source: other-projects/hathitrust/wcsa/extracted-features-solr/trunk/solr-ingest/src/main/java/org/hathitrust/extractedfeatures/ProcessForSolrIngest.java@ 31028

Last change on this file since 31028 was 31028, checked in by davidb, 7 years ago

Support for randonly choosing Solr endpoints added in
  • Property svn:executable set to *
File size: 9.9 KB
Line 
1package org.hathitrust.extractedfeatures;
2
3import java.io.BufferedInputStream;
4import java.io.FileInputStream;
5import java.io.FileNotFoundException;
6import java.io.IOException;
7import java.io.Serializable;
8import java.util.ArrayList;
9
10import org.apache.commons.cli.*;
11
12import org.apache.spark.api.java.*;
13import org.apache.spark.util.DoubleAccumulator;
14import org.apache.spark.util.LongAccumulator;
15import org.hathitrust.extractedfeatures.PerPageJSONFlatmap;
16import org.json.JSONObject;
17import org.apache.spark.SparkConf;
18
19public class ProcessForSolrIngest implements Serializable
20{
21 private static final long serialVersionUID = 1L;
22
23 // Following details on number of partitions to use given in
24 // "Parallelized collections" section of:
25 // https://spark.apache.org/docs/2.0.1/programming-guide.html
26 //
27 // For a more detailed discussion see:
28 // http://blog.cloudera.com/blog/2015/03/how-to-tune-your-apache-spark-jobs-part-2/
29
30 public static final int NUM_CORES = 6;
31 public static final int NUM_PARTITIONS = 2*NUM_CORES; // default would appear to be 2
32
33 protected String _input_dir;
34 protected String _json_list_filename;
35 protected String _solr_url;
36 protected String _output_dir;
37
38 protected int _verbosity;
39
40 public ProcessForSolrIngest(String input_dir, String json_list_filename,
41 String solr_url, String output_dir, int verbosity)
42 {
43 _input_dir = input_dir;
44 _json_list_filename = (json_list_filename != null) ? json_list_filename : input_dir;
45
46 _solr_url = solr_url;
47 _output_dir = output_dir;
48 _verbosity = verbosity;
49 }
50
51 protected String generateSparkAppName(String exec_mode)
52 {
53 String spark_app_name = "[" + exec_mode + "] Extract Features: Process for Solr Ingest";
54 spark_app_name += " [" + _json_list_filename + "]";
55
56 if (_solr_url != null) {
57 spark_app_name += " solr_url=" + _solr_url;
58 }
59
60 if (_output_dir != null) {
61 spark_app_name += " output_dir=" + _output_dir;
62 }
63
64 return spark_app_name;
65 }
66
67 public void execPerVolume()
68 {
69 String spark_app_name = generateSparkAppName("Per Volume");
70
71 SparkConf conf = new SparkConf().setAppName(spark_app_name);
72 JavaSparkContext jsc = new JavaSparkContext(conf);
73
74 if (_verbosity >= 2) {
75 System.out.println("Default Minimum Partions: " + jsc.defaultMinPartitions());
76 System.out.println("Default Parallelism: " + jsc.defaultParallelism());
77 }
78
79 JavaRDD<String> json_list_data = jsc.textFile(_json_list_filename,NUM_PARTITIONS).cache();
80
81 long num_volumes = json_list_data.count();
82 double per_vol = 100.0/(double)num_volumes;
83
84 DoubleAccumulator progress_accum = jsc.sc().doubleAccumulator("Progress Percent");
85
86 System.err.println();
87 System.err.println();
88 System.err.println();
89 System.err.println("****##### _input_dir = " + _input_dir);
90 System.err.println();
91 System.err.println();
92 System.err.println();
93
94 PerVolumeJSON per_vol_json = new PerVolumeJSON(_input_dir,_solr_url,_output_dir,_verbosity, progress_accum,per_vol);
95
96 json_list_data.foreach(per_vol_json);
97
98 long num_ids = num_volumes;
99
100 System.out.println("");
