1 | #!/bin/bash |
---|
2 | |
---|
3 | |
---|
4 | ## Messenger filenames |
---|
5 | FILE_DATE=mesg_01_date_$PATTERNAME.txt ; FILE_RSLT=mesg_02_result_$PATTERNAME.txt |
---|
6 | FILE_STAT=mesg_03_state_$PATTERNAME.txt ; FILE_NEMO=mesg_04_nemo_$PATTERNAME.txt |
---|
7 | FILE_XIOS=mesg_05_xios_$PATTERNAME.txt ; FILE_CMPF=mesg_06_compiler_$PATTERNAME.txt |
---|
8 | FILE_LMPI=mesg_07_mpi_$PATTERNAME.txt ; FILE_NCDF=mesg_08_netcdf_$PATTERNAME.txt |
---|
9 | FILE_INPT=mesg_09_inputs_$PATTERNAME.txt; FILE_TIME=mesg_10_time_$PATTERNAME.txt |
---|
10 | FILE_MEMY=mesg_11_memory_$PATTERNAME.txt; FILE_NOTE=mesg_12_comments_$PATTERNAME.txt |
---|
11 | |
---|
12 | ## Trusting timestamped logfile & archive |
---|
13 | TRUS_FILE=trusting_${DATE}_$PATTERNAME.txt; TRUS_ARCH=trusting_${DATE}_$PATTERNAME.tgz |
---|
14 | |
---|
15 | |
---|
16 | ## Functions in order of use |
---|
17 | print_step() { |
---|
18 | local char_nb=$( echo "$1" | wc -c ) |
---|
19 | local outline=$( printf "%${char_nb}s" ) |
---|
20 | |
---|
21 | printf "\nStep.....\n%s\n%s\n" "$1" ${outline// /-} |
---|
22 | } |
---|
23 | |
---|
24 | init_files() { |
---|
25 | echo 'Date' > ${FILE_DATE}; echo 'Result' > ${FILE_RSLT} |
---|
26 | echo 'State' > ${FILE_STAT}; echo 'NEMOGCM rev.' > ${FILE_NEMO} |
---|
27 | echo 'XIOS rev.' > ${FILE_XIOS}; echo 'Fortran compiler' > ${FILE_CMPF} |
---|
28 | echo 'MPI libs' > ${FILE_LMPI}; echo 'NetCDF libs' > ${FILE_NCDF} |
---|
29 | echo 'Input files' > ${FILE_INPT}; echo 'Elapsed time' > ${FILE_TIME} |
---|
30 | echo 'Memory usage (P/V)' > ${FILE_MEMY}; echo 'Comments' > ${FILE_NOTE} |
---|
31 | |
---|
32 | ## 'Failed' status with 'Unknown error' by default |
---|
33 | echo ${TRUS_RSLT} \ |
---|
34 | >> ${FILE_RSLT} |
---|
35 | echo 'Unknown error' \ |
---|
36 | >> ${FILE_STAT} |
---|
37 | } |
---|
38 | |
---|
39 | get_date() { |
---|
40 | ## UTC time zone for timestamping |
---|
41 | local dat=$( date -ud "${DATE}" +"%F %R %Z" ) |
---|
42 | |
---|
43 | echo $dat \ |
---|
44 | >> ${FILE_DATE} |
---|
45 | } |
---|
46 | |
---|
47 | get_nemo_rev() { |
---|
48 | local dir rev_loc |
---|
49 | local rev=0 |
---|
50 | |
---|
51 | ## Loop on essential NEMO directories |
---|
52 | for dir in ${TRUS_CKOT} ${TRUS_XIOS}; do |
---|
53 | |
---|
54 | ## For time being, just get revision from XIOS with no action on directory |
---|
55 | if [ $dir == ${TRUS_XIOS} ]; then |
---|
56 | rev_loc=$( svn info $dir | awk '/Last Changed Rev/ {print $NF}' ) |
---|
57 | echo 'XIOS '${rev_loc} \ |
---|
58 | >> model.log |
---|
59 | echo "<a href=\"https://forge.ipsl.jussieu.fr/ioserver/changeset/${rev_loc}\" target=\"_blank\">${rev_loc}</a>" \ |
---|
60 | >> ${FILE_XIOS} |
---|
61 | continue |
---|
62 | fi |
---|
63 | |
---|
64 | echo $dir && ${TRUS_SVNA} ${TRUS_NGCM}/$dir |
---|
65 | rev_loc=$( svn info ${TRUS_NGCM}/$dir | awk '/Last Changed Rev/ {print $NF}' ) |
