void
fsd_job_destroy( fsd_job_t *self )
{
fsd_log_enter(( "(%p={job_id=%s})", (void*)self, self->job_id ));
fsd_cond_destroy( &self->status_cond );
fsd_cond_destroy( &self->destroy_cond );
fsd_mutex_destroy( &self->mutex );
fsd_free( self->job_id );
fsd_free( self );
fsd_log_return(( "" ));
}
static void
fsd_iter_destroy( fsd_iter_t *self )
{
int i;
if( self->_own_list && self->_list )
{
for( i = 0; i < self->_length; i++ )
fsd_free( self->_list[i] );
fsd_free( self->_list );
}
fsd_free( self );
}
void
fsd_job_destroy( fsd_job_t *self )
{
fsd_log_enter(( "(%p={job_id=%s})", (void*)self, self->job_id ));
fsd_cond_destroy( &self->status_cond );
fsd_cond_destroy( &self->destroy_cond );
fsd_mutex_destroy( &self->mutex );
fsd_free( self->job_id );
fsd_free( self->execution_hosts );
fsd_free( self->queue );
fsd_free( self->project );
fsd_free( self );
fsd_log_return(( "" ));
}
static void
fsd_template_set_attr( fsd_template_t *self,
const char *name, const char *value )
{
const fsd_attribute_t *attr = NULL;
fsd_log_enter(("(%s=%s)", name, value));
if( name == NULL )
fsd_exc_raise_code( FSD_ERRNO_INVALID_ARGUMENT );
attr = self->by_name( self, name );
if( attr == NULL || attr->is_vector )
fsd_exc_raise_fmt(
FSD_ERRNO_INVALID_ARGUMENT,
"invalid scalar attribute name: %s", name
);
if( value != NULL ) {
if (strlen (value) > DRMAA_MAX_ATTR_LEN)
fsd_exc_raise_fmt(
FSD_ERRNO_INVALID_ARGUMENT,
"Argument length exceeds max size: %d > %d", (int)strlen(value), DRMAA_MAX_ATTR_LEN
);
if( self->attributes[ attr->code ] != NULL ) {
fsd_free(self->attributes[ attr->code ]);
}
self->attributes[ attr->code ] = fsd_strdup( value );
}
else
self->attributes[ attr->code ] = NULL;
}
static fsd_iter_t *
fsd_iter_new_impl( char **list, int length, bool own )
{
fsd_iter_t *volatile self = NULL;
TRY
{
fsd_malloc( self, fsd_iter_t );
self->next = fsd_iter_next;
self->reset = fsd_iter_reset;
self->len = fsd_iter_len;
self->append = fsd_iter_append;
self->destroy = fsd_iter_destroy;
self->_list = list;
self->_position = 0;
if( list == NULL )
self->_length = 0;
else if( length >= 0 )
self->_length = length;
else
{
char **i;
int cnt = 0;
for( i = self->_list; *i != NULL; i++ )
cnt++;
self->_length = cnt;
}
self->_own_list = own;
}
EXCEPT_DEFAULT
{
if( own && list )
{
if( length >= 0 )
{
int i;
for( i = 0; i < length; i++ )
fsd_free( list[i] );
fsd_free( list );
}
else
fsd_free_vector( list );
}
fsd_exc_reraise();
}
END_TRY
return self;
}
void
fsd_drmaa_session_destroy_nowait( fsd_drmaa_session_t *self )
{
fsd_log_enter(( "" ));
fsd_conf_dict_destroy( self->configuration );
fsd_free( self->contact );
if( self->jobs )
self->jobs->destroy( self->jobs );
fsd_mutex_destroy( &self->mutex );
fsd_cond_destroy( &self->wait_condition );
fsd_cond_destroy( &self->destroy_condition );
fsd_mutex_destroy( &self->drm_connection_mutex );
fsd_free( self );
fsd_log_return(( "" ));
}
void
fsd_job_set_destroy( fsd_job_set_t *self )
{
unsigned i;
fsd_job_t *j;
fsd_log_enter(( "()" ));
for( i = 0; i < self->tab_size; i++ )
for( j = self->tab[i]; j != NULL; )
{
fsd_job_t *job = j;
j = j->next;
fsd_mutex_lock( &job->mutex );
job->release( job );
}
fsd_free( self->tab );
fsd_free( self );
fsd_log_return(( "" ));
}
static void
fsd_template_destroy( fsd_template_t *self )
{
unsigned i;
