-
Notifications
You must be signed in to change notification settings - Fork 148
/
Copy pathstructured-streaming-stateful-stream-processing.html
363 lines (318 loc) · 18.1 KB
/
structured-streaming-stateful-stream-processing.html
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
<!doctype html>
<html lang="en">
<head>
<meta charset="utf-8">
<meta name="viewport" content="width=device-width, initial-scale=1.0, maximum-scale=1.0, user-scalable=no">
<title>Apache Spark™ Workshop | Structured Streaming | Stateful Stream Processing</title>
<meta name="description" content="Apache Spark™ Workshop | Structured Streaming | Stateful Stream Processing">
<meta name="author" content="Jacek Laskowski">
<link rel="stylesheet" href="reveal.js/css/reveal.css">
<link rel="stylesheet" href="reveal.js/css/theme/beige.css">
<!-- Theme used for syntax highlighting of code -->
<link rel="stylesheet" href="reveal.js/lib/css/zenburn.css">
<!-- Jacek: custom formatting -->
<link rel="stylesheet" href="revealjs-css/jacek.css">
<!-- Printing and PDF exports -->
<script>
var link = document.createElement('link');
link.rel = 'stylesheet';
link.type = 'text/css';
link.href = window.location.search.match(/print-pdf/gi) ? 'reveal.js/css/print/pdf.css' : 'reveal.js/css/print/paper.css';
document.getElementsByTagName('head')[0].appendChild(link);
</script>
</head>
<body>
<div class="reveal">
<div class="footer">
<footer style="font-size: small;">
© <a href="https://medium.com/@jaceklaskowski">Jacek Laskowski</a> 2019 / <a href="https://twitter.com/jaceklaskowski">@JacekLaskowski</a>
</footer>
</div>
<div class="slides">
<section class="intro" data-transition="zoom" id="home">
<p>
<img width="12%" style="background:none; border:none; box-shadow:none;" data-src="images/spark-logo.png">
<img width="6%" src="images/jacek_laskowski_20141201_512px.png" style="border: 0">
</p>
<h1 style="font-size: 3.17em;">Stateful Stream Processing</h1>
<h3>Apache Spark 2.4.3 / Structured Streaming</h3>
<hr />
<h4 style="font-size: smaller;">
<a href="https://twitter.com/jaceklaskowski">@jaceklaskowski</a> / <a href="https://stackoverflow.com/users/1305344/jacek-laskowski">StackOverflow</a> / <a href="https://github.com/jaceklaskowski">GitHub</a>
<br>
The "Internals" Books: <a href="https://bit.ly/apache-spark-internals">Apache Spark</a> / <a href="https://bit.ly/spark-sql-internals">Spark SQL</a> / <a href="https://bit.ly/spark-structured-streaming">Spark Structured Streaming</a>
</h4>
</section>
<section>
<section id="speaker" style="font-size: 90%" data-markdown>
<textarea data-template>
<p><img width="12%" src="images/jacek_laskowski_20141201_512px.png" style="border: 0"></p>
* **Jacek Laskowski** is a freelance IT consultant
* Specializing in **Spark**, Kafka, Kafka Streams, Scala
* Development | Consulting | Training
* Among contributors to <a href="https://github.com/apache/spark/graphs/contributors">Apache Spark</a>
* Contact me at **[email protected]**
* Follow [@JacekLaskowski](https://twitter.com/jaceklaskowski) on twitter <br>for more #ApacheSpark, #ApacheKafka, #KafkaStreams
</textarea>
</section>
<section id="gitbooks">
<p><img width="12%" src="images/jacek_laskowski_20141201_512px.png" style="border: 0"></p>
<div style="text-align: center">
<p>Jacek is best known by the online "Internals" books:</p>
<p>
<ol>
<li><a href="https://bit.ly/apache-spark-internals">The Internals of Apache Spark</a></li>
<li><a href="https://bit.ly/spark-sql-internals">The Internals of Spark SQL</a></li>
<li><a href="https://bit.ly/spark-structured-streaming">The Internals of Spark Structured Streaming</a></li>
<li><a href="https://bit.ly/kafka-streams-internals">The Internals of Kafka Streams</a></li>
<li><a href="https://bit.ly/apache-kafka-internals">The Internals of Apache Kafka</a></li>
</ol>
</p>
</div>