---|
66 | |
---|
67 | ## Keep last rev. nb |
---|
68 | [ ${rev_loc} -gt $rev ] && rev=${rev_loc} |
---|
69 | done |
---|
70 | |
---|
71 | echo 'NEMOGCM '$rev \ |
---|
72 | >> model.log |
---|
73 | echo "<a href=\"https://forge.ipsl.jussieu.fr/nemo/changeset/$rev\" target=\"_blank\">$rev</a>" \ |
---|
74 | >> ${FILE_NEMO} |
---|
75 | } |
---|
76 | |
---|
77 | get_soft_rel() { |
---|
78 | local soft_rel str |
---|
79 | |
---|
80 | ## Sourcing environment |
---|
81 | if [ -n "${TRUS_ENVI}" ]; then |
---|
82 | if [[ -e ${TRUS_ENVI}.env && $( declare -F | grep ' module' ) ]]; then |
---|
83 | ## .env file if module function is available |
---|
84 | . ${TRUS_ENVI}.env |
---|
85 | else |
---|
86 | ## .path file if existing, if not the given file |
---|
87 | [ -e ${TRUS_ENVI}.path ] && . ${TRUS_ENVI}.path || . ${TRUS_ENVI} |
---|
88 | fi |
---|
89 | fi |
---|
90 | |
---|
91 | ## Problem with `prepend-path` of modulefile that use ':' instead of ' ' as delimiter |
---|
92 | [ $TRUS_HPCC == 'X64_ADA' ] && WRAPPER_LDFLAGS='-L/smplocal/pub/IdrMemMPI/1.4/lib -lidrmem '${WRAPPER_LDFLAGS} |
---|
93 | |
---|
94 | for str in ${TRUS_CMPV} ${TRUS_MPIR} ${TRUS_CDFR} ${TRUS_CDOR}; do |
---|
95 | [ -z "$str" ] && continue |
---|
96 | soft_rel='' |
---|
97 | |
---|
98 | ## Software release: next word after "$soft" in $PATH (case-insensitive) |
---|
99 | soft_rel=$( echo $PATH | sed "s#.*$str\([0-9.a-z_]*\).*#\1#i" ) |
---|
100 | |
---|
101 | ## option --version would work for main compilers (gfortran, intel, pgfortran, ...) |
---|
102 | [ $str == ${TRUS_CMPV} ] && soft_rel=$( $str --version | grep -m1 -oe '\<[0-9. ]*\>' ) |
---|
103 | |
---|
104 | ## Cleaning characters string to display proper soft name |
---|
105 | str=$( echo $str | sed 's#\\##g; s#[/-]$##' ) |
---|
106 | |
---|
107 | echo $str ${soft_rel} \ |
---|
108 | >> model.log |
---|
109 | done |
---|
110 | |
---|
111 | sed -n 3p model.log \ |
---|
112 | >> ${FILE_CMPF} |
---|
113 | sed -n 4p model.log \ |
---|
114 | >> ${FILE_LMPI} |
---|
115 | sed -n 5p model.log \ |
---|
116 | >> ${FILE_NCDF} |
---|
117 | } |
---|
118 | |
---|
119 | get_inputs() { |
---|
120 | ## Extract archive or copy files in case of personal inputs |
---|
121 | [ -z "${TRUS_TARF}" ] && get_io="cp ${TRUS_FORC}/* ." || get_io="tar -vxf ${TRUS_FORC}/${TRUS_TARF}" |
---|
122 | |
---|
123 | ${get_io} > /dev/null |
---|
124 | [ $? -ne 0 ] && get_out 3 || echo 'Success' |
---|
125 | [ $( find -name '*.gz' -print -quit ) ] && gunzip *.gz |
---|
126 | |
---|
127 | ls -lh > inputs_list.txt |
---|
128 | } |
---|
129 | |
---|
130 | diff_inputs() { |
---|
131 | local dif file |
---|
132 | local files_list='' mesg='Same' |
---|
133 | |
---|
134 | ## Simple diff |
---|
135 | for file in 'inputs_list.txt' *namelist_* *.xml cpp_*; do |
---|
136 | dif='' |