for( i = 0; i < self->n_attributes; i++ )
if( self->attributes[i] != NULL )
{
const fsd_attribute_t *attr;
attr = self->by_code( self, i );
if( attr )
{
if( attr->is_vector )
fsd_free_vector( self->attributes[i] );
else
fsd_free( self->attributes[i] );
}
}
fsd_free( self->attributes );
fsd_free( self );
}
static char *
lsfdrmaa_job_quote_command( const char *const *argv )
{
char *volatile result = NULL;
TRY
{
size_t size = 0;
const char *const *i;
const char *j;
char *s;
for( i = argv; *i; i++ )
{
if( i != argv )
size++;
size += 2;
for( j = *i; *j; j++ )
switch( *j )
{
case '"': case '$': case '\\': case '`':
size ++;
default:
size ++;
break;
}
}
fsd_calloc( result, size+1, char );
s = result;
for( i = argv; *i; i++ )
{
if( i != argv )
*s++ = ' ';
*s++ = '"';
for( j = *i; *j; j++ )
switch( *j )
{
case '"': case '$': case '\\': case '`':
*s++ = '\\';
default:
*s++ = *j;
break;
}
*s++ = '"';
}
*s++ = '\0';
}
EXCEPT_DEFAULT
{
fsd_free( result );
fsd_exc_reraise();
}
END_TRY
return result;
}
fsd_job_set_t *
fsd_job_set_new(void)
{
fsd_job_set_t *volatile self = NULL;
const size_t initial_size = 1024;
fsd_log_enter(( "()" ));
TRY
{
fsd_malloc( self, fsd_job_set_t );
self->destroy = fsd_job_set_destroy;
self->add = fsd_job_set_add;
self->remove = fsd_job_set_remove;
self->get = fsd_job_set_get;
self->empty = fsd_job_set_empty;
self->find_terminated = fsd_job_set_find_terminated;
self->get_all_job_ids = fsd_job_set_get_all_job_ids;
self->signal_all = fsd_job_set_signal_all;
self->tab = NULL;
self->n_jobs = 0;
fsd_calloc( self->tab, initial_size, fsd_job_t* );
self->tab_size = initial_size;
self->tab_mask = self->tab_size - 1;
fsd_mutex_init( &self->mutex );
}
EXCEPT_DEFAULT
{
if( self )
{
fsd_free( self->tab );
fsd_free( self );
}
fsd_exc_reraise();
}
END_TRY
fsd_log_return(( " =%p", (void*)self ));
return self;
}
fsd_job_t *
fsd_job_new( char *job_id )
{
fsd_job_t *volatile self = NULL;
fsd_log_enter(( "(%s)", job_id ));
TRY
{
fsd_malloc( self, fsd_job_t );
self->release = fsd_job_release;
self->destroy = fsd_job_destroy;
self->control = fsd_job_control;
self->update_status = fsd_job_update_status;
self->get_termination_status = fsd_job_get_termination_status;
self->on_missing = fsd_job_on_missing;
self->next = NULL;
self->ref_cnt = 1;
self->job_id = job_id;
self->session = NULL;
self->last_update_time = 0;
self->flags = 0;
self->state = DRMAA_PS_UNDETERMINED;
self->exit_status = 0;
self->submit_time = 0;
self->start_time = 0;
self->end_time = 0;
self->cpu_usage = 0;
self->mem_usage = 0;
self->vmem_usage = 0;
self->walltime = 0;
self->n_execution_hosts = 0;
self->execution_hosts = NULL;
self->retry_cnt = 0;
fsd_mutex_init( &self->mutex );
fsd_cond_init( &self->status_cond );
fsd_cond_init( &self->destroy_cond );
fsd_mutex_lock( &self->mutex );
}
EXCEPT_DEFAULT
{
if( self )
self->destroy( self );
else
fsd_free( job_id );
fsd_exc_reraise();
}
END_TRY
fsd_log_return(( "=%p: ref_cnt=%d [lock %s]",
(void*)self, self->ref_cnt, self->job_id ));
return self;
}
fsd_drmaa_session_t *
slurmdrmaa_session_new( const char *contact )
{
slurmdrmaa_session_t *volatile self = NULL;
TRY
{
self = (slurmdrmaa_session_t*)fsd_drmaa_session_new(contact);
fsd_realloc( self, 1, slurmdrmaa_session_t );
self->super.run_job = slurmdrmaa_session_run_job;
self->super.run_bulk = slurmdrmaa_session_run_bulk;
