wonderdog 0.0.1
Sign up to get free protection for your applications and to get access to all the features.
- data/.gitignore +49 -0
- data/.rspec +2 -0
- data/CHANGELOG.md +5 -0
- data/LICENSE.md +201 -0
- data/README.md +175 -0
- data/Rakefile +10 -0
- data/bin/estool +141 -0
- data/bin/estrus.rb +136 -0
- data/bin/wonderdog +93 -0
- data/config/elasticsearch-example.yml +227 -0
- data/config/elasticsearch.in.sh +52 -0
- data/config/logging.yml +43 -0
- data/config/more_settings.yml +60 -0
- data/config/run_elasticsearch-2.sh +42 -0
- data/config/ufo_config.json +12 -0
- data/lib/wonderdog.rb +14 -0
- data/lib/wonderdog/configuration.rb +25 -0
- data/lib/wonderdog/hadoop_invocation_override.rb +139 -0
- data/lib/wonderdog/index_and_mapping.rb +67 -0
- data/lib/wonderdog/timestamp.rb +43 -0
- data/lib/wonderdog/version.rb +3 -0
- data/notes/README-benchmarking.txt +272 -0
- data/notes/README-read_tuning.textile +74 -0
- data/notes/benchmarking-201011.numbers +0 -0
- data/notes/cluster_notes.md +17 -0
- data/notes/notes.txt +91 -0
- data/notes/pigstorefunc.pig +45 -0
- data/pom.xml +80 -0
- data/spec/spec_helper.rb +22 -0
- data/spec/support/driver_helper.rb +15 -0
- data/spec/support/integration_helper.rb +30 -0
- data/spec/wonderdog/hadoop_invocation_override_spec.rb +81 -0
- data/spec/wonderdog/index_and_type_spec.rb +73 -0
- data/src/main/java/com/infochimps/elasticsearch/ElasticSearchInputFormat.java +268 -0
- data/src/main/java/com/infochimps/elasticsearch/ElasticSearchOutputCommitter.java +39 -0
- data/src/main/java/com/infochimps/elasticsearch/ElasticSearchOutputFormat.java +283 -0
- data/src/main/java/com/infochimps/elasticsearch/ElasticSearchSplit.java +60 -0
- data/src/main/java/com/infochimps/elasticsearch/ElasticSearchStreamingInputFormat.java +231 -0
- data/src/main/java/com/infochimps/elasticsearch/ElasticSearchStreamingOutputCommitter.java +37 -0
- data/src/main/java/com/infochimps/elasticsearch/ElasticSearchStreamingOutputFormat.java +88 -0
- data/src/main/java/com/infochimps/elasticsearch/ElasticSearchStreamingRecordReader.java +176 -0
- data/src/main/java/com/infochimps/elasticsearch/ElasticSearchStreamingRecordWriter.java +171 -0
- data/src/main/java/com/infochimps/elasticsearch/ElasticSearchStreamingSplit.java +102 -0
- data/src/main/java/com/infochimps/elasticsearch/ElasticTest.java +108 -0
- data/src/main/java/com/infochimps/elasticsearch/hadoop/util/HadoopUtils.java +100 -0
- data/src/main/java/com/infochimps/elasticsearch/pig/ElasticSearchIndex.java +216 -0
- data/src/main/java/com/infochimps/elasticsearch/pig/ElasticSearchJsonIndex.java +235 -0
- data/src/main/java/com/infochimps/elasticsearch/pig/ElasticSearchStorage.java +355 -0
- data/test/foo.json +3 -0
- data/test/foo.tsv +3 -0
- data/test/test_dump.pig +19 -0
- data/test/test_json_loader.pig +21 -0
- data/test/test_tsv_loader.pig +16 -0
- data/wonderdog.gemspec +32 -0
- metadata +130 -0
@@ -0,0 +1,39 @@
|
|
1
|
+
package com.infochimps.elasticsearch;
|
2
|
+
|
3
|
+
import java.io.IOException;
|
4
|
+
|
5
|
+
import org.apache.hadoop.mapreduce.JobContext;
|
6
|
+
import org.apache.hadoop.mapreduce.OutputCommitter;
|
7
|
+
import org.apache.hadoop.mapreduce.TaskAttemptContext;
|
8
|
+
|
9
|
+
/**
|
10
|
+
* Small committer class that does not do anything.