---|
137 | |
---|
138 | ## Continue even if input file is not in here (see after) |
---|
139 | if [ -e ${TRUS_STOR}/$file ]; then dif=$( diff -q $file ${TRUS_STOR}/$file ); else dif=0; fi |
---|
140 | |
---|
141 | ## Pass over useless file omission in benckmark directory |
---|
142 | [[ -n "$dif" && "$dif" != '0' ]] && { mesg='Different'; echo $dif; files_list+=$file' '; } |
---|
143 | done |
---|
144 | |
---|
145 | [ $mesg == 'Same' ] && echo $mesg |
---|
146 | echo $mesg \ |
---|
147 | >> ${FILE_INPT} |
---|
148 | |
---|
149 | ## List different files for web comment |
---|
150 | [ -n "${files_list}" ] && echo 'Inputs : '${files_list}'differ<br>' \ |
---|
151 | >> temp_${FILE_NOTE} |
---|
152 | } |
---|
153 | |
---|
154 | job_pending() { |
---|
155 | local outline=$( printf "%100s" ) time_elapsed=0 time_increment=30 |
---|
156 | |
---|
157 | sleep ${time_increment} |
---|
158 | |
---|
159 | ## Append a log file while pending |
---|
160 | while [[ $( eval ${TRUS_JSTA} ) && ${time_elapsed} -lt ${TRUS_TOUT} ]]; do |
---|
161 | printf "\n%s\n" ${outline// /#} \ |
---|
162 | >> computation.log |
---|
163 | [ -n "${TRUS_JINF}" ] && eval ${JOB_INFO} \ |
---|
164 | >> computation.log |
---|
165 | sleep ${time_increment} |
---|
166 | time_elapsed=$(( ${time_elapsed} + ${time_increment} )) |
---|
167 | done |
---|
168 | |
---|
169 | sleep ${time_increment} |
---|
170 | |
---|
171 | ## Kill remaining job & stop the test if it's too long |
---|
172 | [ ${time_elapsed} -eq ${TRUS_TOUT} ] && { eval ${JOB_DELE} &> /dev/null; get_out 6; } |
---|
173 | } |
---|
174 | |
---|
175 | diff_results() { |
---|
176 | local file |
---|
177 | local files_list='' mesg='Same' |
---|
178 | |
---|
179 | ## Simple diff |
---|
180 | for file in 'ocean.output' *.stat; do |
---|
181 | ## Stop if no benchmark files (ocean.output, eventual stat files) |
---|
182 | [ ! -e ${TRUS_STOR}/$file ] && { TRUS_RSLT='FAILED'; get_out 7; } |
---|
183 | |
---|
184 | diff -q $file ${TRUS_STOR}/$file |
---|
185 | |
---|
186 | ## Continue even if it differs |
---|
187 | [ $? -ne 0 ] && { TRUS_RSLT='FAILED'; mesg='Different'; files_list+=$file' '; } |
---|
188 | done |
---|
189 | |
---|
190 | [ $mesg == 'Same' ] && echo $mesg |
---|
191 | |
---|
192 | ## List different files for web comment |
---|
193 | [ -n "${files_list}" ] && echo 'Results : '${files_list}'differ<br>' \ |
---|
194 | >> temp_${FILE_NOTE} |
---|
195 | } |
---|
196 | |
---|
197 | diff_restart() { |
---|
198 | local base_name comp dif file list_comp list_tmsp nb_dom time_step tmsp |
---|
199 | local files_list='' dif_sum=0 |
---|
200 | |
---|
201 | ## Stop if no benchmark files (ie time.step) |
---|
202 | [ ! -e ${TRUS_STOR}/time.step ] && { TRUS_RSLT='FAILED'; get_out 7; } |
---|
203 | time_step=$( cat ${TRUS_STOR}/time.step | tr -d [:space:] ) |
---|
204 | |
---|
