Merge pull request #1 from seatme/feature/logical_replication

Add logical replication support to PGX
2017-01-04 15:42:07 -08:00
parent e96c105b55 cf225c8365
commit 9d60e77aa4
5 changed files with 548 additions and 2 deletions
@@ -25,6 +25,7 @@ Pgx supports many additional features beyond what is available through database/
 * Large object support
 * Null mapping to Null* struct or pointer to pointer.
 * Supports database/sql.Scanner and database/sql/driver.Valuer interfaces for custom types
+* Logical replication connections, including receiving WAL and sending standby status updates

 ## Performance

@@ -68,6 +69,7 @@ Then run the following SQL:
    create user pgx_md5 password 'secret';
    create user " tricky, ' } "" \ test user " password 'secret';
    create database pgx_test;
+    create user pgx_replication with replication password 'secret';

 Connect to database pgx_test and run:

@@ -100,6 +102,12 @@ If you are developing on Windows with TCP connections:
    host  pgx_test  pgx_pw    127.0.0.1/32 password
    host  pgx_test  pgx_md5   127.0.0.1/32 md5

+For replication testing, add the following to your postgresql.conf:
+
+    wal_level='logical'
+    max_wal_senders=5
+    max_replication_slots=5
+
 ## Version Policy

 pgx follows semantic versioning for the documented public API on stable releases. Branch ```v2``` is the latest stable release. ```master``` can contain new features or behavior that will change or be removed before being merged to the stable ```v2``` branch (in practice, this occurs very rarely).
@@ -305,6 +305,12 @@ func (c *Conn) connect(config ConnConfig, network, address string, tlsConfig *tl
 				c.log(LogLevelInfo, "Connection established")
 			}

