perso/syncthing-arm
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

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

Browse Source 
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
Download Patch File Download Diff File Expand all files Collapse all files

282
vendor/github.com/lib/pq/copy.go generated vendored
View File

@@ -1,282 +0,0 @@
package pq
import (
"database/sql/driver"
"encoding/binary"
"errors"
"fmt"
"sync"
)
var (
errCopyInClosed = errors.New("pq: copyin statement has already been closed")
errBinaryCopyNotSupported = errors.New("pq: only text format supported for COPY")
errCopyToNotSupported = errors.New("pq: COPY TO is not supported")
errCopyNotSupportedOutsideTxn = errors.New("pq: COPY is only allowed inside a transaction")
errCopyInProgress = errors.New("pq: COPY in progress")
)
// CopyIn creates a COPY FROM statement which can be prepared with
// Tx.Prepare(). The target table should be visible in search_path.
func CopyIn(table string, columns ...string) string {
stmt := "COPY " + QuoteIdentifier(table) + " ("
for i, col := range columns {
if i != 0 {
stmt += ", "
}
stmt += QuoteIdentifier(col)
}
stmt += ") FROM STDIN"
return stmt
}
// CopyInSchema creates a COPY FROM statement which can be prepared with
// Tx.Prepare().
func CopyInSchema(schema, table string, columns ...string) string {
stmt := "COPY " + QuoteIdentifier(schema) + "." + QuoteIdentifier(table) + " ("
for i, col := range columns {
if i != 0 {
stmt += ", "
}
stmt += QuoteIdentifier(col)
}
stmt += ") FROM STDIN"
return stmt
}
type copyin struct {
cn *conn
buffer []byte
rowData chan []byte
done chan bool
closed bool
sync.Mutex // guards err
err error
}
const ciBufferSize = 64 * 1024
// flush buffer before the buffer is filled up and needs reallocation
const ciBufferFlushSize = 63 * 1024
func (cn *conn) prepareCopyIn(q string) (_ driver.Stmt, err error) {
if !cn.isInTransaction() {
return nil, errCopyNotSupportedOutsideTxn
}
ci := &copyin{
cn: cn,
buffer: make([]byte, 0, ciBufferSize),
rowData: make(chan []byte),
done: make(chan bool, 1),
}
// add CopyData identifier + 4 bytes for message length
ci.buffer = append(ci.buffer, 'd', 0, 0, 0, 0)
b := cn.writeBuf('Q')
b.string(q)
cn.send(b)
awaitCopyInResponse:
for {
t, r := cn.recv1()
switch t {
case 'G':
if r.byte() != 0 {
err = errBinaryCopyNotSupported
break awaitCopyInResponse
}
go ci.resploop()
return ci, nil
case 'H':
err = errCopyToNotSupported
break awaitCopyInResponse
case 'E':
err = parseError(r)
case 'Z':
if err == nil {
ci.setBad()
errorf("unexpected ReadyForQuery in response to COPY")
}
cn.processReadyForQuery(r)
return nil, err
default:
ci.setBad()
errorf("unknown response for copy query: %q", t)
}
}
// something went wrong, abort COPY before we return
b = cn.writeBuf('f')
b.string(err.Error())
cn.send(b)
for {
t, r := cn.recv1()
switch t {
case 'c', 'C', 'E':
case 'Z':
// correctly aborted, we're done
cn.processReadyForQuery(r)
return nil, err
default:
ci.setBad()
errorf("unknown response for CopyFail: %q", t)
}
}
}
func (ci *copyin) flush(buf []byte) {
// set message length (without message identifier)
binary.BigEndian.PutUint32(buf[1:], uint32(len(buf)-1))
_, err := ci.cn.c.Write(buf)
if err != nil {
panic(err)
}
}
func (ci *copyin) resploop() {
for {
var r readBuf
t, err := ci.cn.recvMessage(&r)
if err != nil {
ci.setBad()
ci.setError(err)
ci.done <- true
return
}
switch t {
case 'C':
// complete
case 'N':
// NoticeResponse
case 'Z':
ci.cn.processReadyForQuery(&r)
ci.done <- true
return
case 'E':
err := parseError(&r)
ci.setError(err)
default:
ci.setBad()
ci.setError(fmt.Errorf("unknown response during CopyIn: %q", t))
ci.done <- true
return
}
}
}
func (ci *copyin) setBad() {
ci.Lock()
ci.cn.bad = true
ci.Unlock()
}
func (ci *copyin) isBad() bool {
ci.Lock()
b := ci.cn.bad
ci.Unlock()
return b
}
func (ci *copyin) isErrorSet() bool {
ci.Lock()
isSet := (ci.err != nil)
ci.Unlock()
return isSet
}
// setError() sets ci.err if one has not been set already. Caller must not be
// holding ci.Mutex.
func (ci *copyin) setError(err error) {
ci.Lock()
if ci.err == nil {
ci.err = err
}
ci.Unlock()
}
func (ci *copyin) NumInput() int {
return -1
}
func (ci *copyin) Query(v []driver.Value) (r driver.Rows, err error) {
return nil, ErrNotSupported
}
// Exec inserts values into the COPY stream. The insert is asynchronous
// and Exec can return errors from previous Exec calls to the same
// COPY stmt.
//
// You need to call Exec(nil) to sync the COPY stream and to get any
// errors from pending data, since Stmt.Close() doesn't return errors
// to the user.
func (ci *copyin) Exec(v []driver.Value) (r driver.Result, err error) {
if ci.closed {
return nil, errCopyInClosed
}
if ci.isBad() {
return nil, driver.ErrBadConn
}
defer ci.cn.errRecover(&err)
if ci.isErrorSet() {
return nil, ci.err
}
if len(v) == 0 {
return nil, ci.Close()
}
numValues := len(v)
for i, value := range v {
ci.buffer = appendEncodedText(&ci.cn.parameterStatus, ci.buffer, value)
if i < numValues-1 {
ci.buffer = append(ci.buffer, '\t')
}
}
ci.buffer = append(ci.buffer, '\n')
if len(ci.buffer) > ciBufferFlushSize {
ci.flush(ci.buffer)
// reset buffer, keep bytes for message identifier and length
ci.buffer = ci.buffer[:5]
}
return driver.RowsAffected(0), nil
}
func (ci *copyin) Close() (err error) {
if ci.closed { // Don't do anything, we're already closed
return nil
}
ci.closed = true
if ci.isBad() {
return driver.ErrBadConn
}
defer ci.cn.errRecover(&err)
if len(ci.buffer) > 0 {
ci.flush(ci.buffer)
}
// Avoid touching the scratch buffer as resploop could be using it.
err = ci.cn.sendSimpleMessage('c')
if err != nil {
return err
}
<-ci.done
ci.cn.inCopy = false
if ci.isErrorSet() {
err = ci.err
return err
}
return nil
}