205 | ## Find all restart files to rebuild |
---|
206 | if [ $( find -regex ".*_restart.*[0-9]\.nc" -print -quit ) ]; then |
---|
207 | base_name=$( find -regex ".*_restart.*[0-9]\.nc" \ |
---|
208 | | sed "s#^\./\(.*\)_[0-9]*_restart.*#\1#" | sort -u ) |
---|
209 | list_comp=$( find -regex ".*_restart.*[0-9]\.nc" \ |
---|
210 | | sed "s#^.*\(restart[a-z_]*\)_[0-9].*\.nc#\1#" | sort -u ) |
---|
211 | list_tmsp=$( find -regex ".*_restart.*[0-9]\.nc" \ |
---|
212 | | sed "s#^.*\([0-9]\{8\}\)_restart.*#\1#" | sort -u ) |
---|
213 | |
---|
214 | ## Loop on each time step |
---|
215 | for tmsp in ${list_tmsp}; do |
---|
216 | |
---|
217 | for comp in ${list_comp}; do |
---|
218 | file=${base_name}_${tmsp}_${comp} |
---|
219 | nb_dom=$( find -name "${file}_[0-9]*.nc" | wc -l | awk '{ print $1 }' ) |
---|
220 | |
---|
221 | if [ ${nb_dom} -gt 1 ]; then |
---|
222 | ${TRUS_NGCM}/TOOLS/REBUILD_NEMO/rebuild_nemo -t ${TRUS_NPRO} $file ${nb_dom} \ |
---|
223 | > /dev/null |
---|
224 | |
---|
225 | ## Possibility of remaining decomposed restarts (even after rebuild) |
---|
226 | [ $? -eq 0 ] && rm -f ${file}_[0-9]*.nc \ |
---|
227 | > /dev/null |
---|
228 | |
---|
229 | elif [ ${nb_dom} -eq 0 ]; then |
---|
230 | TRUS_RSLT='FAILED' && get_out 8 |
---|
231 | fi |
---|
232 | |
---|
233 | ## Compare restart files at same time step |
---|
234 | if [ $tmsp -eq ${time_step} ]; then |
---|
235 | |
---|
236 | ## Stop if no benchmark files (restart file) |
---|
237 | if [ -e ${TRUS_STOR}/$file.nc ]; then |
---|
238 | |
---|
239 | ## UNIX `cmp` not suitable (timestamp in .nc file) |
---|
240 | dif=$( $TRUS_CDOD $file.nc ${TRUS_STOR}/$file.nc 2> /dev/null \ |
---|
241 | | awk '/records/ {print $0}' | sed '2 s/^/,/' | tr -d '\n' ) |
---|
242 | |
---|
243 | ## CDO can return void stdout with no difference |
---|
244 | if [[ -n "$dif" && $( echo $dif | awk '{print $1}' ) -ne 0 ]]; then |
---|
245 | TRUS_RSLT='FAILED' |
---|
246 | files_list+=$comp' ' && let dif_sum+=$( echo $dif | awk '{print $1}' ) |
---|
247 | echo $file.nc': '$dif |
---|
248 | fi |
---|
249 | |
---|
250 | else |
---|
251 | TRUS_RSLT='FAILED' && get_out 7 |
---|
252 | fi |
---|
253 | |
---|
254 | else |
---|
255 | continue |
---|
256 | fi |
---|
257 | |
---|
258 | done |
---|
259 | |
---|
260 | done |
---|
261 | |
---|
262 | ## List different files for web comment with sum of different records |
---|
263 | if [ ${dif_sum} -ne 0 ]; then |
---|
264 | echo 'Restarts: '${files_list}${dif_sum}' record(s) differ<br>' \ |
---|
265 | >> temp_${FILE_NOTE} |
---|
266 | else |
---|
267 | echo 'Same' |
---|
268 | fi |
---|
269 | |
---|
270 | else |
---|
271 | TRUS_RSLT='FAILED' |
---|
272 | fi |
---|
273 | |
---|
274 | } |
---|
275 | |
---|
276 | get_time() { |
---|
277 | [ -z "${TRUS_JTIM}" ] && return |
---|
278 | |
---|