self->super.new_job = slurmdrmaa_session_new_job;
self->super.load_configuration( &self->super, "slurm_drmaa" );
}
EXCEPT_DEFAULT
{
fsd_free( self );
fsd_exc_reraise();
}
END_TRY
return (fsd_drmaa_session_t*)self;
}
void
lsfdrmaa_job_set_req(
fsd_drmaa_session_t *session,
fsd_expand_drmaa_ph_t *expand,
const fsd_template_t *jt,
struct submit *req,
fsd_environ_t **envp
)
{
const char *input_path_orig = NULL;
const char *output_path_orig = NULL;
const char *error_path_orig = NULL;
char *volatile input_path = NULL;
char *volatile output_path = NULL;
char *volatile error_path = NULL;
bool input_host = false;
bool output_host = false;
bool error_host = false;
bool join_files = false;
bool transfer_input = false;
bool transfer_output = false;
bool transfer_error = false;
const char *job_category = "default";
char **volatile argv = NULL;
const char *value;
const char *const *vector;
/* set default lsf configs */
{
int i = 0;
req->options = 0;
req->options2 = 0;
for( i = 0; i < LSF_RLIM_NLIMITS; i++ )
req->rLimits[i] = DEFAULT_RLIMIT;
req->beginTime = 0;
req->termTime = 0;
}
/* job category */
value = jt->get_attr( jt, DRMAA_JOB_CATEGORY );
if( value )
job_category = value;
{
fsd_conf_option_t *category_value = NULL;
category_value = fsd_conf_dict_get( session->job_categories, job_category );
if( category_value != NULL )
{
if( category_value->type != FSD_CONF_STRING )
fsd_exc_raise_fmt(
FSD_ERRNO_INTERNAL_ERROR,
"configuration error: job category should be string"
);
lsfdrmaa_native_parse( category_value->val.string, req );
}
else
{
if( value != NULL )
fsd_exc_raise_fmt(
FSD_DRMAA_ERRNO_INVALID_ATTRIBUTE_VALUE,
"invalid job category: %s", job_category
);
}
}
/* job working directory */
value = jt->get_attr( jt, DRMAA_WD );
if( value )
{
char *cwd = NULL;
cwd = expand->expand( expand, fsd_strdup(value),
FSD_DRMAA_PH_HD | FSD_DRMAA_PH_INCR );
expand->set( expand, FSD_DRMAA_PH_WD, cwd );
#ifdef SUB3_CWD
req->cwd = fsd_strdup( cwd );
req->options3 |= SUB3_CWD;
#else
fsd_exc_raise_fmt(FSD_ERRNO_INTERNAL_ERROR, "DRMAA_WD attribute is not supported in this version of LSF.");
#endif
}
TRY
{
const char *command = NULL;
unsigned n_args = 0;
const char *const *i;
int j;
/* remote command */
command = jt->get_attr( jt, DRMAA_REMOTE_COMMAND );
if( command == NULL )
fsd_exc_raise_msg(
FSD_DRMAA_ERRNO_CONFLICTING_ATTRIBUTE_VALUES,
"drmaa_remote_command not set for job template"
);
/* arguments list */
vector = jt->get_v_attr( jt, DRMAA_V_ARGV );
if( vector )
{
for( i = vector; *i; i++ )
n_args++;
}
fsd_calloc( argv, n_args+3, char* );
argv[0] = fsd_strdup("exec");
argv[1] = expand->expand( expand, fsd_strdup(command),
FSD_DRMAA_PH_HD | FSD_DRMAA_PH_WD );
if( vector )
{
for( i = vector, j = 2; *i; i++, j++ )
argv[j] = expand->expand( expand, fsd_strdup(*i),
FSD_DRMAA_PH_HD | FSD_DRMAA_PH_WD );
}
req->command = lsfdrmaa_job_quote_command( (const char*const*)argv );
}
FINALLY
{
fsd_free_vector( argv );
}
END_TRY
/* job name */
value = jt->get_attr( jt, DRMAA_JOB_NAME );
if( value )
{
req->jobName = fsd_strdup(value);
req->options |= SUB_JOB_NAME;
}
/* job state at submit */
value = jt->get_attr( jt, DRMAA_JS_STATE );
if( value )
{
if( 0 == strcmp( value, DRMAA_SUBMISSION_STATE_ACTIVE ) )
req->options2 &= !SUB2_HOLD;
