1 | #!/usr/bin/perl
|
---|
2 |
|
---|
3 | use strict;
|
---|
4 | use warnings;
|
---|
5 |
|
---|
6 | # Requires setup.bash to have been sourced
|
---|
7 | BEGIN
|
---|
8 | {
|
---|
9 | die "GSDLHOME not set\n" unless defined $ENV{'GSDLHOME'};
|
---|
10 | die "GSDLOS not set\n" unless defined $ENV{'GSDLOS'};
|
---|
11 | # Ensure Greenstone Perl locations are in INC
|
---|
12 | unshift (@INC, $ENV{'GSDLHOME'} . '/perllib');
|
---|
13 | unshift (@INC, $ENV{'GSDLHOME'} . '/perllib/cpan');
|
---|
14 | # we'll need the perl version number
|
---|
15 | my ($version_number) = `perl-version.pl`;
|
---|
16 | if (defined $ENV{'GSDLEXTS'})
|
---|
17 | {
|
---|
18 | my @extensions = split(/:/,$ENV{'GSDLEXTS'});
|
---|
19 | foreach my $e (@extensions)
|
---|
20 | {
|
---|
21 | my $ext_prefix = $ENV{'GSDLHOME'} . '/ext/' . $e;
|
---|
22 | unshift (@INC, $ext_prefix . '/perllib');
|
---|
23 | unshift (@INC, $ext_prefix . '/perllib/cpan');
|
---|
24 | unshift (@INC, $ext_prefix . '/' . $ENV{'GSDLOS'} . '/lib/perl/' . $version_number);
|
---|
25 | }
|
---|
26 | }
|
---|
27 | }
|
---|
28 |
|
---|
29 | use Devel::Peek;
|
---|
30 | use Sort::Key::Natural qw(natsort);
|
---|
31 | use Time::Local;
|
---|
32 |
|
---|
33 | print "\n===== GS Hadoop Report =====\n";
|
---|
34 | print "Generate a report (CSV) from the output of a parallel processing Green-\n";
|
---|
35 | print "stone import, suitable for feeding into the Gantt chart generator.\n";
|
---|
36 | print "============================\n\n";
|
---|
37 |
|
---|
38 | # Configuration
|
---|
39 | if (!defined $ARGV[0] || !-d $ARGV[0])
|
---|
40 | {
|
---|
41 | die("usage: hadoop_report.pl <path to results>\n");
|
---|
42 | }
|
---|
43 | my $results_path = $ARGV[0];
|
---|
44 |
|
---|
45 | # Read in hadoop.log and parse top level record
|
---|
46 | print ' * Reading and parsing "hadoop.log"... ';
|
---|
47 | my $job_record = {'host'=>'', 'job'=>'', 'start'=>0, 'end'=>0, 'cpu_time'=>0};
|
---|
48 | my $hadoop_log_path = $results_path . '/hadoop.log';
|
---|
49 | if (open(HLIN, '<:utf8', $hadoop_log_path))
|
---|
50 | {
|
---|
51 | while (my $line = <HLIN>)
|
---|
52 | {
|
---|
53 | if ($line =~ /host:(.+)/)
|
---|
54 | {
|
---|
55 | $job_record->{'host'} = $1;
|
---|
56 | }
|
---|
57 | elsif ($line =~ /Running job: job_(\d+_\d+)/)
|
---|
58 | {
|
---|
59 | $job_record->{'job'} = $1;
|
---|
60 | }
|
---|
61 | elsif ($line =~ /CPU time spent \(ms\)=(\d+)/)
|
---|
62 | {
|
---|
63 | $job_record->{'cpu_time'} = $1;
|
---|
64 | }
|
---|
65 | elsif ($job_record->{'start'} == 0 && $line =~ /(\d\d)\/(\d\d)\/(\d\d) (\d\d):(\d\d):(\d\d)/)
|
---|
66 | {
|
---|
67 | $job_record->{'start'} = timelocal($6, $5, $4, $3, ($2 - 1), $1);
|
---|
68 | }
|
---|
69 | elsif ($line =~ /(\d\d)\/(\d\d)\/(\d\d) (\d\d):(\d\d):(\d\d)/)
|
---|
70 | {
|
---|
71 | my $end = timelocal($6, $5, $4, $3, ($2 - 1), $1);
|
---|
72 | if ($end > $job_record->{'end'})
|
---|
73 | {
|
---|
74 | $job_record->{'end'} = $end;
|
---|
75 | }
|
---|
76 | }
|
---|
77 | }
|
---|
78 | close(HLIN);
|
---|
79 | if ($job_record->{'start'} == 0 || $job_record->{'end'} == 0)
|
---|
80 | {
|
---|
81 | die('Error! Failed to parse timing information from log: ' . $hadoop_log_path);
|
---|
82 | }
|
---|
83 | }
|
---|
84 | else
|
