#!/bin/bash # palmrun - script for running PALM jobs #--------------------------------------------------------------------------------# # This file is part of the PALM model system. # # PALM is free software: you can redistribute it and/or modify it under the terms # of the GNU General Public License as published by the Free Software Foundation, # either version 3 of the License, or (at your option) any later version. # # PALM is distributed in the hope that it will be useful, but WITHOUT ANY # WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR # A PARTICULAR PURPOSE. See the GNU General Public License for more details. # # You should have received a copy of the GNU General Public License along with # PALM. If not, see . # # Copyright 2017-2018 Leibniz Universitaet Hannover #--------------------------------------------------------------------------------# # # Current revisions: # ------------------ # # # Former revisions: # ----------------- # $Id: palmrun 4303 2019-11-25 08:55:18Z motisi $ # Abort if temporary working catalog cannot be created # # 3982 2019-05-16 11:07:31Z raasch # bugfix: -b is not added in automatically generated batch jobs for remote hosts # # 3917 2019-04-18 11:46:37Z raasch # in case of batch jobs on remote machines, copying files from SOURCES_FOR_RUN folder has # been rearranged to avoid that input files with original names (e.g. ..._p3d) are copied # into the temporary working catalog # # 3860 2019-04-04 06:23:27Z raasch # check added, that it is not allowed to set option -b if the configuration file has been setup # for running batch jobs on remote hosts, # local time added to terminal output before and after saving output files # # 3725 2019-02-07 10:11:02Z raasch # error messages for failed restarts extended # # 3682 2019-01-18 17:01:54Z knoop # ssh-call for submitting batch jobs on remote systems modified again to avoid output # of login messages on specific systems # # 3574 2018-11-28 13:14:39Z raasch # syntax-error in last commit fixed # # 3573 2018-11-28 13:03:44Z raasch # bugfixes: don't check the .palm.iofiles in restart runs + wrong cd-command # removed, # r3570 reverted # # 3570 2018-11-27 17:44:21Z kanani # Rename job-id by run_id # # 3555 2018-11-22 13:50:37Z raasch # sed-command to remove repeated space-characters replaced by tr-command, to # guarantee mac compatible # # 3549 2018-11-21 15:44:44Z raasch # ssh-call for submitting batch jobs on remote systems modified to avoid output # of login messages on specific systems # # 3534 2018-11-19 15:35:16Z raasch # batch_scp for sending back the job protocol file is called via login-node if # a login-node has been set in the config-file # ssh-calls rearranged to avoid output of system/user-profile messages # # 3505 2018-11-08 09:14:30Z raasch # job folders are created in case that they do not exist # # 3455 2018-10-30 14:12:31Z raasch # options -h and -d renamed -c and -r, respectively # # 3402 2018-10-23 11:49:24Z knoop # job_id is set for compatibility reasons with older versions of config files # # 3313 2018-10-06 15:22:48Z knoop # more detailed error message for outdated fileconnection file # # 3236 2018-09-10 09:07:06Z raasch # warning issued in case of file connection file version mismatch, # host_configuration renamed configuration_identifier, # "host identifier" in header output renamed "config. identifier", # jobname renamed run_identifier, # run_id renamed run_id_number # job_id renamed run_id # option -r renamed -i # # 3151 2018-07-19 08:45:38Z raasch # removed additional listing of files to be compiled, print PALM code revision # in header # # 3132 2018-07-16 11:44:58Z raasch # create consistent error messages in case of failed restarts, # check existence of remote job catalog # # 3043 2018-05-25 10:47:54Z raasch # check for setting of fast_io_catalog # # 3033 2018-05-23 15:26:19Z raasch # -T option not required for interactive runs (tasks per node = # total number of cores is assumed as default, if option is not given), # reformatting of a few messages, # bugfix: destination output directory is created if attribute "pe" is set in # .palm.iofiles # # 2718 2018-01-02 08:49:38Z maronga # bugfix of r2990 # # 2718 2018-01-02 08:49:38Z maronga # bugfix: temporary SOURCES_FOR_RUN_CATALOG now created in base_directory and # not in the working directory from where palmrun is called # # 2718 2018-01-02 08:49:38Z maronga # svfout in the activation string will be replaced now by svfin in case of # automatic restarts # # 2718 2018-01-02 08:49:38Z maronga # Introduction of new ENVIRONMENT variables read_svf and write_svf to consider # the possible output or input of sky view factors # # 2718 2018-01-02 08:49:38Z maronga # "fname" renamed to "jobname" # # 2718 2018-01-02 08:49:38Z maronga # parameters are allowed in output commands (OC:) # # 2718 2018-01-02 08:49:38Z maronga # bugfix: exit if no queue has been given neither with option -q nor in the # configuration file # # 2718 2018-01-02 08:49:38Z maronga # file connection file from trunk/SCRIPTS is taken by default # # 2718 2018-01-02 08:49:38Z maronga # Corrected "Former revisions" section # # 2696 2017-12-14 17:12:51Z kanani # Change in file header (GPL part) # # 2693 2017-12-13 12:33:49Z raasch # replacement of variables given by {{}} in configuration file is now # done in a more general way, variable names are written in small letters # # 2670 2017-12-06 16:32:19Z raasch # test printouts removed # # 2669 2017-12-06 16:03:27Z raasch # file attributes in .palm.iofiles restructured, "loc" attribute completely # removed, # wildcard (*) allowed in .palm.iofiles as file activation string for output # files, # informative messages in case of missing optional input files shortened # bugfix: variable cycle explicitly interpreted with 10 as the number base # # 2638 2017-11-23 12:44:23Z raasch # use of wildcards in file connection statements enabled # # 2605 2017-11-09 15:31:46Z raasch # in case of remote jobs, input files with "job" or "jobopt" (new) attribute # will not be sent with the job file any more, but copied into the # SOURCES_FOR_RUN... folder on the remote host, before the job is submitted # # 2600 2017-11-01 14:11:20Z raasch # cycle numbers are made three digits wide # # 2566 2017-10-20 08:50:47Z raasch # execute command for combine_plot_fields added # "TEMPDIR" renamed "tempdir" # temporary working directory for local batch jobs is created immediately within # the user's palmrun call, due to a requirement of the "grid engine" batch # system, where the working directory is given with batch directive -wd and must # already exist when the job is submitted, # informative messages in non-trace mode reduced and partly reformatted # # 2551 2017-10-18 07:25:11Z raasch # TEMPDIR added as replacement string to be used in batch directives # # 2512 2017-10-04 08:26:59Z raasch # bugfix for determining cycle numbers of NetCDF input files # # 2506 2017-09-29 08:30:37Z raasch # option -V added to check for an existing SOURCES_FOR_RUN_... folder # host configuration added to SOURCES_FOR_RUN_... folder name # host_identifier renamed host_configuration # option -W added to allow for job dependencies # # 2501 2017-09-26 11:41:55Z raasch # default value for number of cores (option -X) set to 1 # bugfix for mechanism which overwrites configuration file settings with values # provided by palmrun options # # 2499 2017-09-22 16:47:58Z kanani # option -h named configuration identifier # # 2480 2017-09-19 06:24:14Z maronga # bugfix for last revision # # 2479 2017-09-19 06:12:16Z raasch # option -A (project account number) added # # 2422 2017-09-08 08:25:41Z raasch # initial revision # #--------------------------------------------------------------------------------# # palmrun - script for running PALM jobs on local and remote hosts #--------------------------------------------------------------------------------# # DECLARATION OF VARIABLES AND THEIR DEFAULT VALUES set +o allexport # SUPPRESS EXPORT OF ALL VARIABLES, SINCE IN THE PAST THIS # LEAD TO PROBLEMS IN ROUTINES CALLED BY PALMRUN # (TOO MANY ARGUMENTS - PROBLEM) set +o noclobber # EXISTING FILES ARE ALLOWED TO BE OVERWRITTEN typeset -i ibd=0 ibdt=0 iec=0 iic=0 iin=0 ioc=0 iout=0 nr_of_input_files=0 typeset -i nr_of_output_files=0 typeset -x -i memory=0 # HAS TO BE EXPORTED HERE, OTHERWISE AN UNKNOWN SIDE # SIDE EFFECT MAY CAUSE DATA LOSS WHEN GETOPTS IS READING THE # SCRIPT-OPTION ARGUMENTS typeset -i cores cputime cpu_hours cpu_minutes cpu_seconds i ii iia iii iio typeset -i icycle inode ival maxcycle mpi_tasks typeset -i nodes remaining_cores run_number tasks_per_node threads_per_task activation_string_list="" AddFilenames="" combine_plot_fields=true compiler_name="" compiler_name_ser="" compiler_options="" configuration_identifier="default" cores=1 cores_atmos=0 cores_ocean=0 coupled_dist="" cpp_options="" cpumax=0 create_batch_job=false create_jobfile_only=false create_remote_batch_job=false dashes=" ----------------------------------------------------------------------------" defaultqueue="" delete_temporary_catalog=true do_compile=true do_trace=false executable="" execute_command="" execution_error=false fileconnection_file=trunk/SCRIPTS/.palm.iofiles global_revision="" hostfile="" hp="" keep_data_from_previous_run=false link_local_input=false link_local_output=false linker_options="" local_jobcatalog="" locat=normal makefile="" max_par_io_str="" prc=$0 while [[ $(echo $prc | grep -c "/") != 0 ]] do prc=`echo $prc | cut -f2- -d"/"` done module_calls="" palmrun_memory="" palmrun_script_name=$prc openmp=false previous_job="" project_account="" queue=none read_svf="" restart_run=false return_address="" remote_jobcatalog="" remote_username="" running_in_batch_mode=false running_in_test_mode=false run_coupled_model=false run_id="" run_id_number="" run_identifier=test silent=false source_list="" source_path="" tasks_per_node=0 threads_per_task=1 transfer_problems=false user_source_path="" use_existing_sources_folder="" use_openmp=false