else if( 0 == strcmp( value, DRMAA_SUBMISSION_STATE_HOLD ) )
req->options2 |= SUB2_HOLD;
else
fsd_exc_raise_msg(
FSD_DRMAA_ERRNO_INVALID_ATTRIBUTE_VALUE,
"invalid value of drmaa_js_state attribute" );
}
/* environment */
vector = jt->get_v_attr( jt, DRMAA_V_ENV );
if( vector )
{
fsd_environ_t *env;
*envp = env = fsd_environ_new( NULL );
env->update( env, vector );
}
/* start time */
value = jt->get_attr( jt, DRMAA_START_TIME );
if( value )
{
req->beginTime = fsd_datetime_parse( value );
fsd_log_debug(( "\n drmaa_start_time: %s -> %ld",
value, (long)req->beginTime ));
}
TRY
{
/* input path */
input_path_orig = jt->get_attr( jt, DRMAA_INPUT_PATH );
if( input_path_orig )
{
input_path = internal_map_file( expand, input_path_orig, &input_host,
"input" );
fsd_log_debug(( "\n drmaa_input_path: %s -> %s",
input_path_orig, input_path ));
}
/* output path */
output_path_orig = jt->get_attr( jt, DRMAA_OUTPUT_PATH );
if( output_path_orig )
{
output_path = internal_map_file( expand, output_path_orig, &output_host,
"output" );
fsd_log_debug(( "\n drmaa_output_path: %s -> %s",
output_path_orig, output_path ));
}
/* error path */
error_path_orig = jt->get_attr( jt, DRMAA_ERROR_PATH );
if( error_path_orig )
{
error_path = internal_map_file( expand, error_path_orig, &error_host,
"error" );
fsd_log_debug(( "\n drmaa_error_path: %s -> %s",
error_path_orig, error_path ));
}
/* join files */
value = jt->get_attr( jt, DRMAA_JOIN_FILES );
if( value )
{
if( (value[0] == 'y' || value[0] == 'Y') && value[1] == '\0' )
join_files = true;
else if( (value[0] == 'n' || value[0] == 'N') && value[1] == '\0' )
join_files = false;
else
fsd_exc_raise_msg(
FSD_DRMAA_ERRNO_INVALID_ATTRIBUTE_VALUE,
"invalid value of drmaa_join_files attribute" );
}
if( join_files )
{
/*
* LSF by default joins output and error streams
* when error file is not set.
*/
if( error_path )
{
if( output_path == NULL )
fsd_exc_raise_msg(
FSD_DRMAA_ERRNO_CONFLICTING_ATTRIBUTE_VALUES,
"drmaa_join_files is set and output file is not given" );
if( 0 != strcmp( output_path, error_path ) )
fsd_log_warning(( "Error file was given but will be ignored "
"since drmaa_join_files was set." ));
fsd_free( error_path ); error_path = NULL;
}
}
else
{
/*
* If error path is not set, we must set it to /dev/null
* to prevent joining files.
*/
if( error_path == NULL && output_path )
error_path = fsd_strdup( "/dev/null" );
if( output_path == NULL && error_path )
output_path = fsd_strdup( "/dev/null" );
if( req->errFile == NULL )
{
req->errFile = fsd_strdup( "/dev/null" );
req->options |= SUB_ERR_FILE;
#ifdef SUB2_OVERWRITE_ERR_FILE
req->options2 &= ~SUB2_OVERWRITE_ERR_FILE;
#endif
}
}
/* transfer files */
value = jt->get_attr( jt, DRMAA_TRANSFER_FILES );
if( value )
{
const char *i;
for( i = value; *i; i++ )
{
switch( *i )
{
case 'i': transfer_input = true; break;
case 'o': transfer_output = true; break;
case 'e': transfer_error = true; break;
default:
fsd_exc_raise_fmt(
FSD_DRMAA_ERRNO_INVALID_ATTRIBUTE_VALUE,
"invalid character '%c' in drmaa_transfer_files: %s",
*i, value
);
}
}
}
# if 0
{
/*
* Input file is transfered by LSF from submission host whenever
* it isn't found on execution host regardless of explicit file transfers.
* When drmaa_transfer_files contains ``i`` input file is send
* explicitly because it may be outdated or otherwise differ.