---|
85 | {
|
---|
86 | die('Error! Failed to open file for reading: ' . $hadoop_log_path);
|
---|
87 | }
|
---|
88 | print "Done!\n";
|
---|
89 |
|
---|
90 | # Read in data_locality.csv (will be matched to task logs)
|
---|
91 | my $data_was_local = {};
|
---|
92 | my $data_locality_csv_path = $results_path . '/data_locality.csv';
|
---|
93 | if (-f $data_locality_csv_path)
|
---|
94 | {
|
---|
95 | print ' * Reading and parsing "data_locality.csv"... ';
|
---|
96 | if (open(DLIN, '<:utf8', $data_locality_csv_path))
|
---|
97 | {
|
---|
98 | while (my $line = <DLIN>)
|
---|
99 | {
|
---|
100 | # note that the line may begin with a taskid or just a taskno (legacy)
|
---|
101 | if ($line =~ /(\d+),\d,(\d)/)
|
---|
102 | {
|
---|
103 | $data_was_local->{$1} = $2;
|
---|
104 | }
|
---|
105 | }
|
---|
106 | close(DLIN);
|
---|
107 | }
|
---|
108 | else
|
---|
109 | {
|
---|
110 | die('Error! Failed to open file for reading: ' . $data_locality_csv_path);
|
---|
111 | }
|
---|
112 | print "Done!\n";
|
---|
113 | }
|
---|
114 | else
|
---|
115 | {
|
---|
116 | print " * Data locality not available or not applicable\n";
|
---|
117 | }
|
---|
118 |
|
---|
119 | # Read in all task logs and parse task records
|
---|
120 | my $task_records;
|
---|
121 | print " * Locating task logs...\n";
|
---|
122 | if (opendir(DH, $results_path))
|
---|
123 | {
|
---|
124 | my @files = readdir(DH);
|
---|
125 | foreach my $file (sort @files)
|
---|
126 | {
|
---|
127 | if ($file =~ /import-hadoop-(\d+_\d+)_m_(\d+)_\d+\.log/)
|
---|
128 | {
|
---|
129 | my $job_no = $1;
|
---|
130 | my $task_no = $2;
|
---|
131 | my $is_data_local = 0;
|
---|
132 | if (defined ($data_was_local->{$task_no}))
|
---|
133 | {
|
---|
134 | $is_data_local = $data_was_local->{$task_no};
|
---|
135 | }
|
---|
136 | my $task_record = {'host'=>'', 'cpu'=>0, 'job' => $job_no, 'task' => $task_no, 'start'=>0, 'end'=>0, 'cpu_time'=>0, 'data_locality'=>$is_data_local, 'file'=>'', 'percom'=>'NA'};
|
---|
137 | print ' - Reading and parsing "' . $file . '"... ';
|
---|
138 | my $task_log_path = $results_path . '/' . $file;
|
---|
139 | my $io_time = 0;
|
---|
140 | if (open(TIN, '<:utf8', $task_log_path))
|
---|
141 | {
|
---|
142 | my $io_start_time = 0;
|
---|
143 | while (my $line = <TIN>)
|
---|
144 | {
|
---|
145 | if ($line =~ /\[Started:(\d+(?:\.\d+)?)\]/)
|
---|
146 | {
|
---|
147 | $task_record->{'start'} = $1;
|
---|
148 | }
|
---|
149 | elsif ($line =~ /\[Host:([^\]]+)\]/)
|
---|
150 | {
|
---|
151 | $task_record->{'host'} = $1;
|
---|
152 | }
|
---|
153 | elsif ($line =~ /\[CPU:(\d+)\]/)
|
---|
154 | {
|
---|
155 | $task_record->{'cpu'} = $1;
|
---|
156 | }
|
---|
157 | elsif ($line =~ /\[Map:([^\>]+)=>/)
|
---|
158 | {
|
---|
159 | $task_record->{'file'} = $1;
|
---|
160 | }
|
---|
161 | elsif ($line =~ /\[IOS:(\d+)\]/)
|
---|
162 | {
|
---|
163 | $io_start_time = $1;
|
---|
164 | }
|
---|
165 | elsif ($io_start_time > 0 && $line =~ /\[IOE:(\d+)\]/)
|
---|
166 | {
|
---|
167 | $io_time += ($1 - $io_start_time);
|
---|
168 | $io_start_time = 0;
|
---|
169 | }
|
---|
170 | elsif ($line =~ /\[Completed:(\d+(?:\.\d+)?)\]/)
|
---|
171 | {
|
---|
172 | my $end_time = $1;
|
---|
173 | $task_record->{'end'} = $end_time;
|
---|
174 | if ($io_start_time > 0)
