int job_recov_binary(
char *filename, /* I */ /* pathname to job save file */
job **pjob, /* M */ /* pointer to a pointer of job structure to fill info */
char *log_buf, /* O */ /* buffer to hold error message */
size_t buf_len) /* I */ /* len of the error buffer */
{
int fds;
job *pj = *pjob;
char *pn;
#ifdef PBS_MOM
char fileid[MAXPATHLEN];
#endif
fds = open(filename, O_RDONLY, 0);
if (fds < 0)
{
snprintf(log_buf, buf_len, "unable to open %s", filename);
return -1;
}
/* read in job quick save sub-structure */
if (read_ac_socket(fds, (char *)&pj->ji_qs, sizeof(pj->ji_qs)) != sizeof(pj->ji_qs) &&
pj->ji_qs.qs_version == PBS_QS_VERSION)
{
snprintf(log_buf, buf_len, "Unable to read %s", filename);
close(fds);
return -1;
}
/* is ji_qs the version we expect? */
if (pj->ji_qs.qs_version != PBS_QS_VERSION)
{
/* ji_qs is older version */
snprintf(log_buf, buf_len,
"%s appears to be from an old version. Attempting to convert.\n",
filename);
log_err(-1, __func__, log_buf);
if (job_qs_upgrade(pj, fds, filename, pj->ji_qs.qs_version) != 0)
{
snprintf(log_buf, buf_len, "unable to upgrade %s\n", filename);
close(fds);
return -1;
}
} /* END if (pj->ji_qs.qs_version != PBS_QS_VERSION) */
/* Does file name match the internal name? */
/* This detects ghost files */
pn = strrchr(filename, (int)'/') + 1;
#ifndef PBS_MOM
if (strncmp(pn, pj->ji_qs.ji_fileprefix, strlen(pj->ji_qs.ji_fileprefix)) != 0)
#else
if(multi_mom != 0)
{
sprintf(fileid,"%s%d",pj->ji_qs.ji_fileprefix,pbs_rm_port);
}
else
{
strcpy(fileid,pj->ji_qs.ji_fileprefix);
}
if (strncmp(pn, fileid, strlen(fileid)) != 0)
#endif
{
/* mismatch, discard job */
snprintf(log_buf, buf_len, "Job Id %s does not match file name for %s",
pj->ji_qs.ji_jobid,
filename);
close(fds);
return -1;
}
/* read in working attributes */
if (recov_attr(
fds,
pj,
job_attr_def,
pj->ji_wattr,
JOB_ATR_LAST,
JOB_ATR_UNKN,
TRUE) != 0)
{
snprintf(log_buf, buf_len, "unable to recover %s (file is likely corrupted)", filename);
close(fds);
return -1;
}
#ifdef PBS_MOM
/* read in tm sockets and ips */
if (recov_tmsock(fds, pj) != 0)
{
snprintf(log_buf, buf_len,
"warning: tmsockets not recovered from %s (written by an older pbs_mom?)",
filename);
log_err(-1, __func__, log_buf);
}
#endif /* PBS_MOM */
close(fds);
return PBSE_NONE;
} /* END job_recov_binary() */
pbs_queue *que_recov(
char *filename) /* pathname to queue save file */
{
int fds;
int i;
pbs_queue *pq;
char namebuf[MAXPATHLEN];
time_t time_now = time(NULL);
pq = que_alloc(filename, TRUE); /* allocate & init queue structure space */
if (pq == NULL)
{
log_err(-1, __func__, "que_alloc failed");
return(NULL);
}
snprintf(namebuf, sizeof(namebuf), "%s%s", path_queues, filename);
fds = open(namebuf, O_RDONLY, 0);
if (fds < 0)
{
log_err(errno, __func__, "open error");
que_free(pq, TRUE);
return(NULL);
}
/* read in queue save sub-structure */
if (read_ac_socket(fds, (char *)&pq->qu_qs, sizeof(queuefix)) !=
sizeof(queuefix))
{
log_err(errno, __func__, "read error");
que_free(pq, TRUE);
close(fds);
return ((pbs_queue *)0);
}
/* read in queue attributes */
if (recov_attr(fds, pq, que_attr_def, pq->qu_attr,
QA_ATR_LAST, 0, TRUE) != 0)
{
log_err(-1, __func__, "recov_attr[common] failed");
que_free(pq, TRUE);
close(fds);
return ((pbs_queue *)0);
}
/*
* now reload the access control lists, these attributes were
* saved separately
*/