279 | ## Interest for checking unusual time computation |
---|
280 | local time_cpu=$( eval ${TRUS_JTIM} ) |
---|
281 | |
---|
282 | printf "Elapsed time: " |
---|
283 | echo ${time_cpu} | tee -a ${FILE_TIME} |
---|
284 | } |
---|
285 | |
---|
286 | get_memy() { |
---|
287 | [[ -z "${TRUS_JPME}" && -z "${TRUS_JVME}" ]] && return |
---|
288 | |
---|
289 | ## Interest for checking unusual memory usage |
---|
290 | local memory_pmax=$( eval ${TRUS_JPME} ) memory_vmax=$( eval ${TRUS_JVME} ) |
---|
291 | |
---|
292 | printf "Memory max usage (physical/virtual): " |
---|
293 | echo ${memory_pmax}' / '${memory_vmax} | tee -a ${FILE_MEMY} |
---|
294 | } |
---|
295 | |
---|
296 | comments() { |
---|
297 | local opat |
---|
298 | local line='' state=$1 |
---|
299 | |
---|
300 | if [ -e ocean.output ]; then |
---|
301 | ## 'W A R N I N G' pattern by default |
---|
302 | opat="-A2 \"^ $state\"" |
---|
303 | [ "$state" == 'E R R O R' ] && opat="-A4 \"$state\"" |
---|
304 | |
---|
305 | ## Select first occurence for web comment |
---|
306 | line=$( eval grep -m1 $opat ocean.output | tr -d '\n' ) |
---|
307 | fi |
---|
308 | |
---|
309 | [ -n "$line" ] && ( echo $line; printf "$line<br>" \ |
---|
310 | >> temp_${FILE_NOTE} ) |
---|
311 | } |
---|
312 | |
---|
313 | log_make() { |
---|
314 | ## Format comments for web |
---|
315 | [ -e temp_${FILE_NOTE} ] && cat temp_${FILE_NOTE} | tr -d '\n' | sed 's/<br>$//' \ |
---|
316 | >> ${FILE_NOTE} |
---|
317 | |
---|
318 | ## Construct txt file with all messenger files |
---|
319 | paste -d ';' mesg_*.txt | tee ${TRUS_FILE} |
---|
320 | } |
---|
321 | |
---|
322 | prod_publish() { |
---|
323 | local cmd |
---|
324 | local rev=$( awk '/NEMOGCM/ {print $NF}' model.log ) |
---|
325 | |
---|
326 | ## Production mode (-p|--prod) |
---|
327 | if [ ${TRUS_PROD} -eq 1 ]; then |
---|
328 | |
---|
329 | ## Create or append trusting logfile |
---|
330 | if [ -f ${TRUS_STOR}/trusting_$PATTERNAME.txt ]; then cmd='tail -1'; else cmd='cat'; fi |
---|
331 | |
---|
332 | $cmd ${TRUS_FILE} \ |
---|
333 | >> ${TRUS_STOR}/trusting_$PATTERNAME.txt |
---|
334 | |
---|
335 | ## Send mail only when FAILED |
---|
336 | if [[ ! -z "${TRUS_MAIL}" && ${TRUS_RSLT} == 'FAILED' ]]; then |
---|
337 | |
---|
338 | ## Content |
---|
339 | cat <<END_MAIL \ |
---|
340 | > trusting.mail |
---|
341 | Dear all, |
---|
342 | |
---|
343 | |
---|
344 | The trusting sequence has not completed successfully on new configuration ${TRUS_CONF} based on ${TRUS_REFE}. |
---|
345 | |
---|
346 | Here is the model summary: |
---|
347 | `cat model.log` |
---|
348 | |
---|
349 | First checking would be on the trusting environment files: |
---|
350 | ${TRUS_USER}.cfg & ${TRUS_HPCC}.cfg |
---|
351 | |
---|
352 | For more details, look into the testing folder at: |
---|
353 | ${TRUS_SCRA} |
---|
354 | |
---|