|
11
|
+
*/
|
12
|
+
public class ElasticSearchOutputCommitter extends OutputCommitter {
|
13
|
+
|
14
|
+
@Override
|
15
|
+
public void abortTask(TaskAttemptContext arg0) throws IOException {
|
16
|
+
}
|
17
|
+
|
18
|
+
@Override
|
19
|
+
public void cleanupJob(JobContext arg0) throws IOException {
|
20
|
+
}
|
21
|
+
|
22
|
+
@Override
|
23
|
+
public void commitTask(TaskAttemptContext arg0) throws IOException {
|
24
|
+
}
|
25
|
+
|
26
|
+
@Override
|
27
|
+
public boolean needsTaskCommit(TaskAttemptContext arg0) throws IOException {
|
28
|
+
return false;
|
29
|
+
}
|
30
|
+
|
31
|
+
@Override
|
32
|
+
public void setupJob(JobContext arg0) throws IOException {
|
33
|
+
}
|
34
|
+
|
35
|
+
@Override
|
36
|
+
public void setupTask(TaskAttemptContext arg0) throws IOException {
|
37
|
+
}
|
38
|
+
|
39
|
+
}
|
@@ -0,0 +1,283 @@
|
|
1
|
+
package com.infochimps.elasticsearch;
|
2
|
+
|
3
|
+
import java.io.File;
|
4
|
+
import java.io.IOException;
|
5
|
+
import java.util.Map;
|
6
|
+
import java.util.HashMap;
|
7
|
+
import java.util.ArrayList;
|
8
|
+
import java.util.List;
|
9
|
+
import java.util.concurrent.atomic.AtomicLong;
|
10
|
+
import java.util.Random;
|
11
|
+
import java.net.URI;
|
12
|
+
|
13
|
+
import org.apache.commons.logging.Log;
|
14
|
+
import org.apache.commons.logging.LogFactory;
|
15
|
+
|
16
|
+
import org.apache.hadoop.conf.Configurable;
|
17
|
+
import org.apache.hadoop.io.*;
|
18
|
+
import org.apache.hadoop.mapreduce.TaskAttemptContext;
|
19
|
+
import org.apache.hadoop.mapreduce.JobContext;
|
20
|
+
import org.apache.hadoop.mapreduce.RecordWriter;
|
21
|
+
import org.apache.hadoop.conf.Configuration;
|
22
|
+
import org.apache.hadoop.fs.FileSystem;
|
23
|
+
import org.apache.hadoop.fs.Path;
|
24
|
+
import org.apache.hadoop.mapreduce.Counter;
|
25
|
+
import org.apache.hadoop.mapreduce.OutputFormat;
|
26
|
+
import org.apache.hadoop.mapreduce.OutputCommitter;
|
27
|
+
import org.apache.hadoop.filecache.DistributedCache;
|
28
|
+
|
29
|
+
import org.elasticsearch.common.xcontent.XContentBuilder;
|
30
|
+
import org.elasticsearch.common.xcontent.XContentFactory;
|
31
|
+
import org.elasticsearch.node.Node;
|
32
|
+
import org.elasticsearch.node.NodeBuilder;
|
33
|
+
import org.elasticsearch.client.Client;
|
34
|
+
import org.elasticsearch.client.Requests;
|
35
|
+
import org.elasticsearch.action.bulk.BulkRequestBuilder;
|
36
|
+
import org.elasticsearch.indices.IndexAlreadyExistsException;
|
37
|
+
import org.elasticsearch.action.bulk.BulkResponse;
|
38
|
+
import org.elasticsearch.ExceptionsHelper;
|
39
|
+
|
40
|
+
import com.infochimps.elasticsearch.hadoop.util.HadoopUtils;
|
41
|
+
|
42
|
+
/**
|
43
|
+
|
44
|
+
Hadoop OutputFormat for writing arbitrary MapWritables (essentially HashMaps) into Elasticsearch. Records are batched up and sent
|
45
|
+
in a one-hop manner to the elastic search data nodes that will index them.
