Changeset 56

Timestamp:

01/08/12 17:21:05 (12 years ago)

Author:

mmamonski

Message:

handle job killing

Location:

trunk/pbs_drmaa

Files:

• job.c (modified) (3 diffs)
• submit.c (modified) (4 diffs)

trunk/pbs_drmaa/job.c

@@ -508 +508 @@
 pbsdrmaa_job_read_exit_status( const char *job_id, const char *job_state_dir_prefix)
 {
-        char *status_file = NULL;
+        char *status_file = NULL, start_file = NULL;
         FILE *fhandle = NULL;
         int exit_status = -1;
@@ -515 +515 @@
 
         status_file = fsd_asprintf("%s/%s.exitcode", job_state_dir_prefix, job_id);
+        start_file = fsd_asprintf("%s/%s.started", job_id, job_state_dir_prefix);
 
         if ((fhandle = fopen(status_file, "r")) == NULL)
          {
+                struct stat tmpstat;
+
                 fsd_log_error(("Failed to open job status file: %s", status_file));
+                if (stat(start_file, &tmpstat) == 0 && (tmpstat.st_mode & S_IFREG))
+                 {
+                        exit_status = 143; /* SIGTERM */
+                        fsd_log_info("But start file exist %s. Assuming that job was killed (exit_status=%d).", start_file, exit_status);
+                 }
          }
         else
@@ -527 +535 @@
 
         fsd_free(status_file);
+        fsd_free(start_file);
 
         return exit_status;

trunk/pbs_drmaa/submit.c

@@ -336 +336 @@
                 if( wd != NULL )
                         script_len += strlen("cd ") + strlen(wd) + strlen("; ");
+                script_len += strlen("touch ") + strlen(((pbsdrmaa_session_t *)self->session)->job_exit_status_file_prefix) + strlen("/$PBS_JOBID.started;");
                 script_len += strlen(executable);
                 if( argv != NULL )
@@ -342 +343 @@
                 if( input_path != NULL )
                         script_len += strlen(" <") + strlen(input_path);
-                if (((pbsdrmaa_session_t *)self->session)->job_exit_status_file_prefix)
-                        script_len += strlen("; echo $? >") + strlen(((pbsdrmaa_session_t *)self->session)->job_exit_status_file_prefix) + strlen("/$PBS_JOBID.exitcode");
+
+                script_len += strlen("; echo $? >") + strlen(((pbsdrmaa_session_t *)self->session)->job_exit_status_file_prefix) + strlen("/$PBS_JOBID.exitcode");
          }
 
@@ -353 +354 @@
                 if( wd != NULL )
                         s += sprintf( s, "cd %s; ", wd );
+                s += sprintf( s, "touch %s/$PBS_JOBID.started;", ((pbsdrmaa_session_t *)self->session)->job_exit_status_file_prefix)
                 s += sprintf( s, "%s", executable );
                 if( argv != NULL )
@@ -359 +361 @@
                 if( input_path != NULL )
                         s += sprintf( s, " <%s", input_path );
-                if (((pbsdrmaa_session_t *)self->session)->job_exit_status_file_prefix)
-                        s += sprintf( s, "; echo $? >%s/$PBS_JOBID.exitcode", ((pbsdrmaa_session_t *)self->session)->job_exit_status_file_prefix);
+
+                s += sprintf( s, "; echo $? >%s/$PBS_JOBID.exitcode", ((pbsdrmaa_session_t *)self->session)->job_exit_status_file_prefix);
 
                 fsd_assert( s == script+script_len );