version="palmrun 1.0 Rev$Rev: 4303 $" working_directory=`pwd` write_binary="" write_svf="" # ERROR HANDLING IN CASE OF EXIT trap 'if [[ $locat != normal && $locat != control_c ]] then # CARRY OUT ERROR-COMMANDS GIVEN IN THE CONFIGURATION FILE (EC:) (( i = 0 )) while (( i < iec )) do (( i = i + 1 )) printf "\n *** Execution of ERROR-command:\n" printf " >>> ${err_command[$i]}\n" eval ${err_command[$i]} done [[ $delete_temporary_catalog = true ]] && (cd; rm -rf $tempdir) printf "\n\n+++ palmrun crashed \n\n" exit 1 elif [[ $locat != normal ]] then [[ $delete_temporary_catalog = true ]] && (cd; rm -rf $tempdir) printf "\n+++ palmrun killed by \"^C\" \n\n" exit 2 else printf "\n --> palmrun finished\n\n" exit 0 fi' exit # ACTIONS IN CASE OF TERMINAL-BREAK (CONTROL-C): trap 'locat=control_c exit 1 ' 2 # READ SHELLSCRIPT-OPTIONS AND REBUILD THE PALMRUN-COMMAND STRING (prc), # WHICH WILL BE USED TO START RESTART-JOBS while getopts :a:A:bBc:CFG:i:jkm:M:O:q:r:R:s:t:T:u:U:vVw:W:xX:yY:zZ option do case $option in (a) activation_string_list=$OPTARG; prc="$prc -a'$OPTARG'";; (A) project_account=$OPTARG; prc="$prc -A'$OPTARG'";; (b) create_batch_job=true; prc="$prc -b";; (B) delete_temporary_catalog=false; prc="$prc -B";; (c) configuration_identifier=$OPTARG; prc="$prc -c$OPTARG";; (C) restart_run=true; prc="$prc -C";; (F) create_jobfile_only=true;; (G) global_revision=$OPTARG; prc="$prc -G'$OPTARG'";; (i) run_id_number=$OPTARG;; (j) running_in_batch_mode=true;; (k) keep_data_from_previous_run=true; prc="$prc -k";; (m) palmrun_memory=$OPTARG; prc="$prc -m$OPTARG";; (M) makefile=$OPTARG; prc="$prc -M$OPTARG";; (O) use_openmp=true; threads_per_task=$OPTARG; prc="$prc -O$OPTARG";; (q) queue=$OPTARG; prc="$prc -q$OPTARG";; (r) run_identifier=$OPTARG; prc="$prc -r$OPTARG";; (R) return_address=$OPTARG;; (s) source_list=$OPTARG;; (t) palmrun_cpumax=$OPTARG; prc="$prc -t$OPTARG";; (T) palmrun_tasks_per_node=$OPTARG; prc="$prc -T$OPTARG";; (u) remote_username=$OPTARG; prc="$prc -u$OPTARG";; (U) return_username=$OPTARG; prc="$prc -U$OPTARG";; (v) silent=true; prc="$prc -v";; (V) use_existing_sources_folder="-V";; (w) max_par_io_str=$OPTARG; prc="$prc -w$OPTARG";; (W) previous_job=$OPTARG;; (x) do_trace=true;set -x; prc="$prc -x";; (X) palmrun_cores=$OPTARG; prc="$prc -X$OPTARG";; (y) ocean_file_appendix=true; prc="$prc -y";; (Y) run_coupled_model=true; coupled_dist=$OPTARG; prc="$prc -Y'$OPTARG'";; (z) running_in_test_mode=true;; (Z) combine_plot_fields=false; prc="$prc -Z";; (\?) printf "\n +++ unknown option $OPTARG \n" printf "\n --> type \"$0 ?\" for available options \n" locat=parameter;exit;; esac done # SKIP GIVEN OPTIONS TO READ POSITIONAL PARAMETER, IF GIVEN # CURRENTLY ONLY PARAMETER ? (TO OUTPUT A SHORT COMMAND INFO) IS ALLOWED (( to_shift = $OPTIND - 1 )) shift $to_shift # PRINT SHORT DESCRIPTION OF PALMRUN OPTIONS if [[ "$1" = "?" ]] then (printf "\n *** Description of available palmrun options:\n" printf "\n Option Description Default-Value" printf "\n -a activation string list \"\" " printf "\n -A project account number ---" printf "\n -b batch-job on local machine ---" printf "\n -B do not delete temporary directory at end ---" printf "\n -c configuration identifier \"default\" " printf "\n -F create batch job file only ---" printf "\n -k keep data from previous run" printf "\n -m memory demand in MB (batch-jobs) 0 MB" printf "\n -M Makefile name Makefile" printf "\n -O threads per openMP task ---" printf "\n -q queue \"$queue\" " printf "\n -r run identifier test" printf "\n -s filenames of routines to be compiled \"\" " printf "\n must end with .f, .f90, .F, or .c !" printf "\n use \"..\" for more than one file and wildcards" printf "\n -s LM compiles all locally modified files" printf "\n -t allowed cpu-time in seconds (batch) 0" printf "\n -T tasks per node ---" printf "\n -u username on remote machine \"\" " printf "\n -v no prompt for confirmation ---" printf "\n -V check if SOURCES_FOR_RUN_... exists ---" printf "\n -w maximum parallel io streams as given by -X" printf "\n -W name of job to wait for ---" printf "\n -x tracing of palmrun for debug purposes ---" printf "\n -X # of processors (on parallel machines) 1" printf "\n -y add appendix \"_O\" to all local output" printf "\n files (ocean precursor runs followed by" printf "\n coupled atmosphere-ocean runs) ---" printf "\n -Y run coupled model, \"#1 #2\" with" printf "\n #1 atmosphere and #2 ocean processors \"#/2 #/2\" depending on -X" printf "\n -Z skip combine_plot_fields at the end of " printf "\n the simulation ---" printf "\n " printf "\n Possible values of positional parameter :" printf "\n \"?\" - this outline \n\n") | more exit elif [[ "$1" != "" ]] then printf "\n +++ positional parameter $1 unknown \n" locat=parameter; exit fi # SHORT STARTING MESSAGE printf "\n*** $version " printf "\n will be executed. Please wait ..." # BUILD THE CONFIGURATION-FILE NAME AND THE SOURCES_FOR_RUN-FOLDER NAME config_file=.palm.config.$configuration_identifier sources_for_run_catalog=SOURCES_FOR_RUN_${configuration_identifier}_$run_identifier # CHECK, IF CONFIGURATION FILE EXISTS if [[ ! -f $config_file ]] then printf "\n\n +++ configuration file: " printf "\n $config_file" printf "\n does not exist" locat=connect; exit fi # CHECK, IF USER PROVIDES OWN FILE CONNECTION FILE if [[ -f .palm.iofiles ]] then # CHECK VERSION MISMATCH # RESTART RUNS DON'T REQUIRE A CHECK, BECAUSE THEY USE A COPY OF THAT # FILE WHICH HAS ALREADY BEEN CHECKED WITHIN THE INITAL PALMRUN CALL if [[ $running_in_batch_mode != true && $restart_run != true ]] then if [[ $(head -n1 $fileconnection_file) != $(head -n1 .palm.iofiles) ]] then printf "\n\n +++ WARNING: A file connection file has been found in your" printf "\n working directory, but its revision does not match" printf "\n the revision of the default (trunk) version." printf "\n You may need to update your connection file" printf "\n \"${working_directory}/.palm.iofiles\" !" fi fi fileconnection_file=.palm.iofiles fi # CHECK, IF FILE CONNECTION FILE EXISTS if [[ ! -f $fileconnection_file ]] then printf "\n\n +++ file connection file: " printf "\n $fileconnection_file" printf "\n does not exist" locat=connect; exit fi # CHECK, IF THE ACTIVATION_STRING_LIST HAS BEEN GIVEN if [[ "$activation_string_list" = "" ]] then printf "\n\n +++ no activation string list given: " printf "\n please set palmrun option \"-a\" " locat=palmrun_option; exit fi # SET VARIABLE TO ACTIVATE PALM BINARY OUTPUT FOR RESTARTS if [[ $(echo $activation_string_list | grep -c "restart") != 0 ]] then write_binary=true else write_binary=false fi # SET VARIABLE TO ACTIVATE WRITING OF SVF DATA if [[ $(echo $activation_string_list | grep -c "svfout") != 0 ]] then write_svf=true else write_svf=false fi # SET VARIABLE TO ACTIVATE READING OF SVF DATA if [[ $(echo $activation_string_list | grep -c "svfin") != 0 ]] then read_svf=true else read_svf=false fi # READ AND EVALUATE THE CONFIGURATION-FILE [[ $silent = false ]] && printf "\n\n Reading the configuration file... " # READ VARIABLE SETTINGS FROM CONFIG FILE LINE BY LINE while read line do # FIRST REPLACE ENVIRONMENT-VARIABLES BY THEIR RESPECTIVE VALUES eval line=\"$line\" # INTERPRET THE LINE if [[ "$(echo $line)" = "" ]] then # EMPTY LINE, NO ACTION continue elif [[ "$(echo $line | cut -c1)" = "#" ]] then # LINE IS A COMMENT LINE continue elif [[ "$(echo $line | cut -c1)" = "%" ]] then # LINE DEFINES AN ENVIRONMENT-VARIABLE var=`echo $line | cut -d" " -s -f1 | cut -c2-` value=`echo $line | cut -d" " -s -f2-` # VALUE FROM THE CONFIGURATION-FILE IS ASSIGNED TO THE # ENVIRONMENT-VARIABLE, BUT ONLY IF NO VALUE HAS BEEN ALREADY # ASSIGNED WITHIN THIS SCRIPT (E.G. BY SCRIPT-OPTIONS). # NON-ASSIGNED VARIABLES HAVE VALUE "" OR 0 (IN CASE OF INTEGER). # HENCE THE GENERAL RULE IS: SCRIPT-OPTION OVERWRITES THE # CONFIGURATION-FILE. if [[ "$(eval echo \$$var)" = "" || "$(eval echo \$$var)" = "0" ]] then eval export $var="\$value" # TERMINAL OUTPUT OF ENVIRONMENT-VARIABLES, IF TRACEBACK IS SWITCHED on if [[ $do_trace = true ]] then printf "\n*** ENVIRONMENT-VARIABLE $var = $value" fi fi elif [[ "$(echo $line | cut -c1-3)" = "BD:" ]] then # LINE DEFINES BATCH-DIRECTIVE (( ibd = ibd + 1 )) line=$(echo $line | cut -c4-) batch_directive[$ibd]="$line" elif [[ "$(echo $line | cut -c1-4)" = "BDT:" ]] then # LINE DEFINES BATCH-DIRECTIVE FOR SENDING BACK THE JOBFILE FROM A # REMOTE TO A LOCAL HOST (( ibdt = ibdt + 1 )) line=$(echo $line | cut -c5-) batch_directive_transfer[$ibdt]="$line" elif [[ "$(echo $line | cut -c1-3)" = "EC:" ]] then # LINE DEFINES ERROR-COMMAND (( iec = iec + 1 )) line=$(echo $line | cut -c4-) err_command[$iec]="$line" elif [[ "$(echo $line | cut -c1-3)" = "IC:" ]] then # LINE DEFINES INPUT-COMMAND (( iic = iic + 1 )) line=$(echo $line | cut -c4-) in_command[$iic]="$line" elif [[ "$(echo $line | cut -c1-3)" = "OC:" ]] then # LINE DEFINES OUTPUT-COMMAND (( ioc = ioc + 1 )) line=$(echo $line | cut -c4-) out_command[$ioc]="$line" else # SKIP ALL OTHER LINES continue fi done < $config_file # CHECK SETTING OF REQUIRED PARAMETERS if [[ "$fast_io_catalog" = "" ]] then printf "\n\n +++ no fast_io_catalog found in $config_file" printf "\n Please add line \"fast_io_catalog ...\" to that file." locat=config_file; exit fi if [[ "$compiler_name" = "" ]] then printf "\n\n +++ no compiler name found in $config_file" printf "\n Please add line \"compiler_name ...\" to that file." locat=config_file; exit fi if [[ "$compiler_name_ser" = "" ]] then printf "\n\n +++ no compiler name for non-paralle compilation found in $config_file" printf "\n Please add line \"compiler_name_ser ...\" to that file." locat=config_file; exit fi if [[ "$compiler_options" = "" ]] then printf "\n\n +++ no compiler options found in $config_file" printf "\n Please add line \"compiler_options ...\" to that file." locat=config_file; exit fi if [[ "$execute_command" = "" ]] then printf "\n\n +++ no execute command found in $config_file" printf "\n Please add line \"execute_command ...