|
46
|
+
|
47
|
+
*/
|
48
|
+
public class ElasticSearchOutputFormat extends OutputFormat<NullWritable, MapWritable> implements Configurable {
|
49
|
+
|
50
|
+
static Log LOG = LogFactory.getLog(ElasticSearchOutputFormat.class);
|
51
|
+
private Configuration conf = null;
|
52
|
+
|
53
|
+
protected class ElasticSearchRecordWriter extends RecordWriter<NullWritable, MapWritable> {
|
54
|
+
|
55
|
+
private Node node;
|
56
|
+
private Client client;
|
57
|
+
private String indexName;
|
58
|
+
private int bulkSize;
|
59
|
+
private int idField;
|
60
|
+
private String idFieldName;
|
61
|
+
private String objType;
|
62
|
+
private String[] fieldNames;
|
63
|
+
|
64
|
+
// Used for bookkeeping purposes
|
65
|
+
private AtomicLong totalBulkTime = new AtomicLong();
|
66
|
+
private AtomicLong totalBulkItems = new AtomicLong();
|
67
|
+
private Random randgen = new Random();
|
68
|
+
private long runStartTime = System.currentTimeMillis();
|
69
|
+
|
70
|
+
// For hadoop configuration
|
71
|
+
private static final String ES_CONFIG_NAME = "elasticsearch.yml";
|
72
|
+
private static final String ES_PLUGINS_NAME = "plugins";
|
73
|
+
private static final String ES_INDEX_NAME = "elasticsearch.index.name";
|
74
|
+
private static final String ES_BULK_SIZE = "elasticsearch.bulk.size";
|
75
|
+
private static final String ES_ID_FIELD_NAME = "elasticsearch.id.field.name";
|
76
|
+
private static final String ES_ID_FIELD = "elasticsearch.id.field";
|
77
|
+
private static final String ES_OBJECT_TYPE = "elasticsearch.object.type";
|
78
|
+
private static final String ES_CONFIG = "es.config";
|
79
|
+
private static final String ES_PLUGINS = "es.path.plugins";
|
80
|
+
|
81
|
+
// Other string constants
|
82
|
+
private static final String COMMA = ",";
|
83
|
+
private static final String SLASH = "/";
|
84
|
+
private static final String NO_ID_FIELD = "-1";
|
85
|
+
|
86
|
+
private volatile BulkRequestBuilder currentRequest;
|
87
|
+
|
88
|
+
/**
|
89
|
+
Instantiates a new RecordWriter for Elasticsearch
|
90
|
+
<p>
|
91
|
+
The properties that <b>MUST</b> be set in the hadoop Configuration object
|
92
|
+
are as follows:
|
93
|
+
<ul>
|
94
|
+
<li><b>elasticsearch.index.name</b> - The name of the elasticsearch index data will be written to. It does not have to exist ahead of time</li>
|
95
|
+
<li><b>elasticsearch.bulk.size</b> - The number of records to be accumulated into a bulk request before writing to elasticsearch.</li>
|
96
|
+
<li><b>elasticsearch.is_json</b> - A boolean indicating whether the records to be indexed are json records. If false the records are assumed to be tsv, in which case <b>elasticsearch.field.names</b> must be set and contain a comma separated list of field names</li>
|
97
|
+
<li><b>elasticsearch.object.type</b> - The type of objects being indexed</li>
|
98
|
+
<li><b>elasticsearch.config</b> - The full path the elasticsearch.yml. It is a local path and must exist on all machines in the hadoop cluster.</li>
|
99
|
+
<li><b>elasticsearch.plugins.dir</b> - The full path the elasticsearch plugins directory. It is a local path and must exist on all machines in the hadoop cluster.</li>
|
100
|
+
</ul>
|
101
|
+
<p>
|
102
|
+
The following fields depend on whether <b>elasticsearch.is_json</b> is true or false.
|
103
|
+
<ul>
|
104
|
+
<li><b>elasticsearch.id.field.name</b> - When <b>elasticsearch.is_json</b> is true, this is the name of a field in the json document that contains the document's id. If -1 is used then the document is assumed to have no id and one is assigned to it by elasticsearch.</li>
|
105
|
+
<li><b>elasticsearch.field.names</b> - When <b>elasticsearch.is_json</b> is false, this is a comma separated list of field names.</li>
|
106
|
+
<li><b>elasticsearch.id.field</b> - When <b>elasticsearch.is_json</b> is false, this is the numeric index of the field to use as the document id. If -1 is used the document is assumed to have no id and one is assigned to it by elasticsearch.</li>
|
107
|
+
</ul>
|
108
|
+
*/
|
109
|
+
public ElasticSearchRecordWriter(TaskAttemptContext context) {
|
110
|
+
Configuration conf = context.getConfiguration();
|
111
|
+
this.indexName = conf.get(ES_INDEX_NAME);
|
112
|
+
this.bulkSize = Integer.parseInt(conf.get(ES_BULK_SIZE));
|
113
|
+
this.idFieldName = conf.get(ES_ID_FIELD_NAME);
|
114
|
+
if (idFieldName.equals(NO_ID_FIELD)) {
|
115
|
+
LOG.info("Documents will be assigned ids by elasticsearch");
|
116
|
+
this.idField = -1;
|
117
|
+
} else {
|
118
|
+
LOG.info("Using field:["+idFieldName+"] for document ids");
|
119
|
+
}
|
120
|
+
this.objType = conf.get(ES_OBJECT_TYPE);
|
121
|
+
|
122
|
+
//
|
123
|
+
// Fetches elasticsearch.yml and the plugins directory from the distributed cache, or
|
124
|
+
// from the local config.
