8, 9)
if err != nil {
return nil, fmt.Errorf("could not parse bytea value: %s", err.Error())
}
result = append(result, byte(r))
s = s[4:]
} else {
// We hit an unescaped, raw byte. Try to read in as many as
// possible in one go.
i := bytes.IndexByte(s, '\\')
if i == -1 {
result = append(result, s...)
break
}
result = append(result, s[:i]...)
s = s[i:]
}
}
}
return result, nil
}
func encodeBytea(serverVersion int, v []byte) (result []byte) {
if serverVersion >= 90000 {
// Use the hex format if we know that the server supports it
result = make([]byte, 2+hex.EncodedLen(len(v)))
result[0] = '\\'
result[1] = 'x'
hex.Encode(result[2:], v)
} else {
// .. or resort to "escape"
for _, b := range v {
if b == '\\' {
result = append(result, '\\', '\\')
} else if b < 0x20 || b > 0x7e {
result = append(result, []byte(fmt.Sprintf("\\%03o", b))...)
} else {
result = append(result, b)
}
}
}
return result
}
// NullTime represents a time.Time that may be null. NullTime implements the
// sql.Scanner interface so it can be used as a scan destination, similar to
// sql.NullString.
type NullTime struct {
Time time.Time
Valid bool // Valid is true if Time is not NULL
}
// Scan implements the Scanner interface.
func (nt *NullTime) Scan(value interface{}) error {
nt.Time, nt.Valid = value.(time.Time)
return nil
}
// Value implements the driver Valuer interface.
func (nt NullTime) Value() (driver.Value, error) {
if !nt.Valid {
return nil, nil
}
return nt.Time, nil
}