Defines | |
| #define | DRMAA_GW_TOTAL_TASKS "${TOTAL_TASKS}" |
| #define | DRMAA_GW_JOB_ID "${JOB_ID}" |
| #define | DRMAA_GW_TASK_ID "${TASK_ID}" |
| #define | DRMAA_GW_PARAM "${PARAM}" |
| #define | DRMAA_GW_MAX_PARAM "${MAX_PARAM}" |
| #define | DRMAA_GW_ARCH "${ARCH}" |
| #define | DRMAA_V_GW_INPUT_FILES "INPUT_FILES" |
| #define | DRMAA_V_GW_OUTPUT_FILES "OUTPUT_FILES" |
| #define | DRMAA_V_GW_RESTART_FILES "RESTART_FILES" |
| #define | DRMAA_GW_RESCHEDULE_ON_FAILURE "RESCHEDULE_ON_FAILURE" |
| #define | DRMAA_GW_NUMBER_OF_RETRIES "NUMBER_OF_RETRIES" |
| #define | DRMAA_GW_RANK "RANK" |
| #define | DRMAA_GW_REQUIREMENTS "REQUIREMENTS" |
| #define | DRMAA_GW_TYPE "TYPE" |
| #define | DRMAA_GW_TYPE_SINGLE "single" |
| #define | DRMAA_GW_TYPE_MPI "mpi" |
| #define | DRMAA_GW_NP "NP" |
| #define DRMAA_GW_ARCH "${ARCH}" |
Pre-defined string to refer to the remote host architecture as returned by the resource selector module. DRMAA_GW_ARCH will be available during job execution and can be used to generate architecture-dependent REMOTE COMMAND executables.
| #define DRMAA_GW_JOB_ID "${JOB_ID}" |
Pre-defined string to refer to the job unique identification as provided by the GridWay system. DRMAA_GW_JOB_ID will be available during job execution and can be used as an ARGUMENT for the REMOTE COMMAND. It is also usefull to generate output filenames, since it is available in the main DRMAA program as returned by drmaa_run_bulk_jobs() and drmaa_run_job() function calls.
| #define DRMAA_GW_MAX_PARAM "${MAX_PARAM}" |
Pre-defined string to refer to the max value of the custom parameter in bulk jobs. This value is equal to <start> + <total_tasks> * <increment>.
| #define DRMAA_GW_NP "NP" |
Pre-defined string to refer to the number of process requested by a MPI job
| #define DRMAA_GW_NUMBER_OF_RETRIES "NUMBER_OF_RETRIES" |
Pre-defined string to refer to the NUMBER_OF_RETRIES GridWay scheduler parameter, the number of times to retry the execution on a given host. Default value is 3.
| #define DRMAA_GW_PARAM "${PARAM}" |
Pre-defined string to refer to a custom parameter in bulk jobs. This value is equal to <start> + <task_id> * <increment>, where <start> and <increment> are drmaa_run_bulk_job() arguments. DRMAA_PLACEHOLDER_INCR should be used for portability reasons
| #define DRMAA_GW_RANK "RANK" |
Pre-defined string to refer to the RANK job template parameter. The RANK is a mathematical expression evaluated for each candidate host (those for which the REQUIREMENTS expression is true). Those candidates with higher ranks are used first to execute your jobs. Example: "(CPU_MHZ * 2) + FREE_MEM_MB;" (NOTE: Must end with ';')
| #define DRMAA_GW_REQUIREMENTS "REQUIREMENTS" |
Pre-defined string to refer to the REQUIREMENTS job template parameter. The REQUIREMENTS is a boolean expression evaluated for each host in the Grid, if it is true the host will be considered to submit the job. Example:"ARCH = "i686" & CPU_MHZ > 1000;" (NOTE: Must end with ';')
| #define DRMAA_GW_RESCHEDULE_ON_FAILURE "RESCHEDULE_ON_FAILURE" |
Pre-defined string to refer to the ON_FAILURE GridWay scheduler parameter. If set to "yes" GridWay will reschedule the job after retrying execution on a given host DRMAA_GW_NUMBER_OF_RETRIES times. Values are "yes" or "no". Default value for this attribute is "no".
| #define DRMAA_GW_TASK_ID "${TASK_ID}" |
Pre-defined string to refer to the task unique identification as provided by the GridWay system. DRMAA_GW_TASK_ID will be available during job execution and can be used as an ARGUMENT for the REMOTE COMMAND. It is also usefull to generate output filenames of bluk jobs. DRMAA_GW_TASK_ID ALWAYS ranges from 0 to DRMAA_GW_TOTAL_TASKS -1. Please note that this attribute name should be used ONLY in conjuntion with a drmaa_run_bulk_jobs() function call.
| #define DRMAA_GW_TOTAL_TASKS "${TOTAL_TASKS}" |
Pre-defined string to refer to the number of total tasks in a bulk job. DRMAA_GW_TOTAL_TASKS will be available during job execution and can be used as an ARGUMENT for the REMOTE COMMAND. This attribute name should be used ONLY in conjuntion with a drmaa_run_bulk_jobs() function call.
| #define DRMAA_GW_TYPE "TYPE" |
Pre-defined string to refer to the job type: "single" or "mpi". Jobs of both types can be further combined in array or workflow structures. MPI jobs spawn within a single resource and NOT across multiple resources.
| #define DRMAA_GW_TYPE_MPI "mpi" |
Pre-defined string to define MPI (Message Passing Interface) jobs.
| #define DRMAA_GW_TYPE_SINGLE "single" |
Pre-defined string to define single (one process) jobs.
| #define DRMAA_V_GW_INPUT_FILES "INPUT_FILES" |
Pre-defined string to refer to the input files of DRMAA_REMOTE_COMMAND. DRMAA_V_GW_INPUT_FILES corresponds to a NULL terminated vector attribute value. Each vector entry is a pair of the form "source destination" filenames. If the destination filename is missing, the source filename will be preserved in the execution host. Input files (sources) ARE RELATIVE TO THE WORKING DIRECTORY or can be a GSIFTP URL. Example: input_file[0]="param."DRMAA_GW_TASK_ID" param" will copy the local file param.2 (for task 2) as param in the remote working directory.
| #define DRMAA_V_GW_OUTPUT_FILES "OUTPUT_FILES" |
Pre-defined string to refer to the input files of DRMAA_REMOTE_COMMAND. DRMAA_V_GW_OUPUT_FILES corresponds to a NULL terminated vector attribute value. Each vector entry is a "source destination" filenames pair. If the destination filename is missing, the source filename will be preserved in the client host. Output files can be a GSIFTP URL. Example: output_file[0]="binary binary."DRMAA_GW_ARCH will copy the output file "binary" to the client host with name binary.i686 (architecture of remote host is i686)
| #define DRMAA_V_GW_RESTART_FILES "RESTART_FILES" |
Pre-defined string to refer to the re-start files generated by DRMAA_REMOTE_COMMAND. DRMAA_V_GW_RESTART_FILES corresponds to a NULL terminated vector attribute value. Each vector entry is the name of a checkpointing file. Re-start files can be used to preserve the execution context (at the application level) of the DRMAA_REMOTE_COMMAND on job migration or stop/resume actions.