\" to that file." locat=config_file; exit fi if [[ "$hostfile" != "" ]] then if [[ $hostfile != auto && ! -f $hostfile ]] then printf "\n\n +++ no hostfile \"$hostfile\" found" printf "\n Please check line \"hostfile ...\" in $config_file" locat=config_file; exit fi fi # DETERMINE THE CALL STATUS if [[ "$return_address" != "" ]] then # I AM RUNNING ON A REMOTE HOST, WHICH ALSO MEANS THAT I AM RUNNING IN # BATCH MODE AND ... running_on_remote=true else # I HAVE BEEN CALLED INTERACTIVELY ON THIS HOST if [[ "$remote_ip" != "" ]] then # I HAVE TO CREATE A BATCH JOB TO RUN PALM ON THE REMOTE HOST create_remote_batch_job=true # CHECK THAT THE REMOTE JOBCATALOG HAS BEEN SET if [[ "$remote_jobcatalog" = "" ]] then printf "\n\n +++ no remote jobcatalog found in $config_file" printf "\n Please add line \"remote_jobcatalog ...\" to that file." locat=config_file; exit fi # CHECK THAT OPTION -b (SUBMITTING BATCH JOB ON LOCAL HOST) HASN'T BEEN SET if [[ $create_batch_job = true ]] then printf "\n\n +++ option -b must not be set because configuration file has been setup" printf "\n for running jobs on a remote host" locat=config_file; exit fi fi running_on_remote=false fi # READ AND EVALUATE THE I/O-FILE LIST [[ $silent = false ]] && printf "\n Reading the I/O files... " # READ THE FILE CONNECTION FILE LINE BY LINE while read line do # REPLACE REPEATING SPACES BETWEEN THE COLUMNS BY A SINGLE SPACE # HERE, TR IS USED INSTEAD OF SED TO GUARANTEE MAC COMPATIBILITY line=`echo "$line" | tr -s " "` # INTERPRET THE LINE if [[ "$(echo $line)" = "" ]] then # EMPTY LINE, NO ACTION continue elif [[ "$(echo $line | cut -c1)" = "#" ]] then # LINE IS A COMMENT LINE true else # LINE DEFINES FILE CONNECTION. READ THE FILE ATTRIBUTES. # s2a: in/out - field # s2b: action - field (optional) s1=`echo "$line" | cut -d" " -f1` s2=`echo "$line" | cut -d" " -s -f2` if [[ $(echo $s2 | grep -c ":") = 0 ]] then s2a=$s2 s2b="" else s2a=`echo $s2 | cut -d":" -f1` s2b=`echo $s2 | cut -d":" -f2` fi s3=`echo "$line" | cut -d" " -f3 | sed 's/*/wildcard /g'` s4=`echo "$line" | cut -d" " -s -f4` eval s4=\"$s4\" # REPLACE ENVIRONMENT-VARIABLES IN PATH BY THEIR RESPECTIVE VALUES s5=`echo "$line" | cut -d" " -s -f5` s6=`echo "$line" | cut -d" " -s -f6` # STORE FILE CONNECTION, IF ACTIVATED BY ACTIVATION-STRING FROM # INPUT- OR OUTPUT-LIST. # VARIABLE S3 MAY CONTAIN A LIST OF ACTIVATION STRINGS (FIELD-SEPERATOR ":"). # IF EXECUTION IS SCHEDULED FOR A REMOTE-MACHINE AND THE FILE IS ONLY # LOCALLY REQUIRED ON THAT MACHINE (I.E. s2b != tr), THE FILE CONNECTION # IS NOT CHECKED AND STORED. IFSALT="$IFS"; IFS="$IFS:" # ADD ":" AS FIELD SEPARATOR if [[ ( "$s2a" = in || "$s2a" = inopt ) && ! ( $create_remote_batch_job = true && "$s2b" != tr ) ]] then found=false for actual in $activation_string_list do for formal in $s3 do [[ $actual = $formal || "$formal" = "-" ]] && found=true done done if [[ $found = true ]] then (( iin = iin + 1 )) localin_pre[$iin]=$s1; actionin_pre[$iin]=$s2b; pathin_pre[$iin]=$s4; endin_pre[$iin]=$s5; extin_pre[$iin]=$s6 if [[ "$s2a" = inopt ]] then optin_pre[$iin]=yes else optin_pre[$iin]=no fi # FILES WITH JOB-ATTRIBUTE ARE STORED IN THE SOURCES_FOR_RUN # FOLDER IF THE JOB IS RUNNING ON A REMOTE HOST if [[ $running_on_remote = true && "$s2b" = tr ]] then pathin_pre[$iin]=${fast_io_catalog}/${sources_for_run_catalog} fi # CHECK FOR MULTIPLE FILES, SET A RESPECTIVE FLAG AND REMOVE # THE WILDCARD FROM THE ENDING if [[ "${s5: -1}" = "*" ]] then if [[ "$s2b" = "di" ]] then printf "\n +++ wildcards (*) not allowed with \"di\" file attribute." printf "\n see file \"$fileconnection_file\", line" printf "\n$line" locat=iofiles_file; exit fi multin[$iin]=true string=${endin_pre[$iin]} endin_pre[$iin]="${string%?}" else multin[$iin]=false fi fi elif [[ "$s2a" = out && ! ( $create_remote_batch_job = true ) ]] then found=false for actual in $activation_string_list do for formal in $s3 do [[ $actual = $formal || $formal = wildcard ]] && found=true done done if [[ $found = true ]] then (( iout = iout + 1 )) localout_pre[$iout]=$s1; actionout_pre[$iout]=$s2b pathout_pre[$iout]=$s4; endout_pre[$iout]=$s5; extout_pre[$iout]=$s6 # CHECK IF WILDCARD IS USED AS ACTIVATION STRING # IN SUCH CASES, NO WARNING WILL LATER BE OUTPUT IF LOCAL FILES DO NOT EXIST if [[ $formal = wildcard ]] then warnout_pre[$iout]=false else warnout_pre[$iout]=true fi # CHECK FOR MULTIPLE FILES, SET A RESPECTIVE FLAG AND REMOVE # THE WILDCARD FROM THE LOCAL FILENAME if [[ "${s1: -1}" = "*" ]] then if [[ "$s2b" = "di" ]] then printf "\n +++ wildcards (*) not allowed with \"di\" file attribute." printf "\n see file \"$fileconnection_file\", line" printf "\n$line" locat=iofiles_file; exit fi multout[$iout]=true string=${localout_pre[$iout]} localout_pre[$iout]="${string%?}" else multout[$iout]=false fi fi elif [[ "$s2a" != in && "$s2a" != inopt && "$s2a" != out ]] then printf "\n +++ I/O-attribute in file $fileconnection_file has invalid" printf "\n value \"$s2\". Only \"in\", \"inopt\", and \"out\" are allowed!" locat=connect; exit fi IFS="$IFSALT" fi done < $fileconnection_file # VALUES OF PALMRUN-OPTIONS OVERWRITE THOSE FROM THE CONFIGURATION-FILE [[ "$palmrun_memory" != "" ]] && memory=$palmrun_memory [[ "$palmrun_cpumax" != "" ]] && cpumax=$palmrun_cpumax [[ "$palmrun_cores" != "" ]] && cores=$palmrun_cores [[ "$max_par_io_str" != "" ]] && maximum_parallel_io_streams=$max_par_io_str [[ "$palmrun_tasks_per_node" != "" ]] && tasks_per_node=$palmrun_tasks_per_node # EVALUATE MODEL COUPLING FEATURES (OPTION -Y) if [[ $run_coupled_model = true ]] then cores_atmos=`echo $coupled_dist | cut -d" " -s -f1` cores_ocean=`echo $coupled_dist | cut -d" " -s -f2` if (( $cores_ocean + $cores_atmos != $cores )) then printf "\n +++ number of processors does not fit to specification by \"-Y\"." printf "\n PEs (total) : $cores" printf "\n PEs (atmosphere): $cores_atmos" printf "\n PEs (ocean) : $cores_ocean" locat=coupling; exit fi fi # IF I AM IN BATCH MODE, CHECK IF EXECUTABLE AND OTHER REQUIRED FILES # HAVE BEEN GENERATED BY PALMBUILD AND STORED IN THE SOURCES_FOR_RUN_... # FOLDER if [[ $running_in_batch_mode = true ]] then if [[ ! -d ${fast_io_catalog}/${sources_for_run_catalog} ]] then printf "\n +++ directory ${fast_io_catalog}/${sources_for_run_catalog} is missing" printf "\n Please check the output of the palmrun-call" printf "\n that you did on your local host." locat=SOURCES_FOR_RUN; exit fi else # CREATE THE SOURCES_FOR_RUN_... FOLDER, BUT NOT IF I AM PART OF AN # AUTOMATIC RESTART RUN # AUTOMATIC RESTART RUNS JUST ACCESS THE DIRECTORY CREATED BY THE INITIAL RUN if [[ $restart_run = false ]] then # COLLECT FILES TO BE COMPILED IN THE SOURCES_FOR_RUN_... FOLDER ON # THE LOCAL HOST if [[ ! -d $source_path ]] then printf "\n\n +++ source path \"$source_path\" on local host" printf "\n \"$(hostname)\" does not exist" locat=source_path; exit fi rm -rf ${base_directory}/${sources_for_run_catalog} mkdir -p ${base_directory}/${sources_for_run_catalog} if [[ "$source_list" = LM ]] then # DETERMINE MODIFIED FILES OF THE SVN WORKING COPY source_list="" cd $source_path # CHECK, IF TRUNK-DIRECTORY IS UNDER SVN CONTROL if [[ ! -d ../.svn ]] then printf "\n\n +++ source directory" printf "\n \"$source_path\" " printf "\n is not under control of \"subversion\"." printf "\n Please do not use palmrun-option \"-s LM\"\n" fi # LIST ALL MODIFIED SOURCE CODE FILES Filenames="" svn status > tmp_svnstatus while read line do firstc=`echo $line | cut -c1` if [[ $firstc = M || $firstc = "?" ]] then Name=`echo "$line" | cut -c8-` extension=`echo $Name | cut -d. -f2` if [[ "$extension" = f90 || "$extension" = F90 || "$extension" = f || "$extension" = F || "$extension" = c ]] then Filenames="$Filenames "$Name fi fi done < tmp_svnstatus rm -rf tmp_svnstatus # COPY FILES TO SOURCES_FOR_RUN_... for filename in $Filenames do cp $filename ${base_directory}/${sources_for_run_catalog} source_list=$source_list"$filename " done cd - > /dev/null # COPY FILES GIVEN BY OPTION -s TO DIRECTORY SOURCES_FOR_RUN_... elif [[ "$source_list" != "" ]] then cd $source_path for filename in $source_list do # SOURCE CODE FILE IS NOT ALLOWED TO INCLUDE PATH if [[ $(echo $filename | grep -c "/") != 0 ]] then printf "\n +++ source code file: $filename" printf "\n must not contain (\"/\") " locat=source; exit fi if [[ ! -f $filename ]] then printf "\n +++ source code file: $filename" printf "\n does not exist" locat=source; exit else cp $filename ${base_directory}/${sources_for_run_catalog} fi done cd - > /dev/null fi # CHECK, IF MAKEFILE EXISTS AND COPY IT TO THE SOURCES_FOR_RUN... DIRECTORY [[ "$makefile" = "" ]] && makefile=$source_path/Makefile if [[ ! -f $makefile ]] then printf "\n +++ file \"$makefile\" does not exist" locat=make; exit else cp $makefile ${base_directory}/${sources_for_run_catalog}/Makefile fi # COPY FILES FROM OPTIONAL SOURCE PATH GIVEN IN THE CONFIGURATION FILE if [[ "$user_source_path" != "" ]] then # DOES THE DIRECTORY EXIST? if [[ ! -d $user_source_path ]] then printf "\n\n *** INFORMATIVE: additional source code directory" printf "\n \"$user_source_path\" " printf "\n does not exist or is not a directory." printf "\n No source code will be used from this directory!