|
---|
175 | {
|
---|
176 | $io_time += ($end_time - $io_start_time);
|
---|
177 | }
|
---|
178 | }
|
---|
179 | }
|
---|
180 | close(TIN);
|
---|
181 | # Calculate CPU time (total time - IO time)
|
---|
182 | $task_record->{'cpu_time'} = $task_record->{'end'} - $task_record->{'start'} - $io_time;
|
---|
183 |
|
---|
184 | # We should now have the filename - use this and try and locate a
|
---|
185 | # convert log for this item (assuming it is multimedia, which it may
|
---|
186 | # not be)
|
---|
187 | if (defined $task_record->{'file'} && $task_record->{'file'} =~ /\/([^\/]+)\.ts/)
|
---|
188 | {
|
---|
189 | my $filename_sans_extension = $1;
|
---|
190 | my $convert_log = $results_path . '/convert-' . $filename_sans_extension . '.log';
|
---|
191 | if (-f $convert_log)
|
---|
192 | {
|
---|
193 | print '[Reading and parsing convert log]... ';
|
---|
194 | if (open(CLIN, '<:utf8', $convert_log))
|
---|
195 | {
|
---|
196 | my $max_percent = 0.00;
|
---|
197 | while (my $line = <CLIN>)
|
---|
198 | {
|
---|
199 | if ($line =~ /.*Encoding: task 1 of 1, (\d+\.\d\d) \%/)
|
---|
200 | {
|
---|
201 | my $percent = $1;
|
---|
202 | if ($percent > $max_percent)
|
---|
203 | {
|
---|
204 | $max_percent = $percent;
|
---|
205 | }
|
---|
206 | }
|
---|
207 | }
|
---|
208 | close(CLIN);
|
---|
209 | $task_record->{'percom'} = $max_percent;
|
---|
210 | }
|
---|
211 | else
|
---|
212 | {
|
---|
213 | print STDERR "Warning! Failed to open log file for reading: " . $convert_log . "\n";
|
---|
214 | }
|
---|
215 | }
|
---|
216 | }
|
---|
217 |
|
---|
218 | # Store this record
|
---|
219 | $task_records->{$task_no} = $task_record;
|
---|
220 | }
|
---|
221 | else
|
---|
222 | {
|
---|
223 | die('Error! Failed to open file for reading: ' . $task_log_path);
|
---|
224 | }
|
---|
225 | print "Done!\n";
|
---|
226 | }
|
---|
227 | }
|
---|
228 | close(DH);
|
---|
229 | }
|
---|
230 | else
|
---|
231 | {
|
---|
232 | die('Error! Failed to open directory for reading: ' . $results_path);
|
---|
233 | }
|
---|
234 | print ' - Processed ' . scalar(keys(%{$task_records})) . " records\n";
|
---|
235 |
|
---|
236 | # Generate compute-node records
|
---|
237 | print ' * Generating compute node records... ';
|
---|
238 | my $node_records;
|
---|
239 | foreach my $taskno (sort keys %{$task_records})
|
---|
240 | {
|
---|
241 | my $task_record = $task_records->{$taskno};
|
---|
242 | my $node_record = {'host'=>'', 'cpu'=>0, 'job' => '', 'start'=>0, 'end'=>0, 'cpu_time'=>0};
|
---|
243 | # - retrieve any existing record
|
---|
244 | my $worker_id = $task_record->{'host'} . '#' . $task_record->{'cpu'};
|
---|
245 | if (defined $node_records->{$worker_id})
|
---|
246 | {
|
---|
247 | $node_record = $node_records->{$worker_id};
|
---|
248 | }
|
---|
249 | if ($node_record->{'host'} eq '')
|
---|
250 | {
|
---|
251 | $node_record->{'host'} = $task_record->{'host'};
|
---|
252 | }
|
---|
253 | if ($node_record->{'cpu'} == 0)
|
---|
254 | {
|
---|
255 | $node_record->{'cpu'} = $task_record->{'cpu'};
|
---|
256 | }
|
---|
257 | if ($node_record->{'job'} eq '')
|
---|
258 | {
|
---|
259 | $node_record->{'job'} = $task_record->{'job'};
|
---|
260 | }
|
---|
261 | if ($node_record->{'start'} == 0 || $task_record->{'start'} < $node_record->{'start'})