</section>
<section id="stackoverflow-spark-structured-streaming">
<div style="text-align: center">
<p>
Jacek is "active" on <a href="http://stackoverflow.com/users/1305344/jacek-laskowski">StackOverflow</a> 🥳
<br />
(<b>spark-structured-streaming</b> tag)
</p>
<p>
<a href="https://stackoverflow.com/tags/spark-structured-streaming/topusers"><img width="65%" src="images/jaceklaskowski-stackoverflow-spark-structured-streaming.png" style="border: 0"></a>
</p>
</div>
</section>
<section id="stackoverflow-apache-spark">
<div style="text-align: center">
<p>
Jacek is "active" on <a href="http://stackoverflow.com/users/1305344/jacek-laskowski">StackOverflow</a> 🥳
<br />
(<b>apache-spark</b> tag)
</p>
<p>
<a href="https://stackoverflow.com/tags/apache-spark/topusers"><img width="60%" src="images/jaceklaskowski-stackoverflow-apache-spark.png" style="border: 0"></a>
</p>
</div>
</section>
</section>
<section id="agenda" style="font-size: 90%" data-markdown>
<textarea data-template>
## Agenda
1. [Structured Streaming](#/intro)
1. [Stateful Stream Processing](#/stateful-stream-processing)
1. [Streaming Aggregation](#/streaming-aggregation)
1. [Streaming Watermark](#/streaming-watermark)
1. [Streaming Join](#/streaming-join)
1. [Arbitrary Stateful Streaming Aggregation](#/arbitrary-stateful-streaming-aggregation)
1. [State Store](#/state-store)
</textarea>
</section>
<section id="intro" data-markdown style="font-size: 80%">
<textarea data-template>
## Structured Streaming
1. **Structured Streaming** (aka **Spark Streams**) is the module of Apache Spark for stream processing
1. Introduces **streaming queries**
* Extension of batch queries in Spark SQL
* Executed by Spark SQL engine
1. Expressed using high-level declarative languages
* **Dataset API**
* **SQL**
* Exactly as in Spark SQL <small>(almost)</small>
1. Logical and physical query plans with operators are based on Spark SQL
1. Supports **fault-tolerant state stores** for stateful operations (e.g. windowed aggregations and joins)
</textarea>
</section>
<section>
<section id="stateful-stream-processing" data-markdown style="font-size: 80%">
<textarea data-template>
## Stateful Stream Processing <small>(1 of 2)</small>
1. **Stateful Stream Processing** is a stream processing with some state (implicit or explicit)
1. In Structured Streaming, a streaming query is stateful when it uses the following high-level declarative operators (or their SQL variants):
* **Dataset.groupBy** for [Streaming Aggregation](#/streaming-aggregation)
* **Dataset.rollup** for [Streaming Aggregation](#/streaming-aggregation)
* **Dataset.cube** for [Streaming Aggregation](#/streaming-aggregation)
* **Dataset.groupByKey** for [Streaming Aggregation](#/streaming-aggregation)
* **Dataset.dropDuplicates** for Streaming Deduplication <small>(_not covered_)</small>
* **Dataset.join** for [Streaming Join](#/streaming-join)
* **Dataset.limit** for Streaming Limit <small>(_not covered_)</small>
</textarea>
</section>
<section data-markdown style="font-size: 80%">
<textarea data-template>
## Stateful Stream Processing <small>(2 of 2)</small>
1. Any structured query (incl. streaming queries) becomes a logical query plan
* High-level declarative operators simply create an **unanalyzed logical plan**
1. Let's rephrase then what was said earlier
1. In Structured Streaming, a streaming query is stateful when it uses the following logical operators:
* **StateStoreRestoreExec** and **StateStoreSaveExec**
* **StreamingDeduplicateExec**
* **FlatMapGroupsWithStateExec**
* **StreamingSymmetricHashJoinExec**
* **StreamingGlobalLimitExec**
</textarea>
</section>
</section>
<section>
<section id="streaming-aggregation" data-markdown style="font-size: 80%">
<textarea data-template>