*/
static const char *name[3]
= {"input", "output", "error"};
static const int options[3]
= { XF_OP_SUB2EXEC, XF_OP_EXEC2SUB, XF_OP_EXEC2SUB };
const char *path[3];
bool host[3], transfer[3];
int i;
path[i=0] = input_path;
path[++i] = output_path;
path[++i] = error_path;
host[i=0] = input_host;
host[++i] = output_host;
host[++i] = error_host;
transfer[i=0] = transfer_input;
transfer[++i] = transfer_output;
transfer[++i] = transfer_error;
for( i = 0; i < 3; i++ )
{
struct xFile *t;
if( !(transfer[i] && path[i] != NULL) )
continue;
if( 0 == strcmp( path[i], "/dev/null" ) )
continue;
if( host[i] )
fsd_log_warning((
"hostname in drmaa_%s_path ignored", name[i] ));
fsd_log_debug(( "setting transfer of %s file (%s) "
"to execution host", name[i], path[i] ));
fsd_realloc( req->xf, req->nxf+1, struct xFile );
t = &req->xf[ req->nxf++ ];
memset( t, 0, sizeof(struct xFile) );
if( sizeof(t->subFn) == MAXFILENAMELEN )
{ /* LSF 6 */
strlcpy( t->subFn, path[i], MAXFILENAMELEN );
strlcpy( t->execFn, path[i], MAXFILENAMELEN );
}
else
{ /* LSF 7 */
*(char**)&t->subFn = fsd_strdup( path[i] );
*(char**)&t->execFn = fsd_strdup( path[i] );
}
t->options = options[i];
}
if( req->nxf > 0 )
req->options |= SUB_OTHER_FILES;
}
# endif /* transfer files */
/* email addresses to send notifications */
vector = jt->get_v_attr( jt, DRMAA_V_EMAIL );
if( vector && vector[0] )
{
/* only to one email address message may be send */
req->mailUser = fsd_strdup( vector[0] );
req->options |= SUB_MAIL_USER | SUB_NOTIFY_END;
#if 0
if( vector[1] != NULL )
fsd_log_warning(( "LSF only supports one e-mail "
"notification address" ));
#endif
}
/* block email */
value = jt->get_attr( jt, DRMAA_BLOCK_EMAIL );
if( value )
{
bool block;
if( strcmp(value, "1") == 0 )
block = true;
else if( strcmp(value, "0") == 0 )
block = false;
else
fsd_exc_raise_msg(
FSD_DRMAA_ERRNO_INVALID_ATTRIBUTE_VALUE,
"invalid value of drmaa_block_email attribute" );
if( block )
{
if( output_path == NULL )
{
fsd_log_debug(( "output path not set and we want to block e-mail, "
"set to /dev/null" ));
output_path = fsd_strdup( "/dev/null" );
}
req->options &= ~SUB_NOTIFY_END;
}
else
{
/* SUB_NOTIFY_END should force sending e-mail even if outfile is set */
req->options |= SUB_NOTIFY_END;
}
}
if( !((lsfdrmaa_session_t*)session)->prepand_report_to_output
&& (req->options & SUB_NOTIFY_END) == 0
&& output_path != NULL )
{
req->options |= SUB_MAIL_USER | SUB_NOTIFY_END;
fsd_free( req->mailUser ); /* when email was set
but notification was blocked */
req->mailUser = fsd_strdup( "notexistent" );
}
if( input_path )
{
req->inFile = input_path;
req->options |= SUB_IN_FILE;
input_path = NULL;
}
if( output_path )
{
req->outFile = output_path;
req->options |= SUB_OUT_FILE;
#ifdef SUB2_OVERWRITE_OUT_FILE
if( 0 != strcmp( output_path, "/dev/null" ) )
req->options2 |= SUB2_OVERWRITE_OUT_FILE;
#endif
output_path = NULL;
}
if( error_path )
{
req->errFile = error_path;
req->options |= SUB_ERR_FILE;
#ifdef SUB2_OVERWRITE_ERR_FILE
if( 0 != strcmp( error_path, "/dev/null" ) )
req->options2 |= SUB2_OVERWRITE_ERR_FILE;
#endif
error_path = NULL;
}
}
FINALLY
{
fsd_free( input_path );
fsd_free( output_path );
fsd_free( error_path );
}
END_TRY
/* deadline time */
value = jt->get_attr( jt, DRMAA_DEADLINE_TIME );
if( value )
req->termTime = fsd_datetime_parse( value );
/* wall clock time hard limit */
value = jt->get_attr( jt, DRMAA_WCT_HLIMIT );
if( value )
req->rLimits[ LSF_RLIMIT_RUN ] = fsd_parse_timedelta( value );
/* wall clock time soft limit */
#ifdef SUB3_RUNTIME_ESTIMATION
value = jt->get_attr( jt, DRMAA_WCT_SLIMIT );
if( value )
{
req->options3 = SUB3_RUNTIME_ESTIMATION;
req->runtimeEstimation = fsd_parse_timedelta( value );
}
#endif
/* duration hard limit */
value = jt->get_attr( jt, DRMAA_DURATION_HLIMIT );