101 System.out.println("############");
102 System.out.println("# Number of volume ids: " + num_ids);
103 System.out.println("############");
104 System.out.println("");
105
106 jsc.close();
107 }
108 public ArrayList<String> extrapolateSolrEndpoints()
109 {
110 ArrayList<String> solr_endpoints = new ArrayList<String>();
111
112 String solr_endpoint_ips = System.getProperty("wcsa-ef-ingest.solr-endpoint-ips",null);
113 if (solr_endpoint_ips != null) {
114 String [] ips = solr_endpoint_ips.split(",");
115 for (String ip : ips) {
116 String solr_endpoint = _solr_url.replaceFirst("//.*?:", "//"+ip+":");
117 solr_endpoints.add(solr_endpoint);
118 }
119 }
120 else {
121 if (_solr_url != null) {
122 solr_endpoints.add(_solr_url);
123 }
124 }
125
126 return solr_endpoints;
127 }
128
129 public void execPerPage()
130 {
131 String spark_app_name = generateSparkAppName("Per Page");
132
133 SparkConf conf = new SparkConf().setAppName(spark_app_name);
134 JavaSparkContext jsc = new JavaSparkContext(conf);
135
136 if (_verbosity >= 2) {
137 System.out.println("Default Minimum Partions: " + jsc.defaultMinPartitions());
138 System.out.println("Default Parallelism: " + jsc.defaultParallelism());
139 }
140
141 JavaRDD<String> json_list_data = jsc.textFile(_json_list_filename,NUM_PARTITIONS).cache();
142
143 long num_volumes = json_list_data.count();
144 double per_vol = 100.0/(double)num_volumes;
145
146 DoubleAccumulator per_vol_progress_accum = jsc.sc().doubleAccumulator("Per Volume Progress Percent");
147
148 PerPageJSONFlatmap paged_solr_json_flatmap
149 = new PerPageJSONFlatmap(_input_dir,_solr_url,_output_dir,_verbosity, per_vol_progress_accum,per_vol);
150 JavaRDD<JSONObject> per_page_jsonobjects = json_list_data.flatMap(paged_solr_json_flatmap).cache();
151
152 //long num_page_ids = per_page_jsonobjects.count(); // trigger lazy eval of: flatmap:per-vol
153
154 LongAccumulator per_page_progress_accum = jsc.sc().longAccumulator("Pages Processed");
155 ArrayList<String> solr_endpoints = extrapolateSolrEndpoints();
156
157 PerPageJSONMap paged_json_id_map
158 = new PerPageJSONMap(_input_dir,solr_endpoints,_output_dir,_verbosity, per_page_progress_accum,1);
159 JavaRDD<String> per_page_ids = per_page_jsonobjects.map(paged_json_id_map);
160
161/*
162 System.out.println("");
163 System.out.println("############");
164 System.out.println("# Progress Accumulator: " + progress_accum.value());
165 System.out.println("############");
166 System.out.println("");
167*/
168
169 long num_page_ids = per_page_ids.count(); // trigger lazy eval of: flatmap:per-vol -> map:per-page
170
171 System.out.println("");
172 System.out.println("############");
173 System.out.println("# Number of page ids: " + num_page_ids);
174 System.out.println("############");
175 System.out.println("");
176
177 /*
178 if (_output_dir != null) {
179 String rdd_save_file = "rdd-solr-json-page-files";
180 json_ids.saveAsTextFile(rdd_save_file);
181 System.out.println("############");
182 System.out.println("# Saved RDD of Solr JSON page files, top-level, as:");
183 System.out.println("# " + rdd_save_file);
184 System.out.println("############");
185 System.out.println("");
186 }
187 */
188
189 jsc.close();
190 }
191
192
193
194
195 public static void print_usage(HelpFormatter formatter, Options options)
196 {
197 formatter.printHelp("RUN.bash [options] input-dir json-filelist.txt", options);
198 }
199
200 public static void main(String[] args) {
201 Options options = new Options();
202
203 Option verbosity_opt = new Option("v", "verbosity", true,