|
125
|
+
//
|
126
|
+
try {
|
127
|
+
String taskConfigPath = HadoopUtils.fetchFileFromCache(ES_CONFIG_NAME, conf);
|
128
|
+
LOG.info("Using ["+taskConfigPath+"] as es.config");
|
129
|
+
String taskPluginsPath = HadoopUtils.fetchArchiveFromCache(ES_PLUGINS_NAME, conf);
|
130
|
+
LOG.info("Using ["+taskPluginsPath+"] as es.plugins.dir");
|
131
|
+
System.setProperty(ES_CONFIG, taskConfigPath);
|
132
|
+
System.setProperty(ES_PLUGINS, taskPluginsPath+SLASH+ES_PLUGINS_NAME);
|
133
|
+
} catch (Exception e) {
|
134
|
+
System.setProperty(ES_CONFIG,conf.get(ES_CONFIG));
|
135
|
+
System.setProperty(ES_PLUGINS,conf.get(ES_PLUGINS));
|
136
|
+
}
|
137
|
+
|
138
|
+
start_embedded_client();
|
139
|
+
initialize_index(indexName);
|
140
|
+
currentRequest = client.prepareBulk();
|
141
|
+
}
|
142
|
+
|
143
|
+
/**
|
144
|
+
Closes the connection to elasticsearch. Any documents remaining in the bulkRequest object are indexed.
|
145
|
+
*/
|
146
|
+
public void close(TaskAttemptContext context) throws IOException {
|
147
|
+
if (currentRequest.numberOfActions() > 0) {
|
148
|
+
try {
|
149
|
+
BulkResponse response = currentRequest.execute().actionGet();
|
150
|
+
} catch (Exception e) {
|
151
|
+
LOG.warn("Bulk request failed: " + e.getMessage());
|
152
|
+
throw new RuntimeException(e);
|
153
|
+
}
|
154
|
+
}
|
155
|
+
LOG.info("Closing record writer");
|
156
|
+
client.close();
|
157
|
+
LOG.info("Client is closed");
|
158
|
+
if (node != null) {
|
159
|
+
node.close();
|
160
|
+
}
|
161
|
+
LOG.info("Record writer closed.");
|
162
|
+
}
|
163
|
+
|
164
|
+
/**
|
165
|
+
Writes a single MapWritable record to the bulkRequest object. Once <b>elasticsearch.bulk.size</b> are accumulated the
|
166
|
+
records are written to elasticsearch.
|
167
|
+
*/
|
168
|
+
public void write(NullWritable key, MapWritable fields) throws IOException {
|
169
|
+
XContentBuilder builder = XContentFactory.jsonBuilder();
|
170
|
+
buildContent(builder, fields);
|
171
|
+
if (idField == -1) {
|
172
|
+
// Document has no inherent id
|
173
|
+
currentRequest.add(Requests.indexRequest(indexName).type(objType).source(builder));
|
174
|
+
} else {
|
175
|
+
try {
|
176
|
+
Text mapKey = new Text(idFieldName);
|
177
|
+
String record_id = fields.get(mapKey).toString();
|
178
|
+
currentRequest.add(Requests.indexRequest(indexName).id(record_id).type(objType).create(false).source(builder));
|
179
|
+
} catch (Exception e) {
|
180
|
+
LOG.warn("Encountered malformed record");
|
181
|
+
}
|
182
|
+
}
|
183
|
+
processBulkIfNeeded();
|
184
|
+
}
|
185
|
+
|
186
|
+
/**
|
187
|
+
Recursively untangles the MapWritable and writes the fields into elasticsearch's XContentBuilder builder.
|
188
|
+
*/
|
189
|
+
private void buildContent(XContentBuilder builder, Writable value) throws IOException {
|
190
|
+
if (value instanceof Text) {
|
191
|
+
builder.value(((Text)value).toString());
|
192
|
+
} else if (value instanceof LongWritable) {
|
193
|
+
builder.value(((LongWritable)value).get());
|
194
|
+
} else if (value instanceof IntWritable) {
|
195
|
+
builder.value(((IntWritable)value).get());
|
196
|
+
} else if (value instanceof DoubleWritable) {
|
197
|
+
builder.value(((DoubleWritable)value).get());
|
198
|
+
} else if (value instanceof FloatWritable) {
|
199
|
+
builder.value(((FloatWritable)value).get());
|
200
|
+
} else if (value instanceof BooleanWritable) {
|
201
|
+
builder.value(((BooleanWritable)value).get());
|
202
|
+
} else if (value instanceof MapWritable) {
|
203
|
+
builder.startObject();
|
204
|
+
for (Map.Entry<Writable,Writable> entry : ((MapWritable)value).entrySet()) {
|
205
|
+
if (!(entry.getValue() instanceof NullWritable)) {
|
206
|
+
builder.field(entry.getKey().toString());
|
207
|
+
buildContent(builder, entry.getValue());
|
208
|
+
}
|
209
|
+
}
|
210
|
+
builder.endObject();
|
211
|
+
} else if (value instanceof ArrayWritable) {
|
212
|
+
builder.startArray();
|
213
|
+
Writable[] arrayOfThings = ((ArrayWritable)value).get();
|
214
|
+
for (int i = 0; i < arrayOfThings.length; i++) {
|
215
|
+
buildContent(builder, arrayOfThings[i]);
|
216
|
+
}
|
217
|
+
builder.endArray();
|
218
|
+
}
|
219
|
+
}
|
220
|
+
|
221
|
+
/**
|
222
|
+
Indexes content to elasticsearch when <b>elasticsearch.bulk.size</b> records have been accumulated.