\n" user_source_path="" if [[ $silent == false ]] then sleep 2 fi else cd $user_source_path Names=$(ls -1 *.f90 2>&1) [[ $(echo $Names | grep -c '*.f90') = 0 ]] && AddFilenames="$Names" Names=$(ls -1 *.F90 2>&1) [[ $(echo $Names | grep -c '*.F90') = 0 ]] && AddFilenames="$AddFilenames $Names" Names=$(ls -1 *.F 2>&1) [[ $(echo $Names | grep -c '*.F') = 0 ]] && AddFilenames="$AddFilenames $Names" Names=$(ls -1 *.f 2>&1) [[ $(echo $Names | grep -c '*.f') = 0 ]] && AddFilenames="$AddFilenames $Names" Names=$(ls -1 *.c 2>&1) [[ $(echo $Names | grep -c '*.c') = 0 ]] && AddFilenames="$AddFilenames $Names" cd - > /dev/null cd ${base_directory}/${sources_for_run_catalog} # COPY MAKEFILE IF EXISTING if [[ -f $user_source_path/Makefile ]] then printf "\n\n *** user Makefile from directory" printf "\n \"$user_source_path\" is used \n" if [[ $silent == false ]] then sleep 1 fi cp $user_source_path/Makefile . fi for filename in $AddFilenames do if [[ -f $filename ]] then printf "\n +++ source code file \"$filename\" found in additional" printf "\n source code directory \"$user_source_path\" " printf "\n but was also given with option \"-s\" which means that it should be taken" printf "\n from directory \"$source_path\"." locat=source; exit fi cp $user_source_path/$filename . source_list="$source_list $filename" # CHECK IF FILE IS CONTAINED IN MAKEFILE if [[ $(grep -c $filename Makefile) = 0 ]] then printf "\n\n +++ user file \"$filename\" " printf "\n is not listed in Makefile \n" locat=source; exit fi done cd - > /dev/null fi fi # COPY CONFIGURATION FILES cp $config_file ${base_directory}/${sources_for_run_catalog} cp $fileconnection_file ${base_directory}/${sources_for_run_catalog} # COPY SHELLSCRIPTS cp ${source_path}/../SCRIPTS/palmrun ${base_directory}/${sources_for_run_catalog} cp ${source_path}/../SCRIPTS/batch_scp ${base_directory}/${sources_for_run_catalog} fi fi # GET THE GLOBAL REVISION-NUMBER OF THE SVN-REPOSITORY # (HANDED OVER TO RESTART-RUNS USING OPTION -G) if [[ "$global_revision" = "" ]] then global_revision=`svnversion $source_path 2>/dev/null` global_revision="Rev: $global_revision" fi # IN CASE OF PARALLEL EXECUTION, CHECK SOME SPECIFICATIONS CONCERNING PROCESSOR NUMBERS if [[ -n $cores ]] then # CHECK, IF THE NUMBER OF CORES PER NODE HAS BEEN GIVEN UND IF IT IS AN # INTEGRAL DIVISOR OF THE TOTAL NUMBER OF CORES GIVEN BY OPTION -X if [[ $tasks_per_node = 0 ]] then if [[ $create_batch_job = true || $create_remote_batch_job = true ]] then printf "\n" printf "\n +++ option \"-T\" (tasks per node) is missing" printf "\n set -T option or define tasks_per_node in the config file" locat=tasks_per_node; (( iec = 0 )); exit else # DEFAULT FOR INTERACTIVE RUN tasks_per_node=$cores fi fi if (( cores < tasks_per_node )) then printf "\n" printf "\n +++ tasks per node (-T) cannot exceed total number of cores (-X)" printf "\n given values: -T $tasks_per_node -X $cores" locat=tasks_per_node; (( iec = 0 )); exit fi (( nodes = cores / ( tasks_per_node * threads_per_task ) )) (( mpi_tasks = cores / threads_per_task )) [[ $mpi_tasks = 0 ]] && (( mpi_tasks = 1 )) (( ii = cores / tasks_per_node )) (( remaining_cores = cores - ii * tasks_per_node )) if (( remaining_cores > 0 )) then printf "\n" printf "\n +++ WARNING: tasks per node (option \"-T\") is not an integral" printf "\n divisor of the total number of cores (option \"-X\")" printf "\n values of this palmrun-call: \"-T $tasks_per_node\" \"-X $cores\"" printf "\n One of the nodes is filled with $remaining_cores instead of $tasks_per_node tasks" (( nodes = nodes + 1 )) fi fi # SET DEFAULT VALUE FOR THE MAXIMUM NUMBER OF PARALLEL IO STREAMS if [[ "$maximum_parallel_io_streams" = "" ]] then maximum_parallel_io_streams=$cores fi # SET PORT NUMBER OPTION FOR CALLS OF SSH/SCP AND batch_scp SCRIPT if [[ "$scp_port" != "" ]] then PORTOPT="-P $scp_port" SSH_PORTOPT="-p $scp_port" fi # DETERMINE THE SSH-OPTION IN CASE THAT AN SSH-KEY IS EXPLICITLY GIVEN IN THE # CONFIG-FILE if [[ "$ssh_key" != "" ]] then ssh_key="-i $HOME/.ssh/$ssh_key" fi # SET QUEUE, IF NOT GIVEN if [[ $create_batch_job = true || $create_remote_batch_job = true ]] then if [[ $queue = none && "$defaultqueue" = "" ]] then printf "\n" printf "\n +++ no default queue given in configuration file and no queue" printf "\n given with option -q" locat=queue; exit fi if [[ $queue = none ]] then queue=$defaultqueue fi fi # GENERATE FULL FILENAMES OF INPUT-FILES, INCLUDING THEIR PATH # CHECK, IF INPUT-FILES EXIST, AND DETERMINE HIGHEST CYCLE NUMBER (IF CYCLES EXIST) (( i = 0 )) (( nr_of_input_files = 0 )) while (( i < iin )) do (( i = i + 1 )) # GENERATE PATH AND FULL FILE NAME (then-BRANCH: FIXED FULL NAME IS GIVEN, I.E. THE # FILE IDENTIFIER IS NOT PART OF THE FILENAME)) if [[ "${actionin_pre[$i]}" = di ]] then eval filename=${pathin_pre[$i]}/${endin_pre[$i]} else eval filename=${pathin_pre[$i]}/${run_identifier}${endin_pre[$i]} fi # CHECK IF FILE EXISTS if ! ls $filename* 1>/dev/null 2>&1 then # FILES WITH ATTRIBUTE opt ARE OPTIONAL. NO ABORT, IF THEY DO NOT EXIST. if [[ "${optin_pre[$i]}" != "yes" ]] then printf "\n\n +++ INPUT-file: " if [[ "${extin_pre[$i]}" = "" || "${extin_pre[$i]}" = " " ]] then printf "\n $filename" else printf "\n $filename.${extin_pre[$i]}" fi printf "\n does not exist\n" locat=input; exit else (( nr_of_input_files = nr_of_input_files + 1 )) localin[$nr_of_input_files]="${localin_pre[$i]}" optin[$nr_of_input_files]="${optin_pre[$i]}" actionin[$nr_of_input_files]="unavailable" pathin[$nr_of_input_files]="${pathin_pre[$i]}" endin[$nr_of_input_files]="${endin_pre[$i]}" extin[$nr_of_input_files]="${extin_pre[$i]}" fi else # FIRST CHECK FOR MULTIPLE NAMES WITH THE SAME BASENAME # ($run_identifier) AND CREATE A LIST FOR THE DETECTED BASENAME # ENDINGS if [[ "${multin[$i]}" = true ]] then # DETERMINE THE EXISTING EXTENSIONS FROM THE LIST OF FILES ls -1 -d ${filename} > filelist 2>/dev/null ls -1 -d ${filename}.* >> filelist 2>/dev/null ls -1 -d ${filename}_* >> filelist 2>/dev/null endings="" while read line do # filename without path (i.e. after the last "/") basefilename=$(basename ${line}) # check if there is an extension and remove it ext=${basefilename##*.} if [[ "$ext" = "${extin_pre[$i]}" ]] then basefilename=${basefilename%.*} fi # check for an existing cycle number and remove it cycle=${basefilename##*.} if [[ $cycle =~ ^-?[0-9]+$ ]] then basefilename=${basefilename%.*} fi # remove the run_identifier from the beginning length_run_identifier=${#run_identifier} ending=${basefilename:${length_run_identifier}} # remove the ending given in the .iofiles from the beginning endingstring="${endin_pre[$i]}" length_ending=${#endingstring} ending=${ending:${length_ending}} if [[ "$ending" = "" ]] then # standard ending as given in the .iofiles if [[ $(echo $endings | grep -c DEFAULT) = 0 ]] then endings="$endings DEFAULT" fi else # ending must start with "_", otherwise its a different file if [[ "${ending:0:1}" = "_" ]] then if [[ $(echo $endings | grep -c "$ending") = 0 ]] then endings="$endings $ending" fi fi fi done filelist 2>/dev/null ls -1 -d $filename.* >> filelist 2>/dev/null while read line do # filename without path (i.e. after the last "/") basefilename=$(basename ${line}) # check if there is an extension extension=${basefilename##*.} if [[ "$extension" = "${extin[$nr_of_input_files]}" ]] then basefilename=${basefilename%.*} fi # check for an existing cycle number cycle=${basefilename##*.} if [[ $cycle =~ ^-?[0-9]+$ ]] then # NUMBERS WITH LEADING ZEROS ARE INTERPRETED AS OCTAL NUMBERS # 10# EXPLICITLY SPECIFIES THE NUMBER BASE AS 10 (( icycle = $((10#$cycle)) )) else (( icycle = 0 )) fi if (( icycle > maxcycle )) then (( maxcycle = icycle )) # FOR COMPATIBILITY REASONS WITH OLDER VERSIONS # CHECK IF CYCLE NUMBER CONTAINS LEADING ZEROS if [[ $(echo $cycle | cut -c1) = 0 ]] then leading_zero=true else leading_zero=false fi fi done 0 )) then if [[ "${extin[$nr_of_input_files]}" != " " && "${extin[$nr_of_input_files]}" != "" ]] then filename=${filename}.$cyclestring.${extin[$nr_of_input_files]} else filename=${filename}.$cyclestring fi else if [[ "${extin[$nr_of_input_files]}" != " " && "${extin[$nr_of_input_files]}" != "" ]] then filename=${filename}.${extin[$nr_of_input_files]} fi fi # STORE FILENAME WITHOUT PATH BUT WITH CYCLE NUMBER, # IS LATER USED FOR TRANSFERRING FILES WIHIN THE JOB (SEE END OF FILE) absnamein[$nr_of_input_files]=$filename if (( maxcycle > 0 )) then if [[ "${actionin[$nr_of_input_files]}" = di ]] then frelin[$nr_of_input_files]=${endin[$nr_of_input_files]}.$cyclestring else frelin[$nr_of_input_files]=${run_identifier}${endin[$nr_of_input_files]}.$cyclestring fi else if [[ "${actionin[$nr_of_input_files]}" = di ]] then frelin[$nr_of_input_files]=${endin[$nr_of_input_files]} else frelin[$nr_of_input_files]=${run_identifier}${endin[$nr_of_input_files]} fi fi done fi done # GENERATE FULL FILENAMES OF OUTPUT-FILES (WITHOUT $ OR ~), # CHECK, IF OUTPUT-FILES EXIST, AND DETERMINE HIGHEST CYCLE NUMBER (IF CYCLES EXIST), # OR, IN CASE THAT FILE DOES NOT EXIST, CHECK, IF IT CAN BE CREATED # THESE ACTIONS ARE NOT CARRIED OUT, IF FILES SHALL BE TRANSFERRED FROM THE REMOTE TO # THE LOCAL HOST (BECAUSE THEIR IS NO DIRECT ACCESS TO THE LOCAL DIRECTORIES FROM THE # REMOTE HOST) (( i = 0 )) while (( i < iout )) do (( i = i + 1 )) if [[ ! ( $running_on_remote = true && ( "${actionout_pre[$i]}" = tr || "${actionout_pre[$i]}" = tra || "${actionout_pre[$i]}" = trpe ) ) ]] then if [[ "${actionout_pre[$i]}" = tr ]] then actionout_pre[$i]="" elif [[ "${actionout_pre[$i]}" = trpe ]] then actionout_pre[$i]=pe elif [[ "${actionout_pre[$i]}" = tra ]] then actionout_pre[$i]=a fi (( maxcycle = 0 )) eval filename=${pathout_pre[$i]}/${run_identifier}${endout_pre[$i]} eval catalogname=${pathout_pre[$i]} if ! ls $filename* 1>/dev/null 2>&1 then # IF OUTPUT-FILE DOES NOT EXIST CHECK, IF IT CAN BE CREATED if cat /dev/null > $filename then rm $filename else # CHECK, IF THE DIRECTORY WHERE FILE SHALL BE COPIED TO EXISTS # IF IT DOES NOT EXIST, TRY TO CREATE IT if [[ ! -d $catalogname ]] then if mkdir -p $catalogname then printf "\n\n *** directory:" printf "\n $catalogname" printf "\n was created\n" else printf "\n\n +++ OUTPUT-file:" printf "\n $filename" printf "\n cannot be created, because directory does not exist" printf "\n and cannot be created either" printf "\n" locat=output ; exit fi 2>/dev/null else printf "\n\n +++ OUTPUT-file:" printf "\n $filename" printf "\n cannot be created, although directory exists" printf "\n" locat=output ; exit fi fi 2>/dev/null fi fi done # DETERMINE THE NAME OF PALMRUN'S TEMPORARY WORKING DIRECTORY if [[ $running_in_batch_mode = false ]] then run_id_number=$RANDOM run_id=${run_identifier}.