forked from influxdata/kapacitor
-
Notifications
You must be signed in to change notification settings - Fork 0
/
udf.go
427 lines (379 loc) · 9.03 KB
/
udf.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
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
package kapacitor
import (
"bufio"
"fmt"
"io"
"log"
"net"
"sync"
"time"
"github.com/cenkalti/backoff"
"github.com/influxdata/kapacitor/command"
"github.com/influxdata/kapacitor/models"
"github.com/influxdata/kapacitor/pipeline"
"github.com/influxdata/kapacitor/udf"
"github.com/pkg/errors"
)
// User defined function
type UDFNode struct {
node
u *pipeline.UDFNode
udf udf.Interface
aborted chan struct{}
wg sync.WaitGroup
mu sync.Mutex
stopped bool
}
// Create a new UDFNode that sends incoming data to child udf
func newUDFNode(et *ExecutingTask, n *pipeline.UDFNode, l *log.Logger) (*UDFNode, error) {
un := &UDFNode{
node: node{Node: n, et: et, logger: l},
u: n,
aborted: make(chan struct{}),
}
// Create the UDF
f, err := et.tm.UDFService.Create(
n.UDFName,
l,
un.abortedCallback,
)
if err != nil {
return nil, err
}
un.udf = f
un.node.runF = un.runUDF
un.node.stopF = un.stopUDF
return un, nil
}
var errNodeAborted = errors.New("node aborted")
func (u *UDFNode) stopUDF() {
u.mu.Lock()
defer u.mu.Unlock()
if !u.stopped {
u.stopped = true
if u.udf != nil {
u.udf.Abort(errNodeAborted)
}
}
}
func (u *UDFNode) runUDF(snapshot []byte) (err error) {
defer func() {
u.mu.Lock()
defer u.mu.Unlock()
//Ignore stopped errors if the udf was stopped externally
if u.stopped && (err == udf.ErrServerStopped || err == errNodeAborted) {
err = nil
}
u.stopped = true
}()
err = u.udf.Open()
if err != nil {
return
}
err = u.udf.Init(u.u.Options)
if err != nil {
return
}
if snapshot != nil {
err = u.udf.Restore(snapshot)
if err != nil {
return
}
}
forwardErr := make(chan error, 1)
go func() {
switch u.Provides() {
case pipeline.StreamEdge:
pointOut := u.udf.PointOut()
for p := range pointOut {
for _, out := range u.outs {
err := out.CollectPoint(p)
if err != nil {
forwardErr <- err
return
}
}
}
case pipeline.BatchEdge:
batchOut := u.udf.BatchOut()
for b := range batchOut {
for _, out := range u.outs {
err := out.CollectBatch(b)
if err != nil {
forwardErr <- err
return
}
}
}
}
forwardErr <- nil
}()
// The abort callback needs to know when we are done writing
// so we wrap in a wait group.
u.wg.Add(1)
go func() {
defer u.wg.Done()
switch u.Wants() {
case pipeline.StreamEdge:
pointIn := u.udf.PointIn()
for p, ok := u.ins[0].NextPoint(); ok; p, ok = u.ins[0].NextPoint() {
u.timer.Start()
select {
case pointIn <- p:
case <-u.aborted:
return
}
u.timer.Stop()
}
case pipeline.BatchEdge:
batchIn := u.udf.BatchIn()
for b, ok := u.ins[0].NextBatch(); ok; b, ok = u.ins[0].NextBatch() {
u.timer.Start()
select {
case batchIn <- b:
case <-u.aborted:
return
}
u.timer.Stop()
}
}
}()
// wait till we are done writing
u.wg.Wait()
// Close the udf
err = u.udf.Close()
if err != nil {
return
}
// Wait/Return any error from the forwarding goroutine
err = <-forwardErr
return
}
func (u *UDFNode) abortedCallback() {
close(u.aborted)
// wait till we are done writing
u.wg.Wait()
}
func (u *UDFNode) snapshot() ([]byte, error) {
return u.udf.Snapshot()
}
// UDFProcess wraps an external process and sends and receives data
// over STDIN and STDOUT. Lines received over STDERR are logged
// via normal Kapacitor logging.
type UDFProcess struct {
server *udf.Server
commander command.Commander
cmdSpec command.Spec
cmd command.Command
stderr io.Reader
// Group for waiting on the process itself
processGroup sync.WaitGroup
logStdErrGroup sync.WaitGroup
mu sync.Mutex
logger *log.Logger
timeout time.Duration
abortCallback func()
}
func NewUDFProcess(
commander command.Commander,
cmdSpec command.Spec,
l *log.Logger,
timeout time.Duration,
abortCallback func(),
) *UDFProcess {
return &UDFProcess{
commander: commander,
cmdSpec: cmdSpec,
logger: l,
timeout: timeout,
abortCallback: abortCallback,
}
}
// Open the UDFProcess
func (p *UDFProcess) Open() error {
p.mu.Lock()
defer p.mu.Unlock()
cmd := p.commander.NewCommand(p.cmdSpec)
stdin, err := cmd.StdinPipe()
if err != nil {
return err
}
stdout, err := cmd.StdoutPipe()
if err != nil {
return err
}
stderr, err := cmd.StderrPipe()
if err != nil {
return err
}
p.stderr = stderr
err = cmd.Start()
if err != nil {
return err
}
p.cmd = cmd
outBuf := bufio.NewReader(stdout)
p.server = udf.NewServer(
outBuf,
stdin,
p.logger,
p.timeout,
p.abortCallback,
cmd.Kill,
)
p.server.Start()
p.logStdErrGroup.Add(1)
go p.logStdErr()
// Wait for process to terminate
p.processGroup.Add(1)
go func() {
// First wait for the pipe read writes to finish
p.logStdErrGroup.Wait()
p.server.WaitIO()
err := cmd.Wait()
if err != nil {
err = fmt.Errorf("process exited unexpectedly: %v", err)
defer p.server.Abort(err)
}
p.processGroup.Done()
}()
return nil
}
// Stop the UDFProcess cleanly.