for (i = 0;i < QA_ATR_LAST;i++)
{
if (pq->qu_attr[i].at_type == ATR_TYPE_ACL)
{
recov_acl(
&pq->qu_attr[i],
&que_attr_def[i],
que_attr_def[i].at_name,
pq->qu_qs.qu_name);
}
}
/* all done recovering the queue */
close(fds);
if ((pq->qu_attr[QA_ATR_MTime].at_flags & ATR_VFLAG_SET) == 0)
{
/* if we are recovering a pre-2.1.2 queue, save a new mtime */
pq->qu_attr[QA_ATR_MTime].at_val.at_long = time_now;
pq->qu_attr[QA_ATR_MTime].at_flags = ATR_VFLAG_SET;
que_save(pq);
}
return(pq);
}
job *job_recov(
char *filename) /* I */ /* pathname to job save file */
{
int fds;
job *pj;
char *pn;
char namebuf[MAXPATHLEN];
int qs_upgrade;
#ifndef PBS_MOM
char parent_id[PBS_MAXSVRJOBID + 1];
job_array *pa;
#endif
qs_upgrade = FALSE;
pj = job_alloc(); /* allocate & initialize job structure space */
if (pj == NULL)
{
/* FAILURE - cannot alloc memory */
return(NULL);
}
strcpy(namebuf, path_jobs); /* job directory path */
strcat(namebuf, filename);
fds = open(namebuf, O_RDONLY, 0);
if (fds < 0)
{
sprintf(log_buffer, "unable to open %s",
namebuf);
log_err(errno, "job_recov", log_buffer);
free((char *)pj);
/* FAILURE - cannot open job file */
return(NULL);
}
/* read in job quick save sub-structure */
if (read(fds, (char *)&pj->ji_qs, quicksize) != (ssize_t)quicksize &&
pj->ji_qs.qs_version == PBS_QS_VERSION)
{
sprintf(log_buffer, "Unable to read %s",
namebuf);
log_err(errno, "job_recov", log_buffer);
free((char *)pj);
close(fds);
return(NULL);
}
/* is ji_qs the version we expect? */
if (pj->ji_qs.qs_version != PBS_QS_VERSION)
{
/* ji_qs is older version */
sprintf(log_buffer,
"%s appears to be from an old version. Attempting to convert.\n",
namebuf);
log_err(-1, "job_recov", log_buffer);
if (job_qs_upgrade(pj, fds, namebuf, pj->ji_qs.qs_version) != 0)
{
sprintf(log_buffer, "unable to upgrade %s\n", namebuf);
log_err(-1, "job_recov", log_buffer);
free((char *)pj);
close(fds);
return(NULL);
}
qs_upgrade = TRUE;
} /* END if (pj->ji_qs.qs_version != PBS_QS_VERSION) */
/* Does file name match the internal name? */
/* This detects ghost files */
pn = strrchr(namebuf, (int)'/') + 1;
if (strncmp(pn, pj->ji_qs.ji_fileprefix, strlen(pj->ji_qs.ji_fileprefix)) != 0)
{
/* mismatch, discard job */
sprintf(log_buffer, "Job Id %s does not match file name for %s",
pj->ji_qs.ji_jobid,
namebuf);
log_err(-1, "job_recov", log_buffer);
free((char *)pj);
close(fds);
return(NULL);
}
/* read in working attributes */
if (recov_attr(
fds,
pj,
job_attr_def,
pj->ji_wattr,
(int)JOB_ATR_LAST,
(int)JOB_ATR_UNKN,
TRUE) != 0)
{
sprintf(log_buffer, "unable to recover %s (file is likely corrupted)",
namebuf);
log_err(-1, "job_recov", log_buffer);
job_free(pj);
close(fds);
return(NULL);
}
#ifdef PBS_MOM
/* read in tm sockets and ips */
if (recov_tmsock(fds, pj) != 0)
{
sprintf(log_buffer, "warning: tmsockets not recovered from %s (written by an older pbs_mom?)",
namebuf);
log_err(-1, "job_recov", log_buffer);
}
if (recov_roottask(fds, pj) != 0)
{
sprintf(log_buffer, "warning: root task not recovered from %s (written by an older pbs_mom?)",
namebuf);
log_err(-1, "job_recov", log_buffer);
}
if (recov_jobflags(fds, pj) != 0)
{
sprintf(log_buffer, "warning: job flags not recovered from %s (written by an older pbs_mom?)", namebuf);
log_err(-1, "job_recov", log_buffer);
}
#else /* PBS_MOM */
if (pj->ji_wattr[(int)JOB_ATR_job_array_request].at_flags & ATR_VFLAG_SET)