+			// Replication connections can't execute the queries to
+			// populate the c.PgTypes and c.pgsqlAfInet
+			if _, ok := msg.options["replication"]; ok {
+				return nil
+			}
+
 			if c.PgTypes == nil {
 				err = c.loadPgTypes()
 				if err != nil {
@@ -0,0 +1,335 @@
+package pgx
+
+import (
+	"errors"
+	"fmt"
+	"net"
+	"time"
+)
+
+const (
+	copyBothResponse    = 'W'
+	walData             = 'w'
+	senderKeepalive     = 'k'
+	standbyStatusUpdate = 'r'
+	initialReplicationResponseTimeout = 5 * time.Second
+)
+
+var epochNano int64
+
+func init() {
+	epochNano = time.Date(2000, 1, 1, 0, 0, 0, 0, time.UTC).UnixNano()
+}
+
+// Format the given 64bit LSN value into the XXX/XXX format,
+// which is the format reported by postgres.
+func FormatLSN(lsn uint64) string {
+	return fmt.Sprintf("%X/%X", uint32(lsn>>32), uint32(lsn))
+}
+
+// Parse the given XXX/XXX format LSN as reported by postgres,
+// into a 64 bit integer as used internally by the wire procotols
+func ParseLSN(lsn string) (outputLsn uint64, err error) {
+	var upperHalf uint64
+	var lowerHalf uint64
+	var nparsed int
+	nparsed, err = fmt.Sscanf(lsn, "%X/%X", &upperHalf, &lowerHalf)
+	if err != nil {
+		return
+	}
+
+	if nparsed != 2 {
+		err = errors.New(fmt.Sprintf("Failed to parsed LSN: %s", lsn))
+		return
+	}
+
+	outputLsn = (upperHalf << 32) + lowerHalf
+	return
+}
+
+// The WAL message contains WAL payload entry data
+type WalMessage struct {
+	// The WAL start position of this data. This
+	// is the WAL position we need to track.
+	WalStart uint64
+	// The server wal end and server time are
+	// documented to track the end position and current
+	// time of the server, both of which appear to be
+	// unimplemented in pg 9.5.
+	ServerWalEnd uint64
+	ServerTime   uint64
+	// The WAL data is the raw unparsed binary WAL entry.
+	// The contents of this are determined by the output
+	// logical encoding plugin.
+	WalData []byte
+}
+
+func (w *WalMessage) Time() time.Time {
+	return time.Unix(0, (int64(w.ServerTime)*1000)+epochNano)
+}
+
+func (w *WalMessage) ByteLag() uint64 {
+	return (w.ServerWalEnd - w.WalStart)
+}
+
+func (w *WalMessage) String() string {
+	return fmt.Sprintf("Wal: %s Time: %s Lag: %d", FormatLSN(w.WalStart), w.Time(), w.ByteLag())
+}
+
+// The server heartbeat is sent periodically from the server,
+// including server status, and a reply request field
+type ServerHeartbeat struct {
+	// The current max wal position on the server,
+	// used for lag tracking
+	ServerWalEnd uint64
+	// The server time, in microseconds since jan 1 2000
+	ServerTime uint64
+	// If 1, the server is requesting a standby status message
+	// to be sent immediately.
+	ReplyRequested byte
+}
+
+func (s *ServerHeartbeat) Time() time.Time {
+	return time.Unix(0, (int64(s.ServerTime)*1000)+epochNano)
+}
+
+func (s *ServerHeartbeat) String() string {
+	return fmt.Sprintf("WalEnd: %s ReplyRequested: %d T: %s", FormatLSN(s.ServerWalEnd), s.ReplyRequested, s.Time())
+}
+
+// The replication message wraps all possible messages from the
+// server received during replication. At most one of the wal message
+// or server heartbeat will be non-nil
+type ReplicationMessage struct {
+	WalMessage      *WalMessage
+	ServerHeartbeat *ServerHeartbeat
+}
+
+// The standby status is the client side heartbeat sent to the postgresql
+// server to track the client wal positions. For practical purposes,
+// all wal positions are typically set to the same value.
+type StandbyStatus struct {
+	// The WAL position that's been locally written
+	WalWritePosition uint64
+	// The WAL position that's been locally flushed
+	WalFlushPosition uint64
+	// The WAL position that's been locally applied
+	WalApplyPosition uint64
+	// The client time in microseconds since jan 1 2000
+	ClientTime uint64
+	// If 1, requests the server to immediately send a
+	// server heartbeat
+	ReplyRequested byte
+}
+
+// Create a standby status struct, which sets all the WAL positions
+// to the given wal position, and the client time to the current time.
+// The wal positions are, in order:
+// WalFlushPosition
+// WalApplyPosition
+// WalWritePosition
+//
+// If only one position is provided, it will be used as the value for all 3
+// status fields. Note you must provide either 1 wal position, or all 3
+// in order to initialize the standby status.
+func NewStandbyStatus(walPositions ...uint64) (status *StandbyStatus, err error) {
+	if len(walPositions) == 1 {
+		status = new(StandbyStatus)
+		status.WalFlushPosition = walPositions[0]