//
// Calling Close should only be done once the owner has stopped writing to the *In channel,
// at which point the remaining data will be processed and the subprocess will be allowed to exit cleanly.
func (p *UDFProcess) Close() error {
p.mu.Lock()
defer p.mu.Unlock()
err := p.server.Stop()
p.processGroup.Wait()
return err
}
// Replay any lines from STDERR of the process to the Kapacitor log.
func (p *UDFProcess) logStdErr() {
defer p.logStdErrGroup.Done()
scanner := bufio.NewScanner(p.stderr)
for scanner.Scan() {
p.logger.Println("I!P", scanner.Text())
}
}
func (p *UDFProcess) Abort(err error) { p.server.Abort(err) }
func (p *UDFProcess) Init(options []*udf.Option) error { return p.server.Init(options) }
func (p *UDFProcess) Snapshot() ([]byte, error) { return p.server.Snapshot() }
func (p *UDFProcess) Restore(snapshot []byte) error { return p.server.Restore(snapshot) }
func (p *UDFProcess) PointIn() chan<- models.Point { return p.server.PointIn() }
func (p *UDFProcess) BatchIn() chan<- models.Batch { return p.server.BatchIn() }
func (p *UDFProcess) PointOut() <-chan models.Point { return p.server.PointOut() }
func (p *UDFProcess) BatchOut() <-chan models.Batch { return p.server.BatchOut() }
func (p *UDFProcess) Info() (udf.Info, error) { return p.server.Info() }
type UDFSocket struct {
server *udf.Server
socket Socket
logger *log.Logger
timeout time.Duration
abortCallback func()
}
type Socket interface {
Open() error
Close() error
In() io.WriteCloser
Out() io.Reader
}
func NewUDFSocket(
socket Socket,
l *log.Logger,
timeout time.Duration,
abortCallback func(),
) *UDFSocket {
return &UDFSocket{
socket: socket,
logger: l,
timeout: timeout,
abortCallback: abortCallback,
}
}
func (s *UDFSocket) Open() error {
err := s.socket.Open()
if err != nil {
return err
}
in := s.socket.In()
out := s.socket.Out()
outBuf := bufio.NewReader(out)
s.server = udf.NewServer(
outBuf,
in,
s.logger,
s.timeout,
s.abortCallback,
func() { s.socket.Close() },
)
return s.server.Start()
}
func (s *UDFSocket) Close() error {
if err := s.server.Stop(); err != nil {
// Always close the socket
s.socket.Close()
return errors.Wrap(err, "stopping UDF server")
}
if err := s.socket.Close(); err != nil {
return errors.Wrap(err, "closing UDF socket connection")
}
return nil
}
func (s *UDFSocket) Abort(err error) { s.server.Abort(err) }
func (s *UDFSocket) Init(options []*udf.Option) error { return s.server.Init(options) }
func (s *UDFSocket) Snapshot() ([]byte, error) { return s.server.Snapshot() }
func (s *UDFSocket) Restore(snapshot []byte) error { return s.server.Restore(snapshot) }
func (s *UDFSocket) PointIn() chan<- models.Point { return s.server.PointIn() }
func (s *UDFSocket) BatchIn() chan<- models.Batch { return s.server.BatchIn() }
func (s *UDFSocket) PointOut() <-chan models.Point { return s.server.PointOut() }
func (s *UDFSocket) BatchOut() <-chan models.Batch { return s.server.BatchOut() }
func (s *UDFSocket) Info() (udf.Info, error) { return s.server.Info() }
type socket struct {
path string
conn *net.UnixConn
}
func NewSocketConn(path string) Socket {
return &socket{
path: path,
}
}
func (s *socket) Open() error {
addr, err := net.ResolveUnixAddr("unix", s.path)
if err != nil {
return err
}
// Connect to socket
b := backoff.NewExponentialBackOff()
b.MaxElapsedTime = time.Minute * 5
err = backoff.Retry(func() error {
conn, err := net.DialUnix("unix", nil, addr)
if err != nil {
return err
}
s.conn = conn
return nil
},
b,
)
return err
}
func (s *socket) Close() error {
return s.conn.Close()
}
type unixCloser struct {
*net.UnixConn
}
func (u unixCloser) Close() error {
// Only close the write end of the socket connection.
// The socket connection as a whole will be closed later.
return u.CloseWrite()
}
func (s *socket) In() io.WriteCloser {
return unixCloser{s.conn}
}
func (s *socket) Out() io.Reader {
return s.conn
}