## Streaming Aggregation <small>(1 of 2)</small>
1. **Streaming Aggregation** is a streaming query that uses the following high-level declarative operators (or their SQL variants):
* **Dataset.groupBy**
* **Dataset.rollup**
* **Dataset.cube**
* **Dataset.groupByKey**
1. Creates **Aggregate** logical operator
1. **window** function allows for **windowed streaming aggregation**
* Tumbling windows
* Sliding windows
</textarea>
</section>
<section id="streaming-aggregation-demo" data-markdown style="font-size: 80%">
<textarea data-template>
## Streaming Aggregation's Demo <small>(2 of 2)</small>
1. Go to [Demo: Streaming Query for Running Counts](https://jaceklaskowski.gitbooks.io/spark-structured-streaming/spark-sql-streaming-demo-groupBy-running-count-complete.html)
</textarea>
</section>
</section>
<section>
<section id="streaming-watermark" data-markdown style="font-size: 80%">
<textarea data-template>
## Streaming Watermark <small>(1 of 2)</small>
1. **Streaming Watermark** (**Event-Time Watermark**) of a stateful streaming query is a time threshold to wait for late and possibly out-of-order events until a streaming state (for a given key) can be considered final
1. Used to mark events that are older than the threshold as "too late", and not "interesting" to update partial non-final aggregates
1. Avoids unbounded streaming state
* Prevents OOMEs
1. Expressed with **Dataset.withWatermark** high-level operator
* Defined on grouping expression(s) of a streaming aggregation (directly or using **window** standard function)
* Creates **EventTimeWatermark** logical operator
</textarea>
</section>
<section id="streaming-watermark-demo" data-markdown style="font-size: 80%">
<textarea data-template>
## Streaming Watermark's Demo <small>(2 of 2)</small>
1. Go to [Demo: Streaming Watermark with Aggregation in Append Output Mode](https://jaceklaskowski.gitbooks.io/spark-structured-streaming/spark-sql-streaming-demo-watermark-aggregation-append.html)
1. Go to [Demo: Streaming Aggregation with Kafka Data Source](https://jaceklaskowski.gitbooks.io/spark-structured-streaming/spark-sql-streaming-demo-kafka-data-source.html) for event-time windows
</textarea>
</section>
</section>
<section>
<section id="streaming-join" data-markdown style="font-size: 80%">
<textarea data-template>
## Streaming Join <small>(1 of 2)</small>
1. **Streaming Join** is a streaming query that uses **Dataset.join** or SQL's **JOIN** high-level query operators
1. Creates **Join** logical operator
1. Streaming joins can be **stateless** or **stateful**
* Joins of a streaming query and a batch query (aka **stream-static joins**) are stateless and no state management is necessary
* Joins of two streaming queries (aka **stream-stream joins**) are stateful and require streaming state (with a streaming watermark)
</textarea>
</section>
<section id="streaming-join-demo" data-markdown style="font-size: 80%">
<textarea data-template>
## Streaming Join's Demo <small>(2 of 2)</small>
1. Go to [Demo: Streaming Join of Streaming Queries and StreamingSymmetricHashJoinExec Physical Operator](https://jaceklaskowski.gitbooks.io/spark-structured-streaming/spark-sql-streaming-demo-join-stream-stream-StreamingSymmetricHashJoinExec.html)
</textarea>
</section>
</section>
<section>
<section id="arbitrary-stateful-streaming-aggregation" data-markdown style="font-size: 90%">
<textarea data-template>
## Arbitrary Stateful Streaming Aggregation <small>(1 of 2)</small>
1. **Arbitrary Stateful Streaming Aggregation** is a streaming query that uses the following **KeyValueGroupedDataset** operators:
* **mapGroupsWithState**
* **flatMapGroupsWithState**
1. Creates **FlatMapGroupsWithState** logical operator
</textarea>
</section>
<section data-markdown style="font-size: 90%">
<textarea data-template>
## Arbitrary Stateful Streaming Aggregation <small>(2 of 2)</small>
```scala