+		status.WalApplyPosition = walPositions[0]
+		status.WalWritePosition = walPositions[0]
+	} else if len(walPositions) == 3 {
+		status = new(StandbyStatus)
+		status.WalFlushPosition = walPositions[0]
+		status.WalApplyPosition = walPositions[1]
+		status.WalWritePosition = walPositions[2]
+	} else {
+		err = errors.New(fmt.Sprintf("Invalid number of wal positions provided, need 1 or 3, got %d", len(walPositions)))
+		return
+	}
+	status.ClientTime = uint64((time.Now().UnixNano() - epochNano) / 1000)
+	return
+}
+
+// Send standby status to the server, which both acts as a keepalive
+// message to the server, as well as carries the WAL position of the
+// client, which then updates the server's replication slot position.
+func (c *Conn) SendStandbyStatus(k *StandbyStatus) (err error) {
+	writeBuf := newWriteBuf(c, copyData)
+	writeBuf.WriteByte(standbyStatusUpdate)
+	writeBuf.WriteInt64(int64(k.WalWritePosition))
+	writeBuf.WriteInt64(int64(k.WalFlushPosition))
+	writeBuf.WriteInt64(int64(k.WalApplyPosition))
+	writeBuf.WriteInt64(int64(k.ClientTime))
+	writeBuf.WriteByte(k.ReplyRequested)
+
+	writeBuf.closeMsg()
+
+	_, err = c.conn.Write(writeBuf.buf)
+	if err != nil {
+		c.die(err)
+	}
+
+	return
+}
+
+// Send the message to formally stop the replication stream. This
+// is done before calling Close() during a clean shutdown.
+func (c *Conn) StopReplication() (err error) {
+	writeBuf := newWriteBuf(c, copyDone)
+
+	writeBuf.closeMsg()
+
+	_, err = c.conn.Write(writeBuf.buf)
+	if err != nil {
+		c.die(err)
+	}
+	return
+}
+
+
+func (c *Conn) readReplicationMessage() (r *ReplicationMessage, err error) {
+	var t byte
+	var reader *msgReader
+	t, reader, err = c.rxMsg()
+	if err != nil {
+		return
+	}
+
+	switch t {
+	case noticeResponse:
+		pgError := c.rxErrorResponse(reader)
+		if c.shouldLog(LogLevelInfo) {
+			c.log(LogLevelInfo, pgError.Error())
+		}
+	case errorResponse:
+		err = c.rxErrorResponse(reader)
+		if c.shouldLog(LogLevelError) {
+			c.log(LogLevelError, err.Error())
+		}
+		return
+	case copyBothResponse:
+		// This is the tail end of the replication process start,
+		// and can be safely ignored
+		return
+	case copyData:
+		var msgType byte
+		msgType = reader.readByte()
+		switch msgType {
+		case walData:
+			walStart := reader.readInt64()
+			serverWalEnd := reader.readInt64()
+			serverTime := reader.readInt64()
+			walData := reader.readBytes(reader.msgBytesRemaining)
+			walMessage := WalMessage{WalStart: uint64(walStart),
+				ServerWalEnd: uint64(serverWalEnd),
+				ServerTime:   uint64(serverTime),
+				WalData:      walData,
+			}
+
+			return &ReplicationMessage{WalMessage: &walMessage}, nil
+		case senderKeepalive:
+			serverWalEnd := reader.readInt64()
+			serverTime := reader.readInt64()
+			replyNow := reader.readByte()
+			h := &ServerHeartbeat{ServerWalEnd: uint64(serverWalEnd), ServerTime: uint64(serverTime), ReplyRequested: replyNow}
+			return &ReplicationMessage{ServerHeartbeat: h}, nil
+		default:
+			if c.shouldLog(LogLevelError) {
+				c.log(LogLevelError,"Unexpected data playload message type %v", t)
+			}
+		}
+	default:
+		if c.shouldLog(LogLevelError) {
+			c.log(LogLevelError,"Unexpected replication message type %v", t)
+		}
+	}
+	return
+}
+
+// Wait for a single replication message up to timeout time.
+//
+// Properly using this requires some knowledge of the postgres replication mechanisms,
+// as the client can receive both WAL data (the ultimate payload) and server heartbeat
+// updates. The caller also must send standby status updates in order to keep the connection
+// alive and working.
+//
+// This returns pgx.ErrNotificationTimeout when there is no replication message by the specified
+// duration.
+func (c *Conn) WaitForReplicationMessage(timeout time.Duration) (r *ReplicationMessage, err error) {
+	var zeroTime time.Time
+
+	deadline := time.Now().Add(timeout)
+
+	// Use SetReadDeadline to implement the timeout. SetReadDeadline will
+	// cause operations to fail with a *net.OpError that has a Timeout()
+	// of true. Because the normal pgx rxMsg path considers any error to
+	// have potentially corrupted the state of the connection, it dies
+	// on any errors. So to avoid timeout errors in rxMsg we set the
+	// deadline and peek into the reader. If a timeout error occurs there
+	// we don't break the pgx connection. If the Peek returns that data
+	// is available then we turn off the read deadline before the rxMsg.
+	err = c.conn.SetReadDeadline(deadline)
+	if err != nil {
+		return nil, err
+	}
+