$run_id_number tempdir=$fast_io_catalog/$run_id # FOR COMPATIBILITY REASONS WITH OLDER VERSIONS SET JOB_ID export job_id=$run_id fi # CHECK SETTINGS REQUIRED FOR BATCH JOBS if [[ $create_batch_job = true || $create_remote_batch_job = true ]] then # CHECK, IF JOB DIRECTIVES HAVE BEEN GIVEN IN CONFIGURATION FILE if [[ $ibd = 0 ]] then printf "\n" printf "\n +++ no batch directives found in configuration file" locat=config_file_batch_directives; (( iec = 0 )); exit fi # CHECK IF CPUTIME IS GIVEN FOR JOB done=false cputime=$cpumax while [[ $done = false ]] do if (( cputime == 0 )) then printf "\n +++ cpu-time is undefined" printf "\n >>> Please type CPU-time in seconds as INTEGER:" printf "\n >>> " read cputime 1>/dev/null 2>&1 else done=true fi done cpumax=$cputime # CHECK THE MEMORY DEMAND done=false while [[ $done = false ]] do if (( memory == 0 )) then printf "\n +++ memory demand is undefined" printf "\n >>> Please type memory in MByte per process as INTEGER:" printf "\n >>> " read memory 1>/dev/null 2>&1 else done=true fi done # IN CASE OF REMOTE-JOBS CHECK, IF A USERNAME FOR THE REMOTE HOST IS GIVEN if [[ $create_remote_batch_job = true && -z $remote_username ]] then while [[ -z $remote_username ]] do printf "\n +++ username on remote host with IP \"$remote_ip\" is undefined" printf "\n >>> Please type username:" printf "\n >>> " read remote_username done fi else if [[ $running_in_batch_mode = false ]] then cputime=10000000 # NO LIMT FOR INTERACTIVE RUNS cpumax=$cputime else cputime=$cpumax fi fi # CALCULATE HOURS/MINUTES/SECONDS, E.G. FOR BATCH-DIRECTIVES (( cpu_hours = cputime / 3600 )) (( resttime = cputime - cpu_hours * 3600 )) (( cpu_minutes = resttime / 60 )) (( cpu_seconds = resttime - cpu_minutes * 60 )) timestring=${cpu_hours}:${cpu_minutes}:${cpu_seconds} # OUTPUT OF THE PALMRUN-HEADER calltime=$(date) printf "\n" printf "#------------------------------------------------------------------------# \n" printf "| %-35s%35s | \n" "$version" "$calltime" printf "| %-35s%35s | \n" "PALM code $global_revision" " " printf "| | \n" column1="called on:"; column2=$(hostname) printf "| %-25s%-45s | \n" "$column1" "$column2" if [[ $create_remote_batch_job = true ]] then column1="execution on:"; column2="$configuration_identifier (username: $remote_username)" else if [[ $running_on_remote = true ]] then column1="config. identifier:"; column2="$configuration_identifier (execute on IP: $remote_ip)" else column1="config. identifier:"; column2="$configuration_identifier (execute on IP: $local_ip)" fi fi printf "| %-25s%-45s | \n" "$column1" "$column2" column1="running in:" if [[ $running_in_batch_mode = true ]] then column2="batch job mode" else if [[ $create_batch_job = true || $create_remote_batch_job = true ]] then column2="job creation mode" else column2="interactive run mode" fi fi printf "| %-25s%-45s | \n" "$column1" "$column2" if [[ $running_in_batch_mode = true || $create_batch_job = true || $create_remote_batch_job = true ]] then if [[ "$project_account" != "" ]] then column1="project account number:" column2="$project_account" printf "| %-25s%-45s | \n" "$column1" "$column2" fi fi if [[ -n $cores ]] then if [[ $run_coupled_model = false ]] then column1="number of cores:"; column2=$cores else column1="number of cores:"; column2="$cores (atmosphere: $cores_atmos, ocean: $cores_ocean)" fi printf "| %-25s%-45s | \n" "$column1" "$column2" fi if [[ -n $tasks_per_node ]] then column1="tasks per node:"; column2="$tasks_per_node (number of nodes: $nodes)" printf "| %-25s%-45s | \n" "$column1" "$column2" if (( remaining_cores > 0 )) then column1=" "; column2="one of the nodes only filled with $remaining_cores tasks" printf "| %-25s%-45s | \n" "$column1" "$column2" fi fi if [[ $maximum_parallel_io_streams != $cores ]] then column1="max par io streams:"; column2="$maximum_parallel_io_streams" printf "| %-25s%-45s | \n" "$column1" "$column2" fi if [[ $use_openmp = true ]] then column1="threads per task:"; column2="$threads_per_task" printf "| %-25s%-45s | \n" "$column1" "$column2" fi if [[ $create_batch_job = true || $create_remote_batch_job = true || $running_in_batch_mode = true ]] then column1="memory demand / PE":; column2="$memory MB" printf "| %-25s%-45s | \n" "$column1" "$column2" column1="job cpu time (h:m:s):"; column2="$timestring" printf "| %-25s%-45s | \n" "$column1" "$column2" fi printf "| | \n" if [[ "$source_list" != "" ]] then if [[ "$make_options" != "" ]] then column1="make options:"; column2=$(echo "$make_options" | cut -c-45) printf "| %-25s%-45s | \n" "$column1" "$column2" line=$(echo "$make_options" | cut -c46-) while [[ "$line" != "" ]] do column1="" column2=$(echo "$line" | cut -c-45) printf "| %-25s%-45s | \n" "$column1" "$column2" line=$(echo "$line" | cut -c46-) done fi fi column1="cpp directives:"; column2=$(echo "$cpp_options" | cut -c-45) printf "| %-25s%-45s | \n" "$column1" "$column2" line=$(echo "$cpp_options" | cut -c46-) while [[ "$line" != "" ]] do column1="" column2=$(echo "$line" | cut -c-45) printf "| %-25s%-45s | \n" "$column1" "$column2" line=$(echo "$line" | cut -c46-) done column1="compiler options:"; column2=$(echo "$compiler_options" | cut -c-45) printf "| %-25s%-45s | \n" "$column1" "$column2" line=$(echo "$compiler_options" | cut -c46-) while [[ "$line" != "" ]] do column1="" column2=$(echo "$line" | cut -c-45) printf "| %-25s%-45s | \n" "$column1" "$column2" line=$(echo "$line" | cut -c46-) done column1="linker options:"; column2=$(echo "$linker_options" | cut -c-45) printf "| %-25s%-45s | \n" "$column1" "$column2" line=$(echo "$linker_options" | cut -c46-) while [[ "$line" != "" ]] do column1="" column2=$(echo "$line" | cut -c-45) printf "| %-25s%-45s | \n" "$column1" "$column2" line=$(echo "$line" | cut -c46-) done if [[ "$login_init_cmd" != "" ]] then column1="login init commands:"; column2=$(echo "$login_init_cmd" | cut -c-45) printf "| %-25s%-45s | \n" "$column1" "$column2" line=$(echo "$login_init_cmd" | cut -c46-) while [[ "$line" != "" ]] do column1="" column2=$(echo "$line" | cut -c-45) printf "| %-25s%-45s | \n" "$column1" "$column2" line=$(echo "$line" | cut -c46-) done fi if [[ "$module_commands" != "" ]] then column1="module commands:"; column2=$(echo "$module_commands" | cut -c-45) printf "| %-25s%-45s | \n" "$column1" "$column2" line=$(echo "$module_commands" | cut -c46-) while [[ "$line" != "" ]] do column1="" column2=$(echo "$line" | cut -c-45) printf "| %-25s%-45s | \n" "$column1" "$column2" line=$(echo "$line" | cut -c46-) done fi printf "| | \n" column1="run identifier:"; column2=$run_identifier printf "| %-25s%-45s | \n" "$column1" "$column2" column1="activation string list:"; column2=$(echo $activation_string_list) printf "| %-25s%-45s | \n" "$column1" "$column2" if [[ "$ocean_file_appendix" = true ]] then printf "| %-35s%-35s | \n" "suffix \"_O\" is added to local files" " " fi if [[ "$source_list" != "" ]] then printf "| | \n" printf "| Files to be compiled: | \n" line=$source_list while [[ "$line" != "" ]] do linestart=$(echo $line | cut -c-70) printf "| %-70s | \n" "$linestart" line=$(echo "$line" | cut -c71-) done fi printf "#------------------------------------------------------------------------#" # OUTPUT OF FILE CONNECTIONS IN CASE OF TRACEBACK if [[ $do_trace = true ]] then (( i = 0 )) while (( i < nr_of_input_files )) do (( i = i + 1 )) if (( i == 1 )) then printf "\n\n >>> INPUT-file assignments:\n" fi printf "\n ${localin[$i]} : ${absnamein[$i]}" done (( i = 0 )) while (( i < iout )) do (( i = i + 1 )) if (( i == 1 )) then printf "\n\n >>> OUTPUT-file assignments:\n" fi printf "\n ${localout[$i]} : ${pathout[$i]}" done (( i = 0 )) while (( i < iic )) do (( i = i + 1 )) if (( i == 1 )) then printf "\n\n >>> INPUT-commands:\n" fi printf "\n ${in_command[$i]}" done (( i = 0 )) while (( i < ioc )) do (( i = i + 1 )) if (( i == 1 )) then printf "\n\n >>> OUTPUT-commands:\n" fi printf "\n ${out_command[$i]}" done fi # QUERY FOR CONTINUE if [[ $silent = false && $running_in_batch_mode = false ]] then antwort=dummy printf "\n\n" printf " >>> everything o.k. (y/n) ? " while read antwort do if [[ "$antwort" != y && "$antwort" != Y && "$antwort" != n && "$antwort" != N ]] then printf " >>> everything o.k. (y/n) ? " else break fi done if [[ $antwort = n || $antwort = N ]] then locat=user_abort; (( iec = 0 )); exit fi if [[ $create_batch_job = true || $create_remote_batch_job = true ]] then printf "\n *** batch-job will be created and submitted" else printf "\n *** PALMRUN will now continue to execute on this machine" fi fi # PROVIDE FILES TO EXECUTE PALM AND CREATE THE EXECUTABLE if [[ $restart_run = false && $running_in_batch_mode = false ]] then if [[ $create_batch_job = true || $create_remote_batch_job = true ]] then printf "\n\n *** creating executable and other sources for the remote host\n" else printf "\n\n *** creating executable and other sources for the local host\n" fi # FIRST CHECK, IF A MAKE DEPOSITORY EXISTS, AND IF NOT, ASK THE USER IF # IT SHALL BE CREATED ask_for_make_depository=false if [[ $create_remote_batch_job = true ]] then line=`grep %base_directory $config_file` make_depository=`echo $line | cut -d" " -s -f2`/MAKE_DEPOSITORY_${configuration_identifier} ssh -q $ssh_key ${remote_username}@${remote_ip} "[[ ! -d ${make_depository} ]] && echo depository not found" 2>&1 | tee ${configuration_identifier}_last_make_protokoll if [[ $(grep -c "depository not found" ${configuration_identifier}_last_make_protokoll) != 0 ]] then printf "\n\n +++ make depository \"${make_depository}\"" printf "\n on remote host not found!" ask_for_make_depository=true fi rm ${configuration_identifier}_last_make_protokoll else # CHECK FOR MAKE_DEPOSITORY ON THE LOCAL HOST make_depository=${base_directory}/MAKE_DEPOSITORY_${configuration_identifier} if [[ ! -d ${make_depository} ]] then printf "\n\n +++ make depository \"${make_depository}\"" printf "\n on local host