if( value )
req->rLimits[ LSF_RLIMIT_CPU ] = fsd_parse_timedelta( value );
/* native specification */
value = jt->get_attr( jt, DRMAA_NATIVE_SPECIFICATION );
if( value )
lsfdrmaa_native_parse( value, req );
lsfdrmaa_dump_submit_req(req);
}
void
slurmdrmaa_job_create(
fsd_drmaa_session_t *session,
const fsd_template_t *jt,
fsd_environ_t **envp,
fsd_expand_drmaa_ph_t *expand,
job_desc_msg_t * job_desc,
int n_job
)
{
const char *input_path_orig = NULL;
const char *output_path_orig = NULL;
const char *error_path_orig = NULL;
char *volatile input_path = NULL;
char *volatile output_path = NULL;
char *volatile error_path = NULL;
bool input_host = false;
bool output_host = false;
bool error_host = false;
bool join_files = false;
const char *value;
const char *const *vector;
const char *job_category = "default";
job_desc->user_id = getuid();
job_desc->group_id = getgid();
job_desc->env_size = 0;
/* job name */
value = jt->get_attr( jt, DRMAA_JOB_NAME );
if( value )
{
job_desc->name = fsd_strdup(value);
fsd_log_debug(("# job_name = %s",job_desc->name));
}
/* job state at submit */
value = jt->get_attr( jt, DRMAA_JS_STATE );
if( value )
{
if( 0 == strcmp( value, DRMAA_SUBMISSION_STATE_ACTIVE ) )
{}
else if( 0 == strcmp( value, DRMAA_SUBMISSION_STATE_HOLD ) )
{
job_desc->priority = 0;
fsd_log_debug(("# hold = user"));
}
else
{
fsd_exc_raise_msg(FSD_DRMAA_ERRNO_INVALID_ATTRIBUTE_VALUE, "invalid value of drmaa_js_state attribute" );
}
}
TRY
{
const char *command = NULL;
char *command_expanded = NULL;
char *temp_script_old = NULL;
char *temp_script = "";
const char *const *i;
int j;
/* remote command */
command = jt->get_attr( jt, DRMAA_REMOTE_COMMAND );
if( command == NULL )
fsd_exc_raise_msg(
FSD_DRMAA_ERRNO_CONFLICTING_ATTRIBUTE_VALUES,
"drmaa_remote_command not set for job template"
);
command_expanded = expand->expand( expand, fsd_strdup(command), FSD_DRMAA_PH_HD | FSD_DRMAA_PH_WD );
temp_script = fsd_asprintf("#!/bin/bash\n%s",command_expanded);
fsd_free(command_expanded);
/* arguments list */
vector = jt->get_v_attr( jt, DRMAA_V_ARGV );
if( vector )
{
for( i = vector, j = 2; *i; i++, j++ )
{
char *arg_expanded = expand->expand( expand, fsd_strdup(*i), FSD_DRMAA_PH_HD | FSD_DRMAA_PH_WD );
temp_script_old = fsd_strdup(temp_script);
if (strcmp(temp_script, "") != 0) {
fsd_free(temp_script);
}
/* add too script */
temp_script = fsd_asprintf("%s '%s'", temp_script_old, arg_expanded);
fsd_free(temp_script_old);
fsd_free(arg_expanded);
}
}
job_desc->script = fsd_asprintf("%s\n", temp_script);
fsd_log_debug(("# Script:\n%s", job_desc->script));
fsd_free(temp_script);
}
END_TRY
/* start time */
value = jt->get_attr( jt, DRMAA_START_TIME );
if( value )
{
job_desc->begin_time = fsd_datetime_parse( value );
fsd_log_debug(( "\n drmaa_start_time: %s -> %ld", value, (long)job_desc->begin_time));
}
/* propagate all environment variables from submission host */
{
extern char **environ;
char **i;
unsigned j = 0;
for ( i = environ; *i; i++) {
job_desc->env_size++;
}
fsd_log_debug(("environ env_size = %d",job_desc->env_size));
fsd_calloc(job_desc->environment, job_desc->env_size+1, char *);
for ( i = environ; *i; i++,j++ ) {
job_desc->environment[j] = fsd_strdup(*i);
}
}
/* environment */
vector = jt->get_v_attr( jt, DRMAA_V_ENV );
if( vector )
{
const char *const *i;
unsigned j = 0;
unsigned env_offset = job_desc->env_size;
for( i = vector; *i; i++ )
{
job_desc->env_size++;
}
fsd_log_debug(("jt env_size = %d",job_desc->env_size));
fsd_log_debug(("# environment ="));
fsd_realloc(job_desc->environment, job_desc->env_size+1, char *);
for( i = vector; *i; i++,j++ )
{
job_desc->environment[j + env_offset] = fsd_strdup(*i);
fsd_log_debug((" %s", job_desc->environment[j+ env_offset]));
}
}
/* wall clock time hard limit */