+	// Wait until there is a byte available before continuing onto the normal msg reading path
+	_, err = c.reader.Peek(1)
+	if err != nil {
+		c.conn.SetReadDeadline(zeroTime) // we can only return one error and we already have one -- so ignore possiple error from SetReadDeadline
+		if err, ok := err.(*net.OpError); ok && err.Timeout() {
+			return nil, ErrNotificationTimeout
+		}
+		return nil, err
+	}
+
+	err = c.conn.SetReadDeadline(zeroTime)
+	if err != nil {
+		return nil, err
+	}
+
+	return c.readReplicationMessage()
+}
+
+// Start a replication connection, sending WAL data to the given replication
+// receiver. This function wraps a START_REPLICATION command as documented
+// here:
+// https://www.postgresql.org/docs/9.5/static/protocol-replication.html
+//
+// Once started, the client needs to invoke WaitForReplicationMessage() in order
+// to fetch the WAL and standby status. Also, it is the responsibility of the caller
+// to periodically send StandbyStatus messages to update the replication slot position.
+//
+// This function assumes that slotName has already been created. In order to omit the timeline argument
+// pass a -1 for the timeline to get the server default behavior.
+func (c *Conn) StartReplication(slotName string, startLsn uint64, timeline int64, pluginArguments ...string) (err error) {
+	var queryString string
+	if timeline >= 0 {
+		queryString = fmt.Sprintf("START_REPLICATION SLOT %s LOGICAL %s TIMELINE %d", slotName, FormatLSN(startLsn), timeline)
+	} else {
+		queryString = fmt.Sprintf("START_REPLICATION SLOT %s LOGICAL %s", slotName, FormatLSN(startLsn))
+	}
+
+	for _, arg := range pluginArguments {
+		queryString += fmt.Sprintf(" %s", arg)
+	}
+
+	if err = c.sendQuery(queryString); err != nil {
+		return
+	}
+
+	// The first replication message that comes back here will be (in a success case)
+	// a empty CopyBoth that is (apparently) sent as the confirmation that the replication has
+	// started. This call will either return nil, nil or if it returns an error
+	// that indicates the start replication command failed
+	var r *ReplicationMessage
+	r, err = c.WaitForReplicationMessage(initialReplicationResponseTimeout)
+	if err != nil && r != nil {
+		if c.shouldLog(LogLevelError) {
+			c.log(LogLevelError, "Unxpected replication message %v", r)
+		}
+	}
+
+	return
+}
@@ -0,0 +1,197 @@
+package pgx_test
+
+import (
+	"github.com/jackc/pgx"
+	"strconv"
+	"strings"
+	"testing"
+	"time"
+	"reflect"
+	"fmt"
+)
+
+// This function uses a postgresql 9.6 specific column
+func getConfirmedFlushLsnFor(t *testing.T, conn *pgx.Conn, slot string) string {
+	// Fetch the restart LSN of the slot, to establish a starting point
+	rows, err := conn.Query(fmt.Sprintf("select confirmed_flush_lsn from pg_replication_slots where slot_name='%s'", slot))
+	if err != nil {
+		t.Fatalf("conn.Query failed: %v", err)
+	}
+	defer rows.Close()
+
+	var restartLsn string
+	for rows.Next() {
+		rows.Scan(&restartLsn)
+	}
+	return restartLsn
+}
+
+// This battleship test (at least somewhat by necessity) does
+// several things all at once in a single run. It:
+// - Establishes a replication connection & slot
+// - Does a series of operations to create some known WAL entries
+// - Replicates the entries down, and checks that the rows it
+//   created come down in order
+// - Sends a standby status message to update the server with the
+//   wal position of the slot
+// - Checks the wal position of the slot on the server to make sure
+//   the update succeeded
+func TestSimpleReplicationConnection(t *testing.T) {
+	t.Parallel()
+
+	var err error
+	var replicationUserConfig pgx.ConnConfig
+	var replicationConnConfig pgx.ConnConfig
+
+	replicationUserConfig = *defaultConnConfig
+	replicationUserConfig.User = "pgx_replication"
+	conn := mustConnect(t, replicationUserConfig)
+	defer closeConn(t, conn)
+
+	replicationConnConfig = *defaultConnConfig
+	replicationConnConfig.User = "pgx_replication"
+	replicationConnConfig.RuntimeParams = make(map[string]string)
+	replicationConnConfig.RuntimeParams["replication"] = "database"
+
+	replicationConn := mustConnect(t, replicationConnConfig)
+	defer closeConn(t, replicationConn)
+
+	_, err = replicationConn.Exec("CREATE_REPLICATION_SLOT pgx_test LOGICAL test_decoding")
+	if err != nil {
+		t.Logf("replication slot create failed: %v", err)
+	}
+
+	// Do a simple change so we can get some wal data
+	_, err = conn.Exec("create table if not exists replication_test (a integer)")
+	if err != nil {
+		t.Fatalf("Failed to create table: %v", err)