355 | An archive has been created to share the questionable configuration for further studies: |
---|
356 | ${TRUS_STOR}/${TRUS_ARCH} |
---|
357 | |
---|
358 | END_MAIL |
---|
359 | |
---|
360 | ## Send with detailed subject |
---|
361 | mail -s "[NEMO Trusting][$rev][${TRUS_BRAN}][${TRUS_REFE}] ${TRUS_RSLT} ${TRUS_RORR}" ${TRUS_MAIL} \ |
---|
362 | < trusting.mail |
---|
363 | fi |
---|
364 | |
---|
365 | fi |
---|
366 | } |
---|
367 | |
---|
368 | get_out() { |
---|
369 | local time_step=0 |
---|
370 | |
---|
371 | TRUS_RORR=$1 |
---|
372 | |
---|
373 | printf "\n\nEnd of test\n" |
---|
374 | |
---|
375 | ## In case of compilation error |
---|
376 | cd ${TRUS_SCRA} |
---|
377 | |
---|
378 | if [ ${TRUS_RSLT} == 'FAILED' ]; then |
---|
379 | echo 'Failure' |
---|
380 | |
---|
381 | ## Error identification |
---|
382 | case ${TRUS_RORR} in |
---|
383 | ## Compilation |
---|
384 | '1') TRUS_RORR='XIOS compilation failed' ;; '2') TRUS_RORR='NEMO compilation failed';; |
---|
385 | ## Submission |
---|
386 | '3') TRUS_RORR='Missing input files' ;; '4') TRUS_RORR='Job submission error' ;; |
---|
387 | ## Computation |
---|
388 | '5') TRUS_RORR='Crashed at time step' ;; '6') TRUS_RORR='Exceeded time limit' ;; |
---|
389 | ## Results |
---|
390 | '7') TRUS_RORR='Missing previous outputs';; '8') TRUS_RORR='New outputs differ' ;; |
---|
391 | ## Other |
---|
392 | '*') TRUS_RORR='Unknown error' ;; |
---|
393 | esac |
---|
394 | |
---|
395 | else |
---|
396 | echo 'Success' && TRUS_RORR='Code is reliable' |
---|
397 | fi |
---|
398 | |
---|
399 | ## Eventual comments from ocean.output |
---|
400 | if [ "${TRUS_RORR}" == 'Crashed at time step' ]; then |
---|
401 | comments 'E R R O R' |
---|
402 | [ -e time.step ] && time_step=$( grep -o [0-9]* time.step ) |
---|
403 | TRUS_RORR+=' '$time_step |
---|
404 | else |
---|
405 | comments 'W A R N I N G' |
---|
406 | [ "${TRUS_RORR}" == 'Exceeded time limit' ] && TRUS_RORR+=' '$(( ${TRUS_TOUT}/3600 ))'h' |
---|
407 | fi |
---|
408 | |
---|
409 | ## Last messenger files |
---|
410 | #export TRUS_RORR |
---|
411 | sed -i "2 s/.*/$TRUS_RSLT/" ${FILE_RSLT}; sed -i "2 s/.*/$TRUS_RORR/" ${FILE_STAT} |
---|
412 | |
---|
413 | ## Save tested configuration if trusting failed in production mode (-p|--prod) |
---|
414 | if [[ ${TRUS_RSLT} == 'FAILED' && ${TRUS_PROD} -eq 1 ]]; then |
---|
415 | echo 'Creating archive '${TRUS_ARCH}' under '${TRUS_STOR} |
---|
416 | tar -czf ${TRUS_STOR}/${TRUS_ARCH} * \ |
---|
417 | -C ${TRUS_NGCM}/CONFIG/${TRUS_CONF}/MY_SRC . \ |
---|
418 | -C ${TRUS_NGCM}/CONFIG/${TRUS_CONF} cpp_${TRUS_CONF}.fcm |
---|
419 | fi |
---|
420 | |
---|
421 | ## Logfile construct & eventual sending of notification email |
---|
422 | printf "\nTrusting digest:\n----------------\n" |
---|
423 | log_make |
---|
424 | prod_publish |
---|
425 | |
---|
426 | exit 0 |
---|
427 | } |
---|