|
223
|
+
*/
|
224
|
+
private void processBulkIfNeeded() {
|
225
|
+
totalBulkItems.incrementAndGet();
|
226
|
+
if (currentRequest.numberOfActions() >= bulkSize) {
|
227
|
+
try {
|
228
|
+
long startTime = System.currentTimeMillis();
|
229
|
+
BulkResponse response = currentRequest.execute().actionGet();
|
230
|
+
totalBulkTime.addAndGet(System.currentTimeMillis() - startTime);
|
231
|
+
if (randgen.nextDouble() < 0.1) {
|
232
|
+
LOG.info("Indexed [" + totalBulkItems.get() + "] in [" + (totalBulkTime.get()/1000) + "s] of indexing"+"[" + ((System.currentTimeMillis() - runStartTime)/1000) + "s] of wall clock"+" for ["+ (float)(1000.0*totalBulkItems.get())/(System.currentTimeMillis() - runStartTime) + "rec/s]");
|
233
|
+
}
|
234
|
+
} catch (Exception e) {
|
235
|
+
LOG.warn("Bulk request failed: " + e.getMessage());
|
236
|
+
throw new RuntimeException(e);
|
237
|
+
}
|
238
|
+
currentRequest = client.prepareBulk();
|
239
|
+
}
|
240
|
+
}
|
241
|
+
|
242
|
+
private void initialize_index(String indexName) {
|
243
|
+
LOG.info("Initializing index");
|
244
|
+
try {
|
245
|
+
client.admin().indices().prepareCreate(indexName).execute().actionGet();
|
246
|
+
} catch (Exception e) {
|
247
|
+
if (ExceptionsHelper.unwrapCause(e) instanceof IndexAlreadyExistsException) {
|
248
|
+
LOG.warn("Index ["+indexName+"] already exists");
|
249
|
+
}
|
250
|
+
}
|
251
|
+
}
|
252
|
+
|
253
|
+
//
|
254
|
+
// Starts an embedded elasticsearch client (ie. data = false)
|
255
|
+
//
|
256
|
+
private void start_embedded_client() {
|
257
|
+
LOG.info("Starting embedded elasticsearch client ...");
|
258
|
+
this.node = NodeBuilder.nodeBuilder().client(true).node();
|
259
|
+
this.client = node.client();
|
260
|
+
}
|
261
|
+
}
|
262
|
+
|
263
|
+
public RecordWriter<NullWritable, MapWritable> getRecordWriter(final TaskAttemptContext context) throws IOException, InterruptedException {
|
264
|
+
return new ElasticSearchRecordWriter(context);
|
265
|
+
}
|
266
|
+
|
267
|
+
public void setConf(Configuration conf) {
|
268
|
+
}
|
269
|
+
|
270
|
+
public Configuration getConf() {
|
271
|
+
return conf;
|
272
|
+
}
|
273
|
+
|
274
|
+
@Override
|
275
|
+
public void checkOutputSpecs(JobContext context) throws IOException, InterruptedException {
|
276
|
+
// TODO Check if the object exists?