value = jt->get_attr( jt, DRMAA_WCT_HLIMIT );
if (value)
{
job_desc->time_limit = slurmdrmaa_datetime_parse( value );
fsd_log_debug(("# wct_hlimit = %s -> %ld",value, (long int)slurmdrmaa_datetime_parse( value )));
}
/*expand->set(expand, FSD_DRMAA_PH_INCR,fsd_asprintf("%d", n_job));*/ /* set current value */
/* TODO: test drmaa_ph_incr */
/* job working directory */
value = jt->get_attr( jt, DRMAA_WD );
if( value )
{
char *cwd_expanded = expand->expand( expand, fsd_strdup(value), FSD_DRMAA_PH_HD | FSD_DRMAA_PH_INCR );
expand->set( expand, FSD_DRMAA_PH_WD, fsd_strdup(cwd_expanded));
fsd_log_debug(("# work_dir = %s",cwd_expanded));
job_desc->work_dir = fsd_strdup(cwd_expanded);
fsd_free(cwd_expanded);
}
else
{
char cwdbuf[4096] = "";
if ((getcwd(cwdbuf, 4095)) == NULL) {
char errbuf[256] = "InternalError";
(void)strerror_r(errno, errbuf, 256); /*on error the default message would be returned */
fsd_log_error(("getcwd failed: %s", errbuf));
job_desc->work_dir = fsd_strdup(".");
} else {
job_desc->work_dir = fsd_strdup(cwdbuf);
}
fsd_log_debug(("work_dir(default:CWD) %s", job_desc->work_dir));
}
TRY
{
/* input path */
input_path_orig = jt->get_attr( jt, DRMAA_INPUT_PATH );
if( input_path_orig )
{
input_path = internal_map_file( expand, input_path_orig, &input_host,"input" );
fsd_log_debug(( "\n drmaa_input_path: %s -> %s", input_path_orig, input_path ));
}
/* output path */
output_path_orig = jt->get_attr( jt, DRMAA_OUTPUT_PATH );
if( output_path_orig )
{
output_path = internal_map_file( expand, output_path_orig, &output_host,"output" );
fsd_log_debug(( "\n drmaa_output_path: %s -> %s", output_path_orig, output_path ));
}
/* error path */
error_path_orig = jt->get_attr( jt, DRMAA_ERROR_PATH );
if( error_path_orig )
{
error_path = internal_map_file( expand, error_path_orig, &error_host,"error" );
fsd_log_debug(( "\n drmaa_error_path: %s -> %s", error_path_orig, error_path ));
}
/* join files */
value = jt->get_attr( jt, DRMAA_JOIN_FILES );
if( value )
{
if( (value[0] == 'y' || value[0] == 'Y') && value[1] == '\0' )
join_files = true;
else if( (value[0] == 'n' || value[0] == 'N') && value[1] == '\0' )
join_files = false;
else
fsd_exc_raise_msg(
FSD_DRMAA_ERRNO_INVALID_ATTRIBUTE_VALUE,
"invalid value of drmaa_join_files attribute" );
}
if( join_files )
{
if( output_path == NULL )
fsd_exc_raise_msg(FSD_DRMAA_ERRNO_CONFLICTING_ATTRIBUTE_VALUES, "drmaa_join_files is set and output file is not given" );
if( error_path!=NULL && 0 != strcmp( output_path, error_path ) )
fsd_log_warning(( "Error file was given but will be ignored since drmaa_join_files was set." ));
if (error_path)
fsd_free(error_path);
error_path = fsd_strdup(output_path);
}
else
{
if( error_path == NULL && output_path )
error_path = fsd_strdup( "/dev/null" );
if( output_path == NULL && error_path )
output_path = fsd_strdup( "/dev/null" );
}
/* email addresses to send notifications */
vector = jt->get_v_attr( jt, DRMAA_V_EMAIL );
if( vector && vector[0] )
{
/* only to one email address message may be send */
job_desc->mail_user = fsd_strdup(vector[0]);
job_desc->mail_type = MAIL_JOB_BEGIN | MAIL_JOB_END | MAIL_JOB_FAIL;
fsd_log_debug(("# mail_user = %s\n",vector[0]));
fsd_log_debug(("# mail_type = %o\n",job_desc->mail_type));
if( vector[1] != NULL )
{
fsd_log_error(( "SLURM only supports one e-mail notification address" ));
fsd_exc_raise_msg(FSD_DRMAA_ERRNO_INVALID_ATTRIBUTE_VALUE,"SLURM only supports one e-mail notification address");
}
}
/* block email */
value = jt->get_attr( jt, DRMAA_BLOCK_EMAIL );
if( value )
{
bool block;
if( strcmp(value, "0") == 0 )
{
block = true;
fsd_log_debug(("# block_email = true"));
fsd_log_debug(("# mail_user delated"));
fsd_free(job_desc->mail_user);
job_desc->mail_user = NULL;
}
else if( strcmp(value, "1") == 0 )