{
/* job is part of an array. We need to put a link back to the server
job array struct for this array. We also have to link this job into
the linked list of jobs belonging to the array. */
array_get_parent_id(pj->ji_qs.ji_jobid, parent_id);
pa = get_array(parent_id);
if (strcmp(parent_id, pj->ji_qs.ji_jobid) == 0)
{
pj->ji_isparent = TRUE;
}
else
{
if (pa == NULL)
{
/* couldn't find array struct, it must not have been recovered,
treat job as indepentent job? perhaps we should delete the job
XXX_JOB_ARRAY: should I unset this?*/
pj->ji_wattr[(int)JOB_ATR_job_array_request].at_flags &= ~ATR_VFLAG_SET;
}
else
{
CLEAR_LINK(pj->ji_arrayjobs);
append_link(&pa->array_alljobs, &pj->ji_arrayjobs, (void*)pj);
pj->ji_arraystruct = pa;
pa->jobs_recovered++;
}
}
}
#endif
close(fds);
/* all done recovering the job */
if (qs_upgrade == TRUE)
{
job_save(pj, SAVEJOB_FULL);
}
return(pj);
} /* END job_recov() */
int svr_recov(
char *svrfile, /* I */
int read_only) /* I */
{
int i;
int sdb;
char log_buf[LOCAL_LOG_BUF_SIZE];
void recov_acl(pbs_attribute *, attribute_def *, const char *, const char *);
sdb = open(svrfile, O_RDONLY, 0);
if (sdb < 0)
{
if (errno == ENOENT)
{
char tmpLine[LOG_BUF_SIZE];
snprintf(tmpLine, sizeof(tmpLine), "cannot locate server database '%s' - use 'pbs_server -t create' to create new database if database has not been initialized.",
svrfile);
log_err(errno, __func__, tmpLine);
}
else
{
log_err(errno, __func__, msg_svdbopen);
}
return(-1);
}
/* read in server structure */
lock_sv_qs_mutex(server.sv_qs_mutex, __func__);
i = read_ac_socket(sdb, (char *) & server.sv_qs, sizeof(server_qs));
if (i != sizeof(server_qs))
{
unlock_sv_qs_mutex(server.sv_qs_mutex, log_buf);
if (i < 0)
log_err(errno, __func__, "read of serverdb failed");
else
log_err(errno, __func__, "short read of serverdb");
close(sdb);
return(-1);
}
/* Save the sv_jobidnumber field in case it is set by the attributes. */
i = server.sv_qs.sv_jobidnumber;
/* read in server attributes */
if (recov_attr(
sdb,
&server,
svr_attr_def,
server.sv_attr,
SRV_ATR_LAST,
0,
!read_only) != 0 )
{
unlock_sv_qs_mutex(server.sv_qs_mutex, log_buf);
log_err(errno, __func__, "error on recovering server attr");
close(sdb);
return(-1);
}
/* Restore the current job number and make it visible in qmgr print server commnad. */
if (!read_only)
{
server.sv_qs.sv_jobidnumber = i;
server.sv_attr[SRV_ATR_NextJobNumber].at_val.at_long = i;
server.sv_attr[SRV_ATR_NextJobNumber].at_flags |= ATR_VFLAG_SET| ATR_VFLAG_MODIFY;
}
unlock_sv_qs_mutex(server.sv_qs_mutex, __func__);
close(sdb);
/* recover the server various acls from their own files */
for (i = 0;i < SRV_ATR_LAST;i++)
{
if (server.sv_attr[i].at_type == ATR_TYPE_ACL)
{
recov_acl(
&server.sv_attr[i],
&svr_attr_def[i],
PBS_SVRACL,
svr_attr_def[i].at_name);
if ((!read_only) && (svr_attr_def[i].at_action != (int (*)(pbs_attribute*, void*, int))0))
{
svr_attr_def[i].at_action(
&server.sv_attr[i],
&server,
ATR_ACTION_RECOV);
}
}
} /* END for (i) */
return(PBSE_NONE);
} /* END svr_recov() */
int svr_recov(
char *svrfile, /* I */
int read_only) /* I */
{
static char *id = "svr_recov";
int i;
int sdb;
void recov_acl(attribute *, attribute_def *, char *, char *);
sdb = open(svrfile, O_RDONLY, 0);
if (sdb < 0)
{
if (errno == ENOENT)
{
char tmpLine[LOG_BUF_SIZE];