|
---|
262 | {
|
---|
263 | $node_record->{'start'} = $task_record->{'start'};
|
---|
264 | }
|
---|
265 | if ($node_record->{'end'} == 0 || $task_record->{'end'} > $node_record->{'end'})
|
---|
266 | {
|
---|
267 | $node_record->{'end'} = $task_record->{'end'};
|
---|
268 | }
|
---|
269 | $node_record->{'cpu_time'} += $task_record->{'cpu_time'};
|
---|
270 | # - store it
|
---|
271 | $node_records->{$worker_id} = $node_record;
|
---|
272 | }
|
---|
273 | print "Done!\n";
|
---|
274 |
|
---|
275 | # Write out CSV of all information
|
---|
276 | my $report_csv_path = $results_path . '/timing.csv';
|
---|
277 | if (open(CSVOUT, '>:utf8', $report_csv_path))
|
---|
278 | {
|
---|
279 | my $row_counter = 1;
|
---|
280 | # Header
|
---|
281 | print CSVOUT "id,name,hostname,start,end,cputime,dl,pid,filename,percom\n";
|
---|
282 | # Master Record
|
---|
283 | print CSVOUT $row_counter . ',M0,' . $job_record->{'host'} . ',' . $job_record->{'start'} . ',' . $job_record->{'end'} . ',' . ($job_record->{'cpu_time'} / 1000) . ",0,0,NA,NA\n";
|
---|
284 | $row_counter++;
|
---|
285 | # For each compute node record
|
---|
286 | my $known_workers = {};
|
---|
287 | foreach my $worker_id (natsort(keys(%{$node_records})))
|
---|
288 | {
|
---|
289 | my $node_record = $node_records->{$worker_id};
|
---|
290 | my $node_id = $row_counter;
|
---|
291 | $row_counter++;
|
---|
292 | my $csv_worker_id = 'W' . $node_record->{'cpu'};
|
---|
293 | # Ensure we haven't used this id before - this should never trigger for
|
---|
294 | # multicore CPUs, but will for clusters (as nearly all nodes will report
|
---|
295 | # themselves as 'W0')
|
---|
296 | if (defined $known_workers->{$csv_worker_id})
|
---|
297 | {
|
---|
298 | # Find a different worker id as this one is already in use
|
---|
299 | my $counter = 0;
|
---|
300 | $csv_worker_id = 'W' . $counter;
|
---|
301 | while (defined $known_workers->{$csv_worker_id})
|
---|
302 | {
|
---|
303 | $counter++;
|
---|
304 | $csv_worker_id = 'W' . $counter;
|
---|
305 | }
|
---|
306 | }
|
---|
307 | $known_workers->{$csv_worker_id} = 1;
|
---|
308 | print CSVOUT $node_id . ',' . $csv_worker_id . ',' . $node_record->{'host'} . ',' . $node_record->{'start'} . ',' . $node_record->{'end'} . ',' . $node_record->{'cpu_time'} . ",0,1,NA,NA\n";
|
---|
309 | # List the child task records
|
---|
310 | foreach my $taskno (sort keys %{$task_records})
|
---|
311 | {
|
---|
312 | my $task_record = $task_records->{$taskno};
|
---|
313 | if ($task_record->{'host'} . '#' . $task_record->{'cpu'} eq $worker_id)
|
---|
314 | {
|
---|
315 | print CSVOUT $row_counter . ',T' . ($task_record->{'task'} + 0) . ',' . $task_record->{'host'} . ',' . $task_record->{'start'} . ',' . $task_record->{'end'} . ',' . $task_record->{'cpu_time'} . ',' . $task_record->{'data_locality'} . ',' . $node_id . ',' . $task_record->{'file'} . ',' . $task_record->{'percom'} . "\n";
|
---|
316 | $row_counter++;
|
---|
317 | }
|
---|
318 | }
|
---|
319 | }
|
---|
320 | close(CSVOUT);
|
---|
321 | }
|
---|
322 | else
|
---|
323 | {
|
---|
324 | die('Error! Failed to open file for writing: ' . $report_csv_path);
|
---|
325 | }
|
---|
326 |
|
---|
327 | print "Complete!\n\n";
|
---|
328 | exit;
|
---|
329 |
|
---|