|
277
|
+
}
|
278
|
+
|
279
|
+
@Override
|
280
|
+
public OutputCommitter getOutputCommitter(TaskAttemptContext context) throws IOException, InterruptedException {
|
281
|
+
return new ElasticSearchOutputCommitter();
|
282
|
+
}
|
283
|
+
}
|
@@ -0,0 +1,60 @@
|
|
1
|
+
package com.infochimps.elasticsearch;
|
2
|
+
|
3
|
+
import java.io.IOException;
|
4
|
+
import java.io.DataInput;
|
5
|
+
import java.io.DataOutput;
|
6
|
+
|
7
|
+
import org.apache.hadoop.io.Text;
|
8
|
+
import org.apache.hadoop.io.Writable;
|
9
|
+
import org.apache.hadoop.mapreduce.InputSplit;
|
10
|
+
|
11
|
+
public class ElasticSearchSplit extends InputSplit implements Writable {
|
12
|
+
|
13
|
+
private String queryString;
|
14
|
+
private long from;
|
15
|
+
private long size;
|
16
|
+
|
17
|
+
public ElasticSearchSplit() {}
|
18
|
+
|
19
|
+
public ElasticSearchSplit(String queryString, long from, long size) {
|
20
|
+
this.queryString = queryString;
|
21
|
+
this.from = from;
|
22
|
+
this.size = size;
|
23
|
+
}
|
24
|
+
|
25
|
+
public String getQueryString() {
|
26
|
+
return queryString;
|
27
|
+
}
|
28
|
+
|
29
|
+
public long getFrom() {
|
30
|
+
return from;
|
31
|
+
}
|
32
|
+
|
33
|
+
public long getSize() {
|
34
|
+
return size;
|
35
|
+
}
|
36
|
+
|
37
|
+
@Override
|
38
|
+
public String[] getLocations() {
|
39
|
+
return new String[] {};
|
40
|
+
}
|
41
|
+
|
42
|
+
@Override
|
43
|
+
public long getLength() {
|
44
|
+
return 0;
|
45
|
+
}
|
46
|
+
|
47
|
+
@Override
|
48
|
+
public void readFields(DataInput in) throws IOException {
|
49
|
+
queryString = Text.readString(in);
|
50
|
+
from = in.readLong();
|
51
|
+
size = in.readLong();
|
52
|
+
}
|
53
|
+
|
54
|
+
@Override
|
55
|
+
public void write(DataOutput out) throws IOException {
|
56
|
+
Text.writeString(out, queryString);
|
57
|
+
out.writeLong(from);
|
58
|
+
out.writeLong(size);
|
59
|
+
}
|
60
|
+
}
|
@@ -0,0 +1,231 @@
|
|
1
|
+
package com.infochimps.elasticsearch;
|
2
|
+
|
3
|
+
import java.io.IOException;
|
4
|
+
import java.io.BufferedReader;
|
5
|
+
import java.io.FileReader;
|
6
|
+
import java.util.List;
|
7
|
+
import java.util.ArrayList;
|
8
|
+
import java.util.Map;
|
9
|
+
import java.util.HashMap;
|
10
|
+
|
11
|
+
import org.apache.commons.logging.Log;
|
12
|
+
import org.apache.commons.logging.LogFactory;
|
13
|
+
|
14
|
+
import org.apache.hadoop.io.*;
|
15
|
+
import org.apache.hadoop.mapred.InputFormat;
|
16
|
+
import org.apache.hadoop.mapred.RecordReader;
|
17
|
+
import org.apache.hadoop.mapred.InputSplit;
|
18
|
+
import org.apache.hadoop.mapred.JobConf;
|
19
|
+
import org.apache.hadoop.mapred.Reporter;
|
20
|
+
|
21
|
+
import org.elasticsearch.common.settings.loader.YamlSettingsLoader;
|
22
|
+
import org.elasticsearch.common.transport.InetSocketTransportAddress;
|
23
|
+
|
24
|
+
import org.elasticsearch.client.transport.TransportClient;
|
25
|
+
import org.elasticsearch.action.search.SearchRequestBuilder;
|
26
|
+
|
27
|
+
import org.elasticsearch.action.search.SearchResponse;
|
28
|
+
import org.elasticsearch.action.search.SearchType;
|
29
|
+
import org.elasticsearch.index.query.FilterBuilders.*;
|
30
|
+
import org.elasticsearch.cluster.ClusterName;
|
31
|
+
|
32
|
+
public class ElasticSearchStreamingInputFormat<K, V> implements InputFormat<K, V> {
|
33
|
+
|
34
|
+
static Log LOG = LogFactory.getLog(ElasticSearchStreamingInputFormat.class);
|
35
|
+
|
36
|
+
// Job settings we need to control directly from Java options.
|
37
|
+
private static final String ES_INDEX_OPT = "elasticsearch.input.index";
|
38
|
+
private static final String ES_DEFAULT_INDEX = "hadoop";
|
39
|
+
private String indexName;
|
40
|
+
|
41
|
+
private static final String ES_MAPPING_OPT = "elasticsearch.input.mapping";
|
42
|
+
private static final String ES_DEFAULT_MAPPING = "streaming_record";
|
43
|
+
private String mappingName;
|
44
|
+
|
45
|
+
private static final String ES_NUM_SPLITS_OPT = "elasticsearch.input.splits";
|
46
|
+
private static final String ES_NUM_SPLITS = "1";
|
47
|
+
private Integer numSplits;
|
48
|
+
|
49
|
+
private static final String ES_QUERY_OPT = "elasticsearch.input.query";
|
50
|
+
private static final String ES_QUERY = "{\"match_all\": {}}";
|
51
|
+
private String queryJSON;
|
52
|
+
|
53
|
+
// Calculated after the first query.
|
54
|
+
private long numHits;
|
55
|
+
private Integer recordsPerSplit;
|
56
|
+
|
57
|
+
// Elasticsearch internal settings required to make a client
|
58
|
+
// connection.