KeyValueGroupedDataset.flatMapGroupsWithState[S: Encoder, U: Encoder](
outputMode: OutputMode,
timeoutConf: GroupStateTimeout)(
func: (K, Iterator[V], GroupState[S]) => Iterator[U]): Dataset[U]
```
1. **GroupState** is a per-group state data
1. Every time the state function func is executed for a **K** key, the state (as **GroupState[S]**) is for this key only
1. **GroupStateTimeout** allows for a timeout based on processing time, event-time or no timeout at all
1. Requires **Append** or **Update** output modes
</textarea>
</section>
</section>
<section id="state-store" data-markdown style="font-size: 80%">
<textarea data-template>
## State Store
1. Structured Streaming uses versioned (and possibly fault-tolerant) **key-value state stores** for streaming state data
1. State stores are on executors in **state** directory under **checkpointLocation**
1. Supports **incremental checkpointing** - only the key-value "Row" pairs that changed can be committed or aborted (other pairs are not touched)
1. Every state store is **identified** (e.g. aggregating operator id, the partition id)
1. **HDFSBackedStateStore** is the default (and only) state store
* Uses a HDFS-compatible file system for versioned state persistence
1. [spark-state-tools](https://github.com/HeartSaVioR/spark-state-tools) for offline manipulation of state stores
</textarea>
</section>
<section id="recap" style="font-size: 90%" data-markdown>
<textarea data-template>
## Recap
1. [Structured Streaming](#/intro)
1. [Stateful Stream Processing](#/stateful-stream-processing)
1. [Streaming Aggregation](#/streaming-aggregation)
1. [Streaming Watermark](#/streaming-watermark)
1. [Streaming Join](#/streaming-join)
1. [Arbitrary Stateful Streaming Aggregation](#/arbitrary-stateful-streaming-aggregation)
1. [State Store](#/state-store)
</textarea>
</section>
<section style="text-align: left" data-markdown id="questions">
<textarea data-template>
# Questions?
* Read [The Internals of Apache Spark](https://bit.ly/apache-spark-internals)
* Read [The Internals of Spark SQL](https://bit.ly/spark-sql-internals)
* Read [The Internals of Spark Structured Streaming](https://bit.ly/spark-structured-streaming)
* Follow [@jaceklaskowski](https://twitter.com/jaceklaskowski) on twitter (DMs open)
* Upvote [my questions and answers on StackOverflow](http://stackoverflow.com/users/1305344/jacek-laskowski)
* Contact me at **[email protected]**
</textarea>
</section>
</div>
</div>
<script src="reveal.js/lib/js/head.min.js"></script>
<script src="reveal.js/js/reveal.js"></script>
<script>
// More info about config & dependencies:
// - https://github.com/hakimel/reveal.js#configuration
// - https://github.com/hakimel/reveal.js#dependencies
Reveal.initialize({
controls: true,
progress: true,
history: true,
center: true,
slideNumber: true,
transition: 'slide', // none/fade/slide/convex/concave/zoom
menu: {
markers: true,
openSlideNumber: true
},
dependencies: [
{ src: 'reveal.js/lib/js/classList.js', condition: function () { return !document.body.classList; } },
{ src: 'reveal.js/plugin/markdown/marked.js' },
{ src: 'reveal.js/plugin/markdown/markdown.js' },
{ src: 'reveal.js/plugin/zoom-js/zoom.js', async: true },
{ src: 'reveal.js/plugin/notes/notes.js', async: true },
{ src: 'reveal.js/plugin/highlight/highlight.js', async: true, callback: function () { hljs.initHighlightingOnLoad(); } }
]
});
</script>
<script>
(function (i, s, o, g, r, a, m) {
i['GoogleAnalyticsObject'] = r; i[r] = i[r] || function () {
(i[r].q = i[r].q || []).push(arguments)
}, i[r].l = 1 * new Date(); a = s.createElement(o),
m = s.getElementsByTagName(o)[0]; a.async = 1; a.src = g; m.parentNode.insertBefore(a, m)
})(window, document, 'script', '//www.google-analytics.com/analytics.js', 'ga');
ga('create', 'UA-45999426-3', 'auto');
ga('send', 'pageview');
</script>
</body>
</html>