not found!" ask_for_make_depository=true fi fi if [[ $ask_for_make_depository = true ]] then antwort=dummy printf "\n\n" printf " >>> Create a new one (y/n) ? " while read antwort do if [[ "$antwort" != y && "$antwort" != Y && "$antwort" != n && "$antwort" != N ]] then printf " >>> Create a new one (y/n) ? " else break fi done if [[ $antwort = n || $antwort = N ]] then locat=user_abort; (( iec = 0 )); exit fi if [[ $do_trace = true ]] then palmbuild -c $configuration_identifier else palmbuild -v -c $configuration_identifier fi if [[ ${PIPESTATUS[0]} != 0 ]] then # ABORT IN CASE OF COMPILATION PROBLEMS printf "\n +++ error while compiling for the MAKE_DEPOSITORY" locat=make_depository exit else echo " *** now continue with creating executable and other sources" fi fi # NOW CREATE THE SOURCES_FOR_RUN FOLDER palmbuild -v $use_existing_sources_folder -c $configuration_identifier -r $run_identifier if [[ ${PIPESTATUS[0]} != 0 ]] then # ABORT IN CASE OF COMPILATION PROBLEMS printf "\n +++ error while creating executable and/or other sources" locat=execution rm -rf ${base_directory}/${sources_for_run_catalog} exit else printf " *** executable and other sources created\n" rm -rf ${base_directory}/${sources_for_run_catalog} fi fi # WHEN CREATING A REMOTE BATCH JOB, THOSE INPUT FILES WITH TRANSFER-ATTRIBUT # WILL BE COPIED TO THE REMOTE HOST if [[ $create_remote_batch_job = true ]] then (( i = 0 )) while (( i < nr_of_input_files )) do (( i = i + 1 )) if [[ "${actionin[$i]}" = tr ]] then eval inputfile=${pathin[$i]}/${frelin[$i]} scp -q $ssh_key $PORTOPT $inputfile ${remote_username}@${remote_ip}:${fast_io_catalog}/${sources_for_run_catalog}/${frelin[$i]} fi done if (( i > 0 )) then printf "\n\n *** input files have been copied to the remote host\n" fi fi # NOW PERFORM THOSE ACTIONS REQUIRED TO EXECUTE THE PROGRAM (PALM) ON THIS MACHINE # (COMPILING/LINKING, EXECUTING, COPYING I/O FILES) if [[ $create_batch_job = false && $create_remote_batch_job = false ]] then # CHANGE TO THE TEMPORARY WORKING DIRECTORY if [[ $running_in_batch_mode = false ]] then # CREATE THE DIRECTORY mkdir -p $tempdir if [[ $? != 0 ]] then printf "\n +++ temporary working directory cannot be created." printf "\n Check setting of variable fast_io_catalog in your config file." locat=execution exit fi chmod go+rx $tempdir cd $tempdir printf "\n *** changed to temporary directory: $tempdir" else # IN BATCH MODE PALMRUN IS CALLED FROM TEMPDIR printf "\n *** running in temporary directory: $tempdir" fi # PROVIDE THE INPUT FILES # LOOP OVER ALL ACTIVATED FILES (LISTED IN THE CONFIGURATION FILE) optional_files_missing=false (( i = 0 )) while (( i < nr_of_input_files )) do (( i = i + 1 )) if (( i == 1 )) then printf "\n\n *** providing INPUT-files:\n$dashes" fi # SKIP OPTIONAL FILES, IF THEY DO NOT EXIST if [[ "${actionin[$i]}" = unavailable ]] then optional_files_missing=true continue fi # CHECK FOR SINGLE FILE (SERIAL RUN) OR DIRECTORY (ONE FILE PER CORE FOR PARELLEL EXECUTION) files_for_cores=false; filetype=file if [[ "${actionin[$i]}" = pe && -n $cores ]] then files_for_cores=true; filetype=files actionin[$i]="" elif [[ "${actionin[$i]}" = pe && ! -n $cores ]] then actionin[$i]="" elif [[ "${actionin[$i]}" = lnpe && -n $cores ]] then files_for_cores=true; filetype=files actionin[$i]="ln" elif [[ "${actionin[$i]}" = lnpe && ! -n $cores ]] then actionin[$i]="ln" fi if [[ $files_for_cores = true ]] then printf "\n >>> INPUT: ${absnamein[$i]}/.... to ${localin[$i]}" else printf "\n >>> INPUT: ${absnamein[$i]} to ${localin[$i]}" fi # INPUT-FILES TO BE LINKED if [[ "${actionin[$i]}" = ln ]] then printf "\n $filetype will be linked" if [[ $files_for_cores = false ]] then if [[ -f "${absnamein[$i]}" ]] then ln ${absnamein[$i]} ${localin[$i]} got_tmp[$i]=true fi else if [[ -d "${absnamein[$i]}" ]] then mkdir -p ${localin[$i]} cd ${absnamein[$i]} for file in $(ls *) do ln $file $tempdir/${localin[$i]} done >|/dev/null 2>&1 cd $tempdir fi # IF "ln -f" HAS FAILED DO A NORMAL COPY "cp -r" if [[ ! -f "${localin[$i]}/_000000" ]] then printf "\n --- WARNING: ln failed, using cp instead (might be time consuming...)" cp -r ${absnamein[$i]}/* ${localin[$i]} fi got_tmp[$i]=true fi fi # FILE IS STORED IN THE RESPECTIVE DIRECTORY GIVEN IN THE CONFIGURATION FILE if [[ "${actionin[$i]}" = "" || "${actionin[$i]}" = "di" || "${actionin[$i]}" = "tr" || "${actionin[$i]}" = "npe" ]] then if [[ "${actionin[$i]}" = "npe" && -n $cores ]] then # FILE COPIES ARE PROVIDED FOR ALL CORES # EACH FILE GETS A UNIQUE FILENAME WITH A FOUR DIGIT NUMBER printf "\n file will be provided for $cores processors" mkdir -p ${localin[$i]} ival=$cores (( ii = 0 )) while (( ii <= ival-1 )) do if (( ii < 10 )) then cp ${absnamein[$i]} ${localin[$i]}/_000$ii elif (( ii < 100 )) then cp ${absnamein[$i]} ${localin[$i]}/_00$ii elif (( ii < 1000 )) then cp ${absnamein[$i]} ${localin[$i]}/_0$ii else cp ${absnamein[$i]} ${localin[$i]}/_$ii fi (( ii = ii + 1 )) done else if [[ $files_for_cores = true ]] then # PROVIDE FILES FOR EACH CORE # FIRST CREATE THE LOCAL DIRECTORY, THEN COPY FILES # FROM THE PERMANENT DIRECTORY BY LINKING THEM TO THE LOCAL ONE printf "\n providing $cores files for the respective cores" mkdir -p ${localin[$i]} if [[ $link_local_input = true ]] then printf " files will be linked\n" cd ${absnamein[$i]} for file in $(ls *) do ln -f $file ${localin[$i]} done cd $tempdir fi # IF "ln -f" FAILED OR IF "$link_local_input = false" DO A NORMAL "cp -r" if [[ ! -f "${localin[$i]}/_000000" ]] then if [[ $link_local_input = true ]] then printf "\n --- WARNING: ln failed, using cp instead (might be time consuming...)" fi cp -r ${absnamein[$i]}/* ${localin[$i]} fi else # PROVIDE FILE FOR RUNS ON A SINGLE CORE if [[ $link_local_input = true ]] then printf " file will be linked\n" ln -f ${absnamein[$i]} ${localin[$i]} fi # If "ln -f" fails or if "$link_local_input = false" do a normal "cp" if [[ ! -f "${localin[$i]}" ]] then if [[ $link_local_input = true ]] then printf "\n --- WARNING: ln failed, using cp instead (might be time consuming...)" fi if [[ $running_on_remote = true && "${actionin[$i]}" = tr ]] then mv ${absnamein[$i]} ${localin[$i]} else cp ${absnamein[$i]} ${localin[$i]} fi fi fi fi fi done if (( i != 0 )) then if [[ $optional_files_missing = true ]] then printf "\n *** INFORMATIVE: some optional INPUT-files are not present" fi printf "\n$dashes\n *** all INPUT-files provided \n" fi # NOW COPY FILES (*.f90, *.o, config files, etc.) FROM SOURCES_FOR_RUN_... TO THE TEMPORARY # WORKING DIRECTORY cp ${fast_io_catalog}/${sources_for_run_catalog}/{*,.[!.]*} $tempdir # EXECUTE INPUT-COMMANDS GIVEN IN THE CONFIGURATION FILE (( i = 0 )) while (( i < iic )) do (( i = i + 1 )) if (( i == 1 )) then printf "\n\n *** execution of INPUT-commands:\n$dashes" fi printf "\n >>> ${in_command[$i]}" eval ${in_command[$i]} if (( i == iic )) then printf "\n$dashes\n" fi done # CHECK IF THE PROGRESS BAR NEEDS TO BE DISABLED if [[ $running_in_batch_mode = true || $running_in_test_mode = true ]] then progress_bar_disabled=true else progress_bar_disabled=false fi # CREATE THE NAMELIST-FILE WITH VALUES OF ENVIRONMENT-VARIABLES REQUIRED BY PALM # (FILE ENVPAR WILL BE READ BY PALM) cat > ENVPAR << EOF &envpar run_identifier = '$run_identifier', host = '$configuration_identifier', write_svf = .${write_svf}., write_binary = .${write_binary}., read_svf = .${read_svf}., tasks_per_node = $tasks_per_node, maximum_parallel_io_streams = $maximum_parallel_io_streams, maximum_cpu_time_allowed = ${cpumax}., revision = '$global_revision', progress_bar_disabled = .${progress_bar_disabled}. / EOF # STARTING THE EXECUTABLE printf "\n\n *** execution starts in directory\n \"`pwd`\"\n$dashes\n" PATH=$PATH:$tempdir # REPLACE PARAMETERS IN THE EXECUTION COMMAND WITH REAL VALUES line=`echo "${execute_command}" | sed 's/{{/$/g' | sed 's/}}//g'` line2=`echo "${execute_command}" | sed 's/{{mpi_tasks}}/1/g' | sed 's/{{tasks_per_node}}/1/g' | sed 's/palm/combine_plot_fields.x/g'` eval line=\"$line\" execute_command="$line" # EXECUTION COMMAND FOR COMBINE_PLOT_FIELDS if [[ "$execute_command_for_combine" = "" ]] then eval line2=\"$line2\" execute_command_for_combine="$line2" fi # PROVIDE A HOSTFILE, IF REQUIRED if [[ "$hostfile" != "" ]] then if [[ $hostfile = auto ]] then # CREATE A NEW HOSTFILE (( ii = 1 )) while (( ii <= cores / threads_per_task )) do echo $(hostname) >> hostfile (( ii = ii + 1 )) done if (( cores / threads_per_task == 0 )) then echo $(hostname) >> hostfile fi else cp $hostfile hostfile fi eval line=\"`head -n $ii hostfile`\" printf "\n *** running on: $line" fi # SET THE NUMBER OF OPENMP-THREADS if [[ $use_openmp = true ]] then export OMP_NUM_THREADS=$threads_per_task printf "\n *** number of OpenMP threads per MPI-task: $OMP_NUM_THREADS" else export OMP_NUM_THREADS=1 fi # PROVIDE DATA FOR ATMOSPHERE OCEAN COUPLING if [[ $run_coupled_model = false ]] then if [[ "$ocean_file_appendix" = true ]] then echo "precursor_ocean" > coupling_steering else echo "precursor_atmos" > coupling_steering fi else (( iia = $cores_atmos / $threads_per_task )) (( iio = $cores_ocean / $threads_per_task )) printf "\n coupled run ($iia atmosphere, $iio ocean)" printf "\n\n" echo "coupled_run $iia $iio" > coupling_steering fi printf "\n *** execute command:" printf "\n \"$execute_command\" \n\n" if [[ $progress_bar_disabled = true ]] then $execute_command < coupling_steering &> >(grep -v --line-buffered -e '^STOP 1$' -e '^1$' &> >(tee STDOUT) ) exit_code=${PIPESTATUS[0]} else $execute_command < coupling_steering &> >(tee STDOUT) exit_code=${PIPESTATUS[0]} fi if [[ ${exit_code} != 0 ]] then # ABORT IN CASE OF RUNTIME ERRORS printf "\n +++ runtime error occured" locat=execution exit else printf "\n$dashes\n *** execution finished \n" fi # CALL OF combine_plot_fields IN ORDER TO MERGE SINGLE FILES WRITTEN # BY EACH CORE INTO ONE FILE if [[ ! -f combine_plot_fields.x ]] then printf "\n\n\n +++ WARNING: no combine_plot_fields found" printf "\n 2d- and/or 3d-data may be incomplete!" printf "\n Your previous palmbuild may have failed. Please check.