|
59
|
+
private static final String ES_CONFIG_OPT = "es.config";
|
60
|
+
private static final String ES_CONFIG = "/etc/elasticsearch/elasticsearch.yml";
|
61
|
+
|
62
|
+
private static final String ES_PLUGINS_OPT = "es.path.plugins";
|
63
|
+
private static final String ES_PLUGINS = "/usr/local/share/elasticsearch/plugins";
|
64
|
+
|
65
|
+
private static final String ES_UNICAST_HOSTS_NAME = "discovery.zen.ping.unicast.hosts";
|
66
|
+
|
67
|
+
private TransportClient client;
|
68
|
+
|
69
|
+
public RecordReader<K, V> getRecordReader(InputSplit split, JobConf conf, Reporter reporter) {
|
70
|
+
setLocalElasticSearchInstallation(conf);
|
71
|
+
return (RecordReader) new ElasticSearchStreamingRecordReader(split, conf);
|
72
|
+
}
|
73
|
+
|
74
|
+
public InputSplit[] getSplits(JobConf conf, int requestedNumSplits) {
|
75
|
+
this.numSplits = requestedNumSplits;
|
76
|
+
|
77
|
+
setLocalElasticSearchInstallation(conf);
|
78
|
+
parseInput(conf);
|
79
|
+
|
80
|
+
startTransportClient(conf);
|
81
|
+
findNumHits();
|
82
|
+
stopTransportClient();
|
83
|
+
|
84
|
+
return createSplits();
|
85
|
+
}
|
86
|
+
|
87
|
+
|
88
|
+
//
|
89
|
+
// == Setup ==
|
90
|
+
//
|
91
|
+
|
92
|
+
public void setLocalElasticSearchInstallation(JobConf conf) {
|
93
|
+
String esConfigPath = conf.get(ES_CONFIG_OPT, ES_CONFIG);
|
94
|
+
String esPluginsPath = conf.get(ES_PLUGINS_OPT, ES_PLUGINS);
|
95
|
+
System.setProperty(ES_CONFIG_OPT, esConfigPath);
|
96
|
+
System.setProperty(ES_PLUGINS_OPT,esPluginsPath);
|
97
|
+
LOG.info("Using Elasticsearch configuration file at "+esConfigPath+" and plugin directory "+esPluginsPath);
|
98
|
+
}
|
99
|
+
|
100
|
+
private void parseInput(JobConf conf) {
|
101
|
+
this.indexName = conf.get(ES_INDEX_OPT, ES_DEFAULT_INDEX);
|
102
|
+
this.mappingName = conf.get(ES_MAPPING_OPT, ES_DEFAULT_MAPPING);
|
103
|
+
// this.numSplits = Integer.parseInt(conf.get(ES_NUM_SPLITS_OPT, ES_NUM_SPLITS));
|
104
|
+
this.queryJSON = conf.get(ES_QUERY_OPT, ES_QUERY);
|
105
|
+
String message = "Using input /"+indexName;
|
106
|
+
if (mappingName != null && mappingName.length() > 0) {
|
107
|
+
message += "/"+mappingName;
|
108
|
+
}
|
109
|
+
if (queryJSON != null && queryJSON.length() > 0) {
|
110
|
+
message += " with query: "+queryJSON;
|
111
|
+
}
|
112
|
+
LOG.info(message);
|
113
|
+
}
|
114
|
+
|
115
|
+
//
|
116
|
+
// == Connecting to Elasticsearch and Querying ==
|
117
|
+
//
|
118
|
+
|
119
|
+
private void startTransportClient(JobConf conf) {
|
120
|
+
this.client = new TransportClient();
|
121
|
+
Map<String,String> settings = parsedSettings(conf);
|
122
|
+
String host = hostname(settings);
|
123
|
+
if (host.toString().length() == 0) {
|
124
|
+
System.exit(1);
|
125
|
+
}
|
126
|
+
LOG.info("Attempting to connect to Elasticsearch node at " + host + ":9300");
|
127
|
+
this.client = new TransportClient().addTransportAddress(new InetSocketTransportAddress(host, 9300));
|
128
|
+
LOG.info("Connected to Elasticsearch cluster");
|
129
|
+
}
|
130
|
+
|
131
|
+
private Map<String,String> parsedSettings(JobConf conf) {
|
132
|
+
String esConfigPath = conf.get(ES_CONFIG_OPT, ES_CONFIG);
|
133
|
+
String esPluginsPath = conf.get(ES_PLUGINS_OPT, ES_PLUGINS);
|
134
|
+
|
135
|
+
try {
|
136
|
+
BufferedReader reader = new BufferedReader( new FileReader(esConfigPath));
|
137
|
+
String line = null;
|
138
|
+
StringBuilder stringBuilder = new StringBuilder();
|
139
|
+
String ls = System.getProperty("line.separator");
|
140
|
+