snprintf(tmpLine, sizeof(tmpLine), "cannot locate server database '%s' - use 'pbs_server -t create' to create new database if database has not been initialized.",
svrfile);
log_err(errno, id, tmpLine);
}
else
{
log_err(errno, id, msg_svdbopen);
}
return(-1);
}
/* read in server structure */
i = read(sdb, (char *) & server.sv_qs, sizeof(struct server_qs));
if (i != sizeof(struct server_qs))
{
if (i < 0)
log_err(errno, id, "read of serverdb failed");
else
log_err(errno, id, "short read of serverdb");
close(sdb);
return(-1);
}
if (strstr((char *)&server.sv_qs,"<server_db>") != NULL)
{
/* attempt to read the server database in xml */
log_event(PBSEVENT_SYSTEM,
PBS_EVENTCLASS_SERVER,
id,
"serverdb appears to be in xml format, attempting to read xml\n");
return(svr_recov_xml(svrfile,read_only));
}
/* Save the sv_jobidnumber field in case it is set by the attributes. */
i = server.sv_qs.sv_jobidnumber;
/* read in server attributes */
if (recov_attr(
sdb,
&server,
svr_attr_def,
server.sv_attr,
(int)SRV_ATR_LAST,
0,
!read_only) != 0 )
{
log_err(errno, id, "error on recovering server attr");
close(sdb);
return(-1);
}
/* Restore the current job number and make it visible in qmgr print server commnad. */
if (!read_only)
{
server.sv_qs.sv_jobidnumber = i;
server.sv_attr[(int)SRV_ATR_NextJobNumber].at_val.at_long = i;
server.sv_attr[(int)SRV_ATR_NextJobNumber].at_flags |= ATR_VFLAG_SET| ATR_VFLAG_MODIFY;
}
close(sdb);
/* recover the server various acls from their own files */
for (i = 0;i < SRV_ATR_LAST;i++)
{
if (server.sv_attr[i].at_type == ATR_TYPE_ACL)
{
recov_acl(
&server.sv_attr[i],
&svr_attr_def[i],
PBS_SVRACL,
svr_attr_def[i].at_name);
if ((!read_only) && (svr_attr_def[i].at_action != (int (*)())0))
{
svr_attr_def[i].at_action(
&server.sv_attr[i],
&server,
ATR_ACTION_RECOV);
}
}
} /* END for (i) */
return(0);
} /* END svr_recov() */
job *job_recov(
char *filename) /* I */ /* pathname to job save file */
{
int fds;
job *pj;
char *pn;
char namebuf[MAXPATHLEN];
char log_buf[LOCAL_LOG_BUF_SIZE];
#ifndef PBS_MOM
char parent_id[PBS_MAXSVRJOBID + 1];
job_array *pa;
#endif
pj = job_alloc(); /* allocate & initialize job structure space */
if (pj == NULL)
{
/* FAILURE - cannot alloc memory */
return(NULL);
}
snprintf(namebuf, MAXPATHLEN, "%s%s", path_jobs, filename); /* job directory path, filename */
fds = open(namebuf, O_RDONLY, 0);
if (fds < 0)
{
snprintf(log_buf, LOCAL_LOG_BUF_SIZE, "unable to open %s", namebuf);
log_err(errno, __func__, log_buf);
#ifndef PBS_MOM
unlock_ji_mutex(pj, __func__, "1", LOGLEVEL);
free(pj->ji_mutex);
#endif
free((char *)pj);
/* FAILURE - cannot open job file */
return(NULL);
}
/* read in job quick save sub-structure */
if (read_ac_socket(fds, (char *)&pj->ji_qs, sizeof(pj->ji_qs)) != sizeof(pj->ji_qs) &&
pj->ji_qs.qs_version == PBS_QS_VERSION)
{
snprintf(log_buf, LOCAL_LOG_BUF_SIZE, "Unable to read %s", namebuf);
log_err(errno, __func__, log_buf);
#ifndef PBS_MOM
unlock_ji_mutex(pj, __func__, "2", LOGLEVEL);
free(pj->ji_mutex);
#endif
free((char *)pj);
close(fds);
return(NULL);
}
/* is ji_qs the version we expect? */
if (pj->ji_qs.qs_version != PBS_QS_VERSION)
{
/* ji_qs is older version */
snprintf(log_buf, LOCAL_LOG_BUF_SIZE,
"%s appears to be from an old version. Attempting to convert.\n",
namebuf);
log_err(-1, __func__, log_buf);
if (job_qs_upgrade(pj, fds, namebuf, pj->ji_qs.qs_version) != 0)