\n" elif [[ "$combine_plot_fields" == true ]] then printf "\n\n\n *** post-processing: now executing \"$execute_command_for_combine\" ..." $execute_command_for_combine else # TEMPORARY SOLUTION TO SKIP combine_plot_fields. THIS IS REQUIRED IN CASE OF HUGE AMOUNT OF # DATA OUTPUT printf "\n\n\n *** post-processing: skipping combine_plot_fields (-Z option set) ..." fi # EXECUTE OUTPUT-COMMANDS GIVEN IN THE CONFIGURATION FILE (( i = 0 )) while (( i < ioc )) do (( i = i + 1 )) if (( i == 1 )) then printf "\n\n *** execution of OUTPUT-commands:\n$dashes" fi # REPLACE PARAMETERS IN THE OUTPUT COMMAND WITH REAL VALUES out_command[$i]=`echo "${out_command[$i]}" | sed 's/{{/$/g' | sed 's/}}//g'` printf "\n >>> ${out_command[$i]}" eval ${out_command[$i]} if (( i == ioc )) then printf "\n$dashes\n" fi done # IN A FIRST PASS, ADD ADDITIONAL OUTPUT FILE CONNECTIONS IN CASE OF # WILDCARDS (( i = 0 )) (( nr_of_output_files = 0 )) while (( i < iout )) do (( i = i + 1 )) # FIRST CHECK FOR MULTIPLE NAMES WITH THE SAME LOCAL NAME AND # CREATE A LIST FOR THE DETECTED ENDINGS if [[ "${multout[$i]}" = true ]] then # DETERMINE THE EXISTING EXTENSIONS FROM THE LIST OF FILES ls -1 -d ${localout_pre[$i]} > filelist 2>/dev/null ls -1 -d ${localout_pre[$i]}_* >> filelist 2>/dev/null endings="DEFAULT" while read line do # remove the local name from the beginning localnamestring="${localout_pre[$i]}" length_localname=${#localnamestring} ending=${line:${length_localname}} if [[ "$ending" != "" ]] then endings="$endings $ending" fi done filelist 2>/dev/null ls -1 -d $filename.* >> filelist 2>/dev/null while read line do # filename without path (i.e. after the last "/") basefilename=$(basename ${line}) # check if there is an extension extension=${basefilename##*.} if [[ "$extension" = "${extout[$i]}" ]] then basefilename=${basefilename%.*} fi # check for an existing cycle number cycle=${basefilename##*.} if [[ $cycle =~ ^-?[0-9]+$ ]] then # NUMBERS WITH LEADING ZEROS ARE INTERPRETED AS OCTAL NUMBERS # 10# EXPLICITLY SPECIFIES THE NUMBER BASE AS 10 (( icycle = $((10#$cycle)) + 1 )) else (( icycle = 1 )) fi if (( icycle > maxcycle )) then (( maxcycle = icycle )) fi done = cycnum[$i] )) then (( cycnum[$i] = run_number )) else if (( run_number > 0 )) then printf "\n --- INFORMATIVE: The following file cannot get a unified cycle number" fi fi fi if (( cycnum[$i] > 0 )) then cyclestring=`printf "%03d" ${cycnum[$i]}` pathout[$i]=${pathout[$i]}.$cyclestring fi fi # CHECK FOR SINGLE FILE (SERIAL RUN) OR DIRECTORY (ONE FILE PER CORE FOR PARELLEL EXECUTION) files_for_cores=false; filetype=file link_local_output=false if [[ "${actionout[$i]}" = pe && -n $cores ]] then files_for_cores=true; filetype=directory actionout[$i]="" elif [[ "${actionout[$i]}" = pe && ! -n $cores ]] then actionout[$i]="" elif [[ "${actionout[$i]}" = lnpe && -n $cores ]] then files_for_cores=true; filetype=directory link_local_output=true actionout[$i]="" elif [[ "${actionout[$i]}" = lnpe && ! -n $cores ]] then link_local_output actionout[$i]="" elif [[ "${actionout[$i]}" = trpe && -n $cores ]] then files_for_cores=true; filetype=directory actionout[$i]="tr" elif [[ "${actionout[$i]}" = trpe && ! -n $cores ]] then actionout[$i]="tr" fi if [[ ! -f ${localout[$i]} && $files_for_cores = false ]] then if [[ ${warnout[$i]} = true ]] then printf "\n +++ temporary OUTPUT-file ${localout[$i]} does not exist\n" fi elif [[ ! -d ${localout[$i]} && $files_for_cores = true ]] then if [[ ${warnout[$i]} = true ]] then printf "\n +++ temporary OUTPUT-file ${localout[$i]}/.... does not exist\n" fi else # COPY VIA SCP TO LOCAL HOST (ALWAYS IN BINARY MODE USING batch_scp option -m) # IF TARGET DIRECTORY DOES NOT EXISTS, TRY TO CREATE IT if [[ "${actionout[$i]}" = tr || "${actionout[$i]}" = tra ]] then if [[ $running_on_remote = true ]] then # SET OPTIONS FOR TRANSFER if [[ "${actionout[$i]}" = tr ]] then if [[ $files_for_cores = false ]] then catalog_option="" catalog_string="" else catalog_option="-c" catalog_string="/" fi append_option="" append_string="" else append_option="-A" append_string="append" fi transfer_failed=false printf "\n >>> OUTPUT: ${localout[$i]}$catalog_string $append_string by SCP to" printf "\n ${pathout[$i]}/${configuration_identifier}_${run_identifier}${endout[$i]}$catalog_string\n" # TRANSFER VIA SCP if [[ "$remote_loginnode" != "" ]] then ssh -q $remote_username@$remote_loginnode "cd $tempdir; ${fast_io_catalog}/${sources_for_run_catalog}/batch_scp $PORTOPT $catalog_option $append_option -b -m $usecycle_option -u $local_username $return_address ${localout[$i]} \"${pathout[$i]}\" ${configuration_identifier}_${run_identifier}${endout[$i]} ${extout[$i]}" else batch_scp $PORTOPT $catalog_option $append_option -b -m $usecycle_option -u $local_username $return_address ${localout[$i]} "${pathout[$i]}" ${configuration_identifier}_${run_identifier}${endout[$i]} ${extout[$i]} fi [[ ${PIPESTATUS[0]} != 0 ]] && transfer_failed=true # IF TRANSFER FAILED, CREATE BACKUP COPY ON THIS MACHINE if [[ $transfer_failed = true ]] then printf " +++ transfer failed. Trying to save a copy on this host under:\n" printf " ${pathout[$i]}/${configuration_identifier}_${run_identifier}${endout[$i]}_$run_id_number\n" # FIRST CHECK, IF DIRECTORY EXISTS, AND CREATE IT, IF NECESSARY eval local_catalog=${pathout[$i]} if [[ ! -d $local_catalog ]] then printf " *** local directory does not exist. Trying to create:\n" printf " $local_catalog \n" mkdir -p $local_catalog fi eval cp ${localout[$i]} ${pathout[$i]}/${configuration_identifier}_${run_identifier}${endout[$i]}_$run_id_number transfer_problems=true fi else # UNSET actionout. DUE TO THIS SETTING, FILE WILL LATER JUST # BE COPIED OR APPENDED ON THIS MACHINE if [[ "${actionout[$i]}" = tr ]] then actionout[$i]="" else actionout[$i]="a" fi fi fi # APPEND ON THIS MACHINE if [[ "${actionout[$i]}" = "a" ]] then if [[ "${extout[$i]}" != " " && "${extout[$i]}" != "" ]] then printf "\n >>> OUTPUT: ${localout[$i]} append to" printf "\n ${pathout[$i]}.${extout[$i]}\n" cat ${localout[$i]} >> ${pathout[$i]}.${extout[$i]} else printf "\n >>> OUTPUT: ${localout[$i]} append to" printf "\n ${pathout[$i]}\n" cat ${localout[$i]} >> ${pathout[$i]} fi fi # COPY ON THIS MACHINE # COPY HAS TO BE USED, BECAUSE MOVE DOES NOT WORK IF FILE-ORIGIN AND TARGET ARE # ON DIFFERENT FILE-SYSTEMS if [[ "${actionout[$i]}" = "" && $files_for_cores = false ]] then # COPY IN CASE OF RUNS ON SINGLE CORES if [[ "${extout[$i]}" != " " && "${extout[$i]}" != "" ]] then printf "\n >>> OUTPUT: ${localout[$i]} to" printf "\n ${pathout[$i]}.${extout[$i]}\n" if [[ $link_local_output = true ]] then printf " file will be linked\n" ln -f ${localout[$i]} ${pathout[$i]}.${extout[$i]} fi # If "ln -f" fails of if "$link_local_output = false" do a normal "cp" if [[ ! -f "${pathout[$i]}.${extout[$i]}" ]] then if [[ $link_local_output = true ]] then printf " --- WARNING: ln failed, using cp instead (might be time consuming...)\n" fi cp ${localout[$i]} ${pathout[$i]}.${extout[$i]} else printf "+++ no copy because file ${pathout[$i]}.${extout[$i]} exists\n" fi else printf "\n >>> OUTPUT: ${localout[$i]} to" printf "\n ${pathout[$i]}\n" if [[ $link_local_output = true ]] then printf " file will be linked\n" ln -f ${localout[$i]} ${pathout[$i]} fi # If "ln -f" fails of if "$link_local_output = false" do a normal "cp" if [[ ! -f "${pathout[$i]}" ]] then if [[ $link_local_output = true ]] then printf " --- WARNING: ln failed, using cp instead (might be time consuming...)\n" fi cp ${localout[$i]} ${pathout[$i]} else printf "+++ no copy because file ${pathout[$i]} exists\n" fi fi elif [[ "${actionout[$i]}" = "" && $files_for_cores = true ]] then # FILES FROM THE DIFFERENT CORES ARE MOVED WITH ln-COMMAND TO THE PERMANENT DIRECTORY # AS A FIRST STEP, THE PERMANENT DIRECTORY IS CREATED printf "\n >>> OUTPUT: ${localout[$i]}/_.... to" printf "\n ${pathout[$i]}\n" if [[ $link_local_output = true ]] then printf " files will be linked\n" mkdir -p ${pathout[$i]} cd ${localout[$i]} for file in $(ls *) do ln -f $file ${pathout[$i]} done >|/dev/null 2>&1 cd $tempdir fi # IF "ln -f" HAS FAILED OR IF "$link_local_output = false" DO A NORMAL COPY "cp -r" if [[ ! -f "${pathout[$i]}/_000000" ]] then if [[ $link_local_output = true ]] then printf " --- WARNING: ln failed, using cp instead (might be time consuming...)\n" fi [[ ! -d "${pathout[$i]}" ]] && mkdir ${pathout[$i]} cp -r ${localout[$i]}/* ${pathout[$i]} fi fi fi done if (( i != 0 )) then timestring=`date +%H:%M:%S` if [[ $transfer_problems = true ]] then printf "\n$dashes\n *** OUTPUT-files saved local time: $timestring" printf "\n +++ WARNING: some data transfers failed! \n" else printf "\n$dashes\n *** all OUTPUT-files saved local time: $timestring \n" fi fi # IF REQUIRED, START A RESTART-JOB # FILE CONTINUE_RUN MUST HAVE BEEN CREATED BY THE EXECUTABLE (PALM) if [[ -f CONTINUE_RUN ]] then # ADD RESTART-OPTIONS TO THE PALMRUN-CALL (IF THEY ARE NOT USED ALREADY): # -C TELLS PALMRUN THAT IT IS A RESTART-RUN # -v SILENT MODE WITHOUT INTERACTIVE QUERIES # -b START A BATCH JOB [[ $(echo $prc | grep -c "\-C") = 0 ]] && prc="$prc -C" [[ $(echo $prc | grep -c "\-v") = 0 ]] && prc="$prc -v" [[ $(echo $prc | grep -c "\-b") = 0 && $running_on_remote = false ]] && prc="$prc -b" # REPLACE SVFOUT IN THE ACTIVATION STRINGS (GIVEN WITH OPTION -a) # SO THAT RESTARTS USE SVF DATA CREATED BY THE INITIAL RUN if [[ $(echo $prc | grep -c "svfout") != 0 ]] then prc=`echo $prc | sed 's/svfout/svfin/g'` fi # REPLACE THE HASH IN THE ACTIVATION STRINGS (GIVEN WITH OPTION -a) # SO THAT RESTARTS ACCESS DIFFERENT FILES THAN THE INITIAL RUN if [[ $(echo $prc | grep -c "#") != 0 ]] then prc=`echo $prc | sed 's/#/r/g'` fi # START THE RESTART-JOB printf "\n\n *** initiating restart-run on \"$local_ip\" using command:\n" echo " $prc" printf "\n$dashes\n" if [[ $running_on_remote = true ]] then echo "*** ssh will be used to initiate restart-runs!" echo " return_address=\"$return_address\" " echo " return_username=\"$local_username\" " if [[ "$remote_loginnode" != "" ]] then ssh -q $remote_username@$remote_loginnode "ssh -q $SSH_PORTOPT $local_username@$return_address \" PATH=\\\$PATH:$LOCAL_PALMRUN_PATH; cd $LOCAL_PWD; $prc\" " | tee palmrun_restart.log else ssh -q $SSH_PORTOPT $local_username@$return_address " PATH=\\\$PATH:$LOCAL_PALMRUN_PATH; cd $LOCAL_PWD; $prc" | tee palmrun_restart.log fi # WAIT TO ALLOW THE RESTART-JOB TO BE QUEUED, BEFORE THE CURRENT JOB IS FINISHED sleep 30 else # START THE RESTART JOB ON THE LOCAL HOST eval $prc | tee palmrun_restart.log # THE ' MUST BE EVALUATED fi # CHECK, IF RESTART JOB HAS BEEN STARTED if [[ $( grep -c "+++ palmrun crashed" palmrun_restart.log ) != 0 ]] then printf "\n$dashes\n +++ creating restart run failed \n" locat=create_restart rm palmrun_restart.log exit elif [[ $( grep -c "*** palmrun finished" palmrun_restart.log ) != 1 ]] then printf "\n$dashes\n +++ creating restart run failed, probably due to network problems\n" locat=create_restart rm palmrun_restart.log exit else printf "\n$dashes\n *** restart run initiated \n" rm palmrun_restart.log fi # DELETE INPUT-(RESTART)FILES, WHICH HAVE BEEN FETCHED FROM THE TEMPORARY DATA # DIRECTORY, BACAUSE THEY ARE NOT REQUIRED BY THE RESTART-JOB. # THIS IS DONE IN ORDER TO AVOID EXCEEDING DISC QUOTAS OR DISC SPACE (RESTART-FILES # MAY BE VERY HUGE) (( i = 0 )) while (( i < nr_of_input_files )) do (( i = i + 1 )) if [[ "${got_tmp[$i]}" = true && $keep_data_from_previous_run = false ]] then rm -r ${absnamein[$i]} fi done fi # ALL ACTIONS FINISHED, TEMPORARY WORKING-DIRECTORY CAN BE DELETED cd $HOME [[ $delete_temporary_catalog = true ]] && rm -rf $tempdir else # PREPARING ACTIONS, # IF A BATCH-JOB IS TO BE GENERATED AND TO BE STARTED ON A LOCAL OR REMOTE-MACHINE # BUILD THE PALMRUN-COMMAND TO BE CALLED IN THE BATCH-JOB palmrun_com="$palmrun_script_name -r $run_identifier -c $configuration_identifier -m $memory -t $cpumax -q $queue -i $run_id_number -U $local_username" [[ "$activation_string_list" != "" ]] && palmrun_com=${palmrun_com}" -a \"$activation_string_list\"" [[ "$global_revision" != "" ]] && palmrun_com=${palmrun_com}" -G \"$global_revision\"" [[ $keep_data_from_previous_run = true ]] && palmrun_com=${palmrun_com}" -k" [[ $do_trace = true ]] && palmrun_com=${palmrun_com}" -x" [[ "$cores" != "" ]] && palmrun_com=${palmrun_com}" -X $cores" [[ $use_openmp = true ]] && palmrun_com=${palmrun_com}" -O $threads_per_task" [[ $tasks_per_node != 0 ]] && palmrun_com=${palmrun_com}" -T $tasks_per_node" [[ $delete_temporary_catalog = false ]] && palmrun_com=${palmrun_com}" -B" [[ "$ocean_file_appendix" = true ]] && palmrun_com=${palmrun_com}" -y" [[ $run_coupled_model = true ]] && palmrun_com=${palmrun_com}" -Y \"$coupled_dist\"" [[ "$combine_plot_fields" = false ]] && palmrun_com=${palmrun_com}" -Z" [[ "$max_par_io_str" != "" ]] && palmrun_com=${palmrun_com}" -w $max_par_io_str" [[ "$project_account" != "" ]] && palmrun_com=${palmrun_com}" -A $project_account" if [[ $create_remote_batch_job = true ]] then palmrun_com=${palmrun_com}" -j -u $remote_username -R $local_ip" if [[ $do_trace = true ]] then printf "\n *** PALMRUN-command on remote host:\n $palmrun_com \n" fi elif [[ $create_batch_job = true ]] then palmrun_com=${palmrun_com}" -j" if [[ $do_trace = true ]] then printf "\n *** PALMRUN-command on local host:\n $palmrun_com \n" fi fi # DETERMINE THE FULL PATHS FOR THE JOB PROTOCOL FILES ON THE LOCAL AND # REMOTE HOST job_protocol_file_local=${local_jobcatalog}/${configuration_identifier}_${run_id} job_protocol_file=$job_protocol_file_local if [[ $create_remote_batch_job = true ]] then job_protocol_file_remote=${remote_jobcatalog}/${configuration_identifier}_${run_id} job_protocol_file=$job_protocol_file_remote job_transfer_protocol_file=${remote_jobcatalog}/last_job_transfer_protocol scpjob_file=${remote_jobcatalog}/scpjob.$run_id_number fi # BUILD THE JOB-SCRIPTS ON FILE jobfile jobfile=jobfile.$run_id_number # FIRST CREATE THE BATCH DIRECTIVES (( i = 0 )) while (( i < ibd )) do (( i = i + 1 )) line=`echo "${batch_directive[$i]}" | sed 's/{{/$/g' | sed 's/}}//g'` eval line=\"$line\" echo "$line" >> $jobfile done echo " " >> $jobfile # FOR BATCH JOBS ON REMOTE HOSTS, ADD THE JOBFILE TO SEND BACK THE JOB # PROTOCOL if [[ $create_remote_batch_job = true ]] then echo "set +vx" >> $jobfile echo "trap '" >> $jobfile echo "set +vx" >> $jobfile echo "cd ${remote_jobcatalog}" >> $jobfile echo "cat > scpjob.$run_id_number << %%END%%" >> $jobfile # ADD THE BATCH DIRECTIVES (( i = 0 )) while (( i < ibdt )) do (( i = i + 1 )) line=`echo "${batch_directive_transfer[$i]}" | sed 's/{{/$/g' | sed 's/}}//g'` eval line=\"$line\" echo "$line" >> $jobfile done echo " " >> $jobfile echo "set -x" >> $jobfile if [[ "$remote_loginnode" != "" ]] then echo "ssh -q $remote_username@$remote_loginnode \"cd ${remote_jobcatalog}; ${fast_io_catalog}/${sources_for_run_catalog}/batch_scp $PORTOPT -d -w 10 -u $local_username $local_ip $job_protocol_file_remote \\\"$local_jobcatalog\\\" ${configuration_identifier}_${run_identifier}\" " >> $jobfile else echo "${fast_io_catalog}/${sources_for_run_catalog}/batch_scp $PORTOPT -d -w 10 -u $local_username $local_ip $job_protocol_file_remote \"$local_jobcatalog\" ${configuration_identifier}_${run_identifier}" >> $jobfile fi echo "%%END%%" >> $jobfile echo "echo \" *** submitting job for transfering the job protocol file to $local_ip\" " >> $jobfile echo "$submit_command $scpjob_file" >> $jobfile echo "rm $scpjob_file" >> $jobfile echo "rm -rf $job_transfer_protocol_file" >> $jobfile echo "set -x" >> $jobfile echo " ' exit" >> $jobfile fi # ACTIVATE ERROR-TRACEBACK if [[ $do_trace = true ]] then echo "set -x" >> $jobfile else echo "set +vx" >> $jobfile fi # INITIALIZE THE ENVIRONMENT AND LOAD MODULES if [[ "$login_init_cmd" != "" ]] then echo "$login_init_cmd" >> $jobfile fi if [[ "$module_commands" != "" ]] then echo "$module_commands" >> $jobfile fi # CREATE TEMPORARY DIRECTORY AND SWITCH TO IT if [[ $create_remote_batch_job = true ]] then echo "mkdir $tempdir" >> $jobfile echo "if [[ \$? != 0 ]]" >> $jobfile echo "then" >> $jobfile echo " echo \"+++ temporary working directory cannot be created.\" " >> $jobfile echo " echo \" Check setting of variable fast_io_catalog in your config file.\" " >> $jobfile echo " exit" >> $jobfile echo "fi" >> $jobfile echo "chmod go+rx $tempdir" >> $jobfile else # DIRECTORY FOR LOCAL BATCH JOBS IS CREATED NOW, DUE TO A # REQUIREMENT OF THE GRID ENGINE BATCH SYSTEM (WORKING DIR IS GIVEN IN # BATCH DIRECTIVE -wd AND MUST ALREADY EXIST WHEN THE JOB IS SUBMITTED) mkdir $tempdir if [[ $? != 0 ]] then printf "\n +++ temporary working directory cannot be created." printf "\n Check setting of variable fast_io_catalog in your config file." locat=execution exit fi chmod go+rx $tempdir fi echo "cd $tempdir" >> $jobfile echo "export tempdir=$tempdir" >> $jobfile echo "cp ${fast_io_catalog}/${sources_for_run_catalog}/.[!.]* ." >> $jobfile echo "export PATH=.:${fast_io_catalog}/${sources_for_run_catalog}:\$PATH" >> $jobfile echo "export execute_palmrun=true" >> $jobfile # PROVIDE NAME OF THE CURRENT WORKING-DIRECTORY ON THE LOCAL MACHINE (FROM WHERE THE JOB IS # STARTED) BY SETTING AN ENVIRONMENT-VARIABLE. THIS INFORMATION IS USED IN THE JOB BY PALMRUN # IN CASE THAT RESTART-RUNS HAVE TO BE GENERATED echo "LOCAL_PWD=$working_directory" >> $jobfile echo "export LOCAL_PWD" >> $jobfile # PROVIDE THE PATH OF THE LOCAL PALMRUN-SCRIPT FOR THE SAME REASON echo "LOCAL_PALMRUN_PATH=${source_path}/../SCRIPTS" >> $jobfile echo "export LOCAL_PALMRUN_PATH" >> $jobfile # CALL PALMRUN WITHIN THE JOB # AS FINAL ACTION, REMOVE THE TEMPORARY DIRECTORY CREATED AT THE BEGINNING OF THE JOB echo "set -x" >> $jobfile echo "[[ \$execute_palmrun = true ]] && $palmrun_com" >> $jobfile # TRANSFER JOBFILE TO THE TARGET HOST if [[ $create_jobfile_only = false ]] then # CREATE LOCAL JOB FOLDER eval local_jobcatalog=$local_jobcatalog if [[ ! -d $local_jobcatalog ]] then echo " " echo " *** local jobcatalog \"$local_jobcatalog\" does not exist and will be created now" mkdir $local_jobcatalog fi if [[ $create_remote_batch_job = true ]] then echo " " echo " *** transfer of job to remote host via scp" # CREATE REMOTE JOB FOLDER, IF IT DOES NOT EXIST ssh -q $ssh_key $SSH_PORTOPT ${remote_username}@${remote_ip} "[[ ! -d $remote_jobcatalog ]] && mkdir $remote_jobcatalog" 2>&1 # COPY THE JOB FILE if [[ $do_trace = true ]] then echo " scp $ssh_key $PORTOPT $jobfile ${remote_username}@${remote_ip}:${remote_jobcatalog}/${configuration_identifier}_${run_id}" fi scp $ssh_key $PORTOPT $jobfile ${remote_username}@${remote_ip}:${remote_jobcatalog}/${configuration_identifier}_${run_id} > /dev/null # SUBMIT THE JOB printf " *** submit the job (output of submit command, \n" printf " e.g. the job-id given by the batch-system, may follow) \n" if [[ $do_trace = true ]] then echo " ssh -q $ssh_key $SSH_PORTOPT ${remote_username}@${remote_ip} \"cd $remote_jobcatalog; $submit_command ${configuration_identifier}_${run_id}; rm ${configuration_identifier}_${run_id}\" " fi ssh -q $ssh_key $SSH_PORTOPT ${remote_username}@${remote_ip} "cd $remote_jobcatalog; $submit_command ${configuration_identifier}_${run_id}; rm ${configuration_identifier}_${run_id}" elif [[ $create_batch_job = true ]] then eval local_jobcatalog=$local_jobcatalog cp $jobfile ${local_jobcatalog}/${configuration_identifier}_${run_id} cd $local_jobcatalog echo " " echo " *** submit the job" if [[ $do_trace = true ]] then echo "$submit_command ${configuration_identifier}_${run_id}" fi $submit_command ${configuration_identifier}_${run_id} rm ${configuration_identifier}_${run_id} cd - > /dev/null fi rm -rf $jobfile else printf "\n *** jobfile created under name \"$jobfile\" " printf "\n no batch-job has been sent!" fi fi # END OF REMOTE-PART