while( ( line = reader.readLine() ) != null ) {
|
141
|
+
stringBuilder.append( line );
|
142
|
+
stringBuilder.append( ls );
|
143
|
+
}
|
144
|
+
return new YamlSettingsLoader().load(stringBuilder.toString());
|
145
|
+
} catch (IOException e) {
|
146
|
+
LOG.error("Could not find or read the configuration file " + esConfigPath + ".");
|
147
|
+
return new HashMap<String,String>();
|
148
|
+
}
|
149
|
+
}
|
150
|
+
|
151
|
+
private String hostname(Map<String,String> settings) {
|
152
|
+
String hostsString = settings.get(ES_UNICAST_HOSTS_NAME);
|
153
|
+
if (hostsString.toString().length() == 0) {
|
154
|
+
LOG.error("Could not find hosts. Did you set the '" + ES_UNICAST_HOSTS_NAME + "' key?");
|
155
|
+
return "";
|
156
|
+
}
|
157
|
+
|
158
|
+
String[] hosts = hostsString.split(",");
|
159
|
+
if (hosts.length > 0) {
|
160
|
+
String host = hosts[0];
|
161
|
+
if (host.toString().length() == 0) {
|
162
|
+
LOG.error("Could not parse hosts from '" + ES_UNICAST_HOSTS_NAME + "' key.");
|
163
|
+
return "";
|
164
|
+
} else {
|
165
|
+
return host;
|
166
|
+
}
|
167
|
+
} else {
|
168
|
+
LOG.error("Could not find any hosts in the '" + ES_UNICAST_HOSTS_NAME + "' key.");
|
169
|
+
return "";
|
170
|
+
}
|
171
|
+
}
|
172
|
+
|
173
|
+
private void stopTransportClient() {
|
174
|
+
if (client != null) client.close();
|
175
|
+
LOG.info("Disconnected from Elasticsearch cluster");
|
176
|
+
}
|
177
|
+
|
178
|
+
private void findNumHits() {
|
179
|
+
SearchRequestBuilder request = client.prepareSearch(indexName);
|
180
|
+
if (mappingName != null && mappingName.length() > 0) {
|
181
|
+
request.setTypes(mappingName);
|
182
|
+
}
|
183
|
+
request.setSearchType(SearchType.COUNT);
|
184
|
+
if (queryJSON != null && queryJSON.length() > 0) {
|
185
|
+
request.setQuery(queryJSON);
|
186
|
+
}
|
187
|
+
SearchResponse response = request.execute().actionGet();
|
188
|
+
this.numHits = response.hits().totalHits();
|
189
|
+
|
190
|
+
LOG.info("Ran query: "+String.valueOf(numHits)+" hits");
|
191
|
+
}
|
192
|
+
|
193
|
+
//
|
194
|
+
// == Setting splits ==
|
195
|
+
//
|
196
|
+
|
197
|
+
private void readjustSplitsByHits() {
|
198
|
+
|
199
|
+
}
|
200
|
+
|
201
|
+
private InputSplit[] createSplits() {
|
202
|
+
// Say that
|
203
|
+
//
|
204
|
+
// numHits = 7
|
205
|
+
// numSplits = 2
|
206
|
+
if((long) numSplits > numHits) {
|
207
|
+
numSplits = (int) numHits;
|
208
|
+
}
|
209
|
+
|
210
|
+
this.recordsPerSplit = (int) (numHits/((long)numSplits)); // == 3 records/split
|
211
|
+
|
212
|
+
List<InputSplit> splits = new ArrayList<InputSplit>(numSplits);
|
213
|
+
|
214
|
+
// i == 0, 1
|
215
|
+
for(int i = 0; i < numSplits; i++) {
|
216
|
+
Integer from = i * recordsPerSplit;
|
217
|
+
splits.add(new ElasticSearchStreamingSplit(indexName, mappingName, numSplits, queryJSON, numHits, from, recordsPerSplit));
|
218
|
+
}
|
219
|
+
// 7 is > (2 * 3) == 6
|
220
|
+
if (numHits > ((long) (numSplits * recordsPerSplit))) {
|
221
|
+
Integer from = numSplits * recordsPerSplit;
|
222
|
+
Integer size = (int) (numHits - ((long) from));
|
223
|
+
splits.add(new ElasticSearchStreamingSplit(indexName, mappingName, numSplits, queryJSON, numHits, from, size));
|
224
|
+
}
|
225
|
+
|
226
|
+
LOG.info("Splitting "+String.valueOf(numHits)+" hits across "+String.valueOf(splits.size())+" splits ("+String.valueOf(recordsPerSplit)+" hits/split)");
|
227
|
+
|
228
|
+
return splits.toArray(new InputSplit[splits.size()]);
|
229
|
+
}
|
230
|
+
|
231
|
+
}
|