204 "Set to control the level of debugging output [0=none, 1=some, 2=lots]");
205 verbosity_opt.setRequired(false);
206 options.addOption(verbosity_opt);
207
208 Option properties_opt = new Option("p", "properties", true,
209 "Read in the specified Java properties file");
210 properties_opt.setRequired(false);
211 options.addOption(properties_opt);
212
213 Option output_dir_opt = new Option("o", "output-dir", true,
214 "If specified, save BZipped Solr JSON files to this directory");
215 output_dir_opt.setRequired(false);
216 options.addOption(output_dir_opt);
217
218 Option solr_url_opt = new Option("u", "solr-url", true,
219 "If specified, the URL to post the Solr JSON data to");
220 solr_url_opt.setRequired(false);
221 options.addOption(solr_url_opt);
222
223 Option read_only_opt = new Option("r", "read-only", false,
224 "Used to initiate a run where the files are all read in, but nothing is ingested/saved");
225 read_only_opt.setRequired(false);
226 options.addOption(read_only_opt);
227
228 // Need to work with CLI v1.2 as this is the JAR that is bundled with Hadoop/Spark
229 CommandLineParser parser = new GnuParser();
230 //CommandLineParser parser = new DefaultParser(); // if working with CLI v1.3 and above
231
232 HelpFormatter formatter = new HelpFormatter();
233 CommandLine cmd = null;
234
235 try {
236 cmd = parser.parse(options, args);
237 }
238 catch (ParseException e) {
239 System.err.println(e.getMessage());
240 print_usage(formatter,options);
241 System.exit(1);
242 }
243
244
245 String verbosity_str = cmd.getOptionValue("verbosity","1");
246 int verbosity = Integer.parseInt(verbosity_str);
247
248 String property_filename = cmd.getOptionValue("properties",null);
249
250 String output_dir = cmd.getOptionValue("output-dir",null);
251 String solr_url = cmd.getOptionValue("solr-url",null);
252 boolean read_only = cmd.hasOption("read-only");
253
254 String[] filtered_args = cmd.getArgs();
255
256 if (filtered_args.length != 2) {
257 print_usage(formatter,options);
258 System.exit(1);
259 }
260
261 if (property_filename != null) {
262 try {
263 FileInputStream fis = new FileInputStream(property_filename);
264 BufferedInputStream bis = new BufferedInputStream(fis);
265
266 System.getProperties().load(bis);
267 }
268 catch (FileNotFoundException e) {
269 // TODO Auto-generated catch block
270 e.printStackTrace();
271 System.err.println("File not found: '" + property_filename + "'. Skipping property file read");
272 }
273 catch (IOException e) {
274 System.err.println("IO Exception for: '" + property_filename + "'. Malformed syntax? Skipping property file read");
275 }
276 }
277
278 if (!read_only && ((output_dir == null) && (solr_url==null))) {
279 System.err.println("Need to specify either --solr-url or --output-dir otherwise generated files are not ingested/saved");
280 print_usage(formatter,options);
281 System.exit(1);
282 }
283 if (read_only) {
284 // For this case, need to ensure solr-url and output-dir are null
285 output_dir = null;
286 solr_url = null;
287 }
288
289 String input_dir = filtered_args[0];
290 String json_list_filename = filtered_args[1];
291
292 ProcessForSolrIngest prep_for_ingest
293 = new ProcessForSolrIngest(input_dir,json_list_filename,solr_url,output_dir,verbosity);
294
295 String process_ef_json_mode = System.getProperty("wcsa-ef-ingest.process-ef-json-mode","per-page");
296 if (process_ef_json_mode.equals("per-volume")) {
297 prep_for_ingest.execPerVolume();
298 }
299 else {
300 prep_for_ingest.execPerPage();
301 }
302 }
303}
Note: See TracBrowser for help on using the repository browser.