block = false;
else
fsd_exc_raise_msg(FSD_DRMAA_ERRNO_INVALID_ATTRIBUTE_VALUE,"invalid value of drmaa_block_email attribute" );
if( block && output_path == NULL )
{
fsd_log_debug(( "output path not set and we want to block e-mail, set to /dev/null" ));
output_path = fsd_strdup( "/dev/null" );
}
}
if( input_path )
{
job_desc->std_in = fsd_strdup(input_path);
fsd_log_debug(("# input = %s", input_path));
}
if( output_path )
{
job_desc->std_out = fsd_strdup(output_path);
fsd_log_debug(("# output = %s", output_path));
}
if( error_path )
{
job_desc->std_err = fsd_strdup(error_path);
fsd_log_debug(("# error = %s", error_path));
}
}
FINALLY
{
fsd_free( input_path );
fsd_free( output_path );
fsd_free( error_path );
input_path = NULL;
output_path = NULL;
error_path = NULL;
}
END_TRY
/* job category */
value = jt->get_attr( jt, DRMAA_JOB_CATEGORY );
if( value )
job_category = value;
{
fsd_conf_option_t *category_value = NULL;
category_value = fsd_conf_dict_get( session->job_categories, job_category );
if( category_value != NULL )
{
if( category_value->type != FSD_CONF_STRING )
fsd_exc_raise_fmt(
FSD_ERRNO_INTERNAL_ERROR,
"configuration error: job category should be string"
);
fsd_log_debug(("# Job category %s : %s\n",value,category_value->val.string));
slurmdrmaa_parse_native(job_desc,category_value->val.string);
}
else
{
if( value != NULL )
fsd_exc_raise_fmt(
FSD_DRMAA_ERRNO_INVALID_ATTRIBUTE_VALUE,
"invalid job category: %s", job_category
);
}
}
/* set defaults for constraints - ref: slurm.h */
fsd_log_debug(("# Setting defaults for tasks and processors" ));
job_desc->num_tasks = 1;
job_desc->min_cpus = 0;
job_desc->cpus_per_task = 0;
job_desc->pn_min_cpus = 0;
/* native specification */
value = jt->get_attr( jt, DRMAA_NATIVE_SPECIFICATION );
if( value )
{
fsd_log_debug(("# Native specification: %s\n", value));
slurmdrmaa_parse_native(job_desc, value);
}
}
char *
fsd_drmaa_session_wait_for_any_job(
fsd_drmaa_session_t *self,
const struct timespec *timeout,
int *status, fsd_iter_t **rusage,
bool dispose
)
{
fsd_job_set_t *set = self->jobs;
fsd_job_t *volatile job = NULL;
char *volatile job_id = NULL;
volatile bool locked = false;
fsd_log_enter(( "" ));
TRY
{
while( job == NULL )
{
bool signaled = true;
if( self->destroy_requested )
fsd_exc_raise_code( FSD_DRMAA_ERRNO_NO_ACTIVE_SESSION );
if( !self->enable_wait_thread )
self->update_all_jobs_status( self );
locked = fsd_mutex_lock( &self->mutex );
if( set->empty( set ) )
fsd_exc_raise_msg( FSD_DRMAA_ERRNO_INVALID_JOB,
"No job found to be waited for" );
if( (job = set->find_terminated( set )) != NULL )
break;
if( self->destroy_requested )
fsd_exc_raise_code( FSD_DRMAA_ERRNO_NO_ACTIVE_SESSION );
if( self->enable_wait_thread )
{
fsd_log_debug(( "wait_for_any_job: waiting for wait thread" ));
if( timeout )
signaled = fsd_cond_timedwait(
&self->wait_condition, &self->mutex, timeout );
else
fsd_cond_wait( &self->wait_condition, &self->mutex );
}
else
{
fsd_log_debug(( "wait_for_any_job: waiting for next check" ));
self->wait_for_job_status_change( self,
&self->wait_condition, &self->mutex, timeout );
}
locked = fsd_mutex_unlock( &self->mutex );
fsd_log_debug((
"wait_for_any_job: woken up; signaled=%d", signaled ));
if( !signaled )
fsd_exc_raise_code( FSD_DRMAA_ERRNO_EXIT_TIMEOUT );
}
fsd_log_debug(( "wait_for_any_job: waiting finished" ));
job_id = fsd_strdup( job->job_id );
job->get_termination_status( job, status, rusage );
}
EXCEPT_DEFAULT
{
if( job_id )
fsd_free( job_id );
fsd_exc_reraise();
}
FINALLY
{
if( job )
{
if( fsd_exc_get() == NULL && dispose )
{
set->remove( set, job );
job->flags |= FSD_JOB_DISPOSED;
}
job->release( job );
}
if( locked )
fsd_mutex_unlock( &self->mutex );
}
END_TRY
fsd_log_return(( " =%s", job_id ));
return job_id;
}