+	}
+
+	err = replicationConn.StartReplication("pgx_test", 0, -1)
+	if err != nil {
+		t.Fatalf("Failed to start replication: %v", err)
+	}
+
+	var i int32
+	var insertedTimes []int64
+	for i < 5 {
+		var ct pgx.CommandTag
+		currentTime := time.Now().Unix()
+		insertedTimes = append(insertedTimes, currentTime)
+		ct, err = conn.Exec("insert into replication_test(a) values($1)", currentTime)
+		if err != nil {
+			t.Fatalf("Insert failed: %v", err)
+		}
+		t.Logf("Inserted %d rows", ct.RowsAffected())
+		i++
+	}
+
+	i = 0
+	var foundTimes []int64
+	var foundCount int
+	var maxWal uint64
+	for {
+		var message *pgx.ReplicationMessage
+
+		message, err = replicationConn.WaitForReplicationMessage(time.Duration(1 * time.Second))
+		if err != nil {
+			if err != pgx.ErrNotificationTimeout {
+				t.Fatalf("Replication failed: %v %s", err, reflect.TypeOf(err))
+			}
+		}
+		if message != nil {
+			if message.WalMessage != nil {
+				// The waldata payload with the test_decoding plugin looks like:
+				// public.replication_test: INSERT: a[integer]:2
+				// What we wanna do here is check that once we find one of our inserted times,
+				// that they occur in the wal stream in the order we executed them.
+				walString := string(message.WalMessage.WalData)
+				if strings.Contains(walString, "public.replication_test: INSERT") {
+					stringParts := strings.Split(walString, ":")
+					offset, err := strconv.ParseInt(stringParts[len(stringParts)-1], 10, 64)
+					if err != nil {
+						t.Fatalf("Failed to parse walString %s", walString)
+					}
+					if foundCount > 0 || offset == insertedTimes[0] {
+						foundTimes = append(foundTimes, offset)
+						foundCount++
+					}
+				}
+				if message.WalMessage.WalStart > maxWal {
+					maxWal = message.WalMessage.WalStart
+				}
+
+			}
+			if message.ServerHeartbeat != nil {
+				t.Logf("Got heartbeat: %s", message.ServerHeartbeat)
+			}
+		} else {
+			t.Log("Timed out waiting for wal message")
+			i++
+		}
+		if i > 3 {
+			t.Log("Actual timeout")
+			break
+		}
+	}
+
+	if foundCount != len(insertedTimes) {
+		t.Fatalf("Failed to find all inserted time values in WAL stream (found %d expected %d)", foundCount, len(insertedTimes))
+	}
+
+	for i := range insertedTimes {
+		if foundTimes[i] != insertedTimes[i] {
+			t.Fatalf("Found %d expected %d", foundTimes[i], insertedTimes[i])
+		}
+	}
+
+	t.Logf("Found %d times, as expected", len(foundTimes))
+
+	// Before closing our connection, let's send a standby status to update our wal
+	// position, which should then be reflected if we fetch out our current wal position
+	// for the slot
+	status, err := pgx.NewStandbyStatus(maxWal)
+	if err != nil {
+		t.Errorf("Failed to create standby status %v", err)
+	}
+	replicationConn.SendStandbyStatus(status)
+	replicationConn.StopReplication()
+
+	// Let's push the boundary conditions of the standby status and ensure it errors correctly
+	status, err = pgx.NewStandbyStatus(0,1,2,3,4)
+	if err == nil {
+		t.Errorf("Expected error from new standby status, got %v",status)
+	}
+
+	// And if you provide 3 args, ensure the right fields are set
+	status, err = pgx.NewStandbyStatus(1,2,3)
+	if err != nil {
+		t.Errorf("Failed to create test status: %v", err)
+	}
+	if status.WalFlushPosition != 1 {
+		t.Errorf("Unexpected flush position %d", status.WalFlushPosition)
+	}
+	if status.WalApplyPosition != 2 {
+		t.Errorf("Unexpected apply position %d", status.WalApplyPosition)
+	}
+	if status.WalWritePosition != 3 {
+		t.Errorf("Unexpected write position %d", status.WalWritePosition)
+	}
+
+	err = replicationConn.Close()
+	if err != nil {
+		t.Fatalf("Replication connection close failed: %v", err)
+	}
+
+	restartLsn := getConfirmedFlushLsnFor(t, conn, "pgx_test")
+	integerRestartLsn, _ := pgx.ParseLSN(restartLsn)
+	if integerRestartLsn != maxWal {
+		t.Fatalf("Wal offset update failed, expected %s found %s", pgx.FormatLSN(maxWal), restartLsn)
+	}
+
+	_, err = conn.Exec("select pg_drop_replication_slot($1)", "pgx_test")
+	if err != nil {
+		t.Fatalf("Failed to drop replication slot: %v", err)
+	}
+
+}
@@ -30,7 +30,7 @@ func ensureConnValid(t *testing.T, db *sql.DB) {

 	rows, err := db.Query("select generate_series(1,$1)", 10)
 	if err != nil {
-		t.Fatalf("db.Query failed: ", err)
+		t.Fatalf("db.Query failed: %v", err)
 	}
 	defer rows.Close()

@@ -42,7 +42,7 @@ func ensureConnValid(t *testing.T, db *sql.DB) {
 	}

 	if rows.Err() != nil {
-		t.Fatalf("db.Query failed: ", err)
+		t.Fatalf("db.Query failed: %v", err)
 	}

 	if rowCount != 10 {