{
snprintf(log_buf, LOCAL_LOG_BUF_SIZE, "unable to upgrade %s\n", namebuf);
log_err(-1, __func__, log_buf);
#ifndef PBS_MOM
unlock_ji_mutex(pj, __func__, "3", LOGLEVEL);
free(pj->ji_mutex);
#endif
free((char *)pj);
close(fds);
return(NULL);
}
} /* END if (pj->ji_qs.qs_version != PBS_QS_VERSION) */
/* Does file name match the internal name? */
/* This detects ghost files */
pn = strrchr(namebuf, (int)'/') + 1;
if (strncmp(pn, pj->ji_qs.ji_fileprefix, strlen(pj->ji_qs.ji_fileprefix)) != 0)
{
/* mismatch, discard job */
snprintf(log_buf, LOCAL_LOG_BUF_SIZE, "Job Id %s does not match file name for %s",
pj->ji_qs.ji_jobid,
namebuf);
log_err(-1, __func__, log_buf);
#ifndef PBS_MOM
unlock_ji_mutex(pj, __func__, "4", LOGLEVEL);
free(pj->ji_mutex);
#endif
free((char *)pj);
close(fds);
return(NULL);
}
/* read in working attributes */
if (recov_attr(
fds,
pj,
job_attr_def,
pj->ji_wattr,
JOB_ATR_LAST,
JOB_ATR_UNKN,
TRUE) != 0)
{
snprintf(log_buf, LOCAL_LOG_BUF_SIZE, "unable to recover %s (file is likely corrupted)", namebuf);
log_err(-1, __func__, log_buf);
#ifndef PBS_MOM
unlock_ji_mutex(pj, __func__, "5", LOGLEVEL);
job_free(pj, FALSE);
#else
mom_job_free(pj);
#endif
close(fds);
return(NULL);
}
#ifndef PBS_MOM
/* Comment out the mother superior tracking. Will be debugged later
if (pj->ji_wattr[JOB_ATR_exec_host].at_val.at_str != NULL)
{*/
/* add job to the mother superior list for it's node */
/* char *ms = strdup(pj->ji_wattr[JOB_ATR_exec_host].at_val.at_str);
char *end = strchr(ms, '/');
if (end != NULL)
*end = '\0';
if ((end = strchr(ms, '+')) != NULL)
*end = '\0';
add_to_ms_list(ms, pj);
free(ms);
}*/
#endif
#ifdef PBS_MOM
/* read in tm sockets and ips */
if (recov_tmsock(fds, pj) != 0)
{
snprintf(log_buf, LOCAL_LOG_BUF_SIZE,
"warning: tmsockets not recovered from %s (written by an older pbs_mom?)",
namebuf);
log_err(-1, __func__, log_buf);
}
#else /* not PBS_MOM */
if (strchr(pj->ji_qs.ji_jobid, '[') != NULL)
{
/* job is part of an array. We need to put a link back to the server
job array struct for this array. We also have to link this job into
the linked list of jobs belonging to the array. */
array_get_parent_id(pj->ji_qs.ji_jobid, parent_id);
pa = get_array(parent_id);
if (pa == NULL)
{
job_abt(&pj, (char *)"Array job missing array struct, aborting job");
close(fds);
return NULL;
}
strcpy(pj->ji_arraystructid, parent_id);
if (strcmp(parent_id, pj->ji_qs.ji_jobid) == 0)
{
pj->ji_is_array_template = TRUE;
}
else
{
pa->job_ids[(int)pj->ji_wattr[JOB_ATR_job_array_id].at_val.at_long] = strdup(pj->ji_qs.ji_jobid);
pa->jobs_recovered++;
/* This is a bit of a kluge, but for some reason if an array job was
on hold when the server went down the ji_wattr[JOB_ATR_hold].at_val.at_long
value is 0 on recovery even though pj->ji_qs.ji_state is JOB_STATE_HELD and
the substate is JOB_SUBSTATE_HELD
*/
if ((pj->ji_qs.ji_state == JOB_STATE_HELD) &&
(pj->ji_qs.ji_substate == JOB_SUBSTATE_HELD))
{
pj->ji_wattr[JOB_ATR_hold].at_val.at_long = HOLD_l;
pj->ji_wattr[JOB_ATR_hold].at_flags = ATR_VFLAG_SET;
}
}
if (pa != NULL)
{
unlock_ai_mutex(pa, __func__, "1", LOGLEVEL);
}
}
#endif
close(fds);
pj->ji_commit_done = 1;
/* all done recovering the job */
job_save(pj, SAVEJOB_FULL, 0);
return(pj);
} /* END job_recov() */
job *job_recov(
char *filename) /* I */ /* pathname to job save file */
{
int fds;
job