forked from asonawalla/gazette
/
replicate_client.go
213 lines (189 loc) · 5.46 KB
/
replicate_client.go
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
package gazette
import (
"bufio"
"bytes"
"errors"
"fmt"
"io"
"io/ioutil"
"net"
"net/http"
"net/http/httputil"
"net/url"
"strconv"
"sync"
"time"
log "github.com/sirupsen/logrus"
"github.com/LiveRamp/gazette/pkg/httpdump"
"github.com/LiveRamp/gazette/pkg/journal"
"github.com/LiveRamp/gazette/pkg/keepalive"
)
const (
CommitDeltaHeader = "X-Commit-Delta"
FragmentLastModifiedHeader = "X-Fragment-Last-Modified"
FragmentLocationHeader = "X-Fragment-Location"
FragmentNameHeader = "X-Fragment-Name"
RouteTokenHeader = "X-Route-Token"
WriteHeadHeader = "X-Write-Head"
ReplicateClientIdlePoolSize = 6
)
type ReplicateClient struct {
endpoint *CachedURL
idlePool chan replicaClientConn
}
type replicaClientConn struct {
raw net.Conn
buf *bufio.ReadWriter
}
type replicaClientTransaction struct {
client ReplicateClient
chunker io.WriteCloser
conn replicaClientConn
request *http.Request
}
func NewReplicateClient(ep *CachedURL) ReplicateClient {
// Use a global map of pools keyed on |ep.Base| to facilitate connection
// re-use.
idlePoolsMu.Lock()
idlePool, ok := idlePools[ep.Base]
if !ok {
idlePool = make(chan replicaClientConn, ReplicateClientIdlePoolSize)
idlePools[ep.Base] = idlePool
}
idlePoolsMu.Unlock()
return ReplicateClient{
endpoint: ep,
idlePool: idlePool,
}
}
func (c ReplicateClient) Replicate(op journal.ReplicateOp) {
transaction := replicaClientTransaction{client: c}
go transaction.start(op)
}
func (t *replicaClientTransaction) start(op journal.ReplicateOp) {
conn, err := t.takeConn()
if err != nil {
op.Result <- journal.ReplicateResult{Error: err}
return
}
req, _ := t.client.endpoint.NewHTTPRequest("REPLICATE", "/"+op.Journal.String(), nil)
queryArgs := url.Values{
"newSpool": {strconv.FormatBool(op.NewSpool)},
"writeHead": {strconv.FormatInt(op.WriteHead, 10)},
"routeToken": {string(op.RouteToken)},
}
req.URL.RawQuery = queryArgs.Encode()
req.Header.Add("Expect", "100-continue")
req.Header.Add("Trailer", CommitDeltaHeader)
req.TransferEncoding = []string{"chunked"}
reqBytes, err := httpdump.DumpRequest(req, false)
if err != nil {
op.Result <- journal.ReplicateResult{Error: err}
return
}
// Flush the replication request to the peer.
conn.buf.Write(reqBytes)
if err = conn.buf.Flush(); err != nil {
op.Result <- journal.ReplicateResult{Error: err}
return
}
// Wait up to a minute for a 100-continue response.
// TODO(johnny): HTTP/2 to peers would remove the need for this timeout.
conn.raw.SetReadDeadline(time.Now().Add(time.Minute))
resp, err := http.ReadResponse(conn.buf.Reader, req)
if err != nil {
op.Result <- journal.ReplicateResult{Error: err}
return
} else if resp.StatusCode != http.StatusContinue {
var body bytes.Buffer
io.Copy(&body, resp.Body)
var remoteWriteHead int64
if s := resp.Header.Get(WriteHeadHeader); s != "" {
remoteWriteHead, err = strconv.ParseInt(s, 16, 64)
if err != nil {
log.WithFields(log.Fields{"err": err, "arg": s}).
Error("failed to parse replica head")
}
}
// Finish the request by writing an empty chunk and trailing headers.
conn.buf.WriteString("0\r\n\r\n")
if err := conn.buf.Flush(); !resp.Close && err == nil {
// Connection is still okay. Retain for the next round.
t.putConn(conn)
}
op.Result <- journal.ReplicateResult{Error: errors.New(body.String()),
ErrorWriteHead: remoteWriteHead}
return
}
// We've now opened a transaction stream.
conn.raw.SetReadDeadline(time.Time{}) // Clear timeout.
t.chunker = httputil.NewChunkedWriter(conn.buf)
t.conn = conn
t.request = req
op.Result <- journal.ReplicateResult{Writer: t}
return
}
func (t *replicaClientTransaction) takeConn() (replicaClientConn, error) {
var conn replicaClientConn
select {
case conn = <-t.client.idlePool:
return conn, nil
default:
}
url, err := t.client.endpoint.ResolveURL()
if err != nil {
return replicaClientConn{}, err
}
raw, err := keepalive.Dialer.Dial("tcp", url.Host)
if err != nil {
t.client.endpoint.InvalidateResolution()
return replicaClientConn{}, err
}
return replicaClientConn{raw,
bufio.NewReadWriter(bufio.NewReader(raw), bufio.NewWriter(raw))}, nil
}
func (t *replicaClientTransaction) putConn(conn replicaClientConn) {
conn.raw.SetReadDeadline(time.Time{}) // Clear timeout.
select {
case t.client.idlePool <- conn:
default:
}
}
func (t *replicaClientTransaction) Write(p []byte) (n int, err error) {
return t.chunker.Write(p)
}
func (t *replicaClientTransaction) Commit(delta int64) error {
// Close the chunker and write the commit delta as a trailing header.
t.chunker.Close()
fmt.Fprintf(t.conn.buf, "%s: %x\r\n\r\n", CommitDeltaHeader, delta)
if err := t.conn.buf.Flush(); err != nil {
return err
}
// Wait up to a minute for a commit response.
// TODO(johnny): HTTP/2 to peers would remove the need for this timeout.
t.conn.raw.SetReadDeadline(time.Now().Add(time.Minute))
resp, err := http.ReadResponse(t.conn.buf.Reader, t.request)
if err != nil {
return err
}
// Success is indicated by 204 No Content.
if resp.StatusCode != http.StatusNoContent {
var body bytes.Buffer
io.Copy(&body, resp.Body)
err = errors.New(body.String())
} else {
io.Copy(ioutil.Discard, resp.Body)
}
if !resp.Close {
t.putConn(t.conn)
}
return err
}
var (
// Pool idle connections keyed on Base of an endpoint.
idlePools map[string]chan replicaClientConn
idlePoolsMu sync.Mutex
)
func init() {
idlePools = make(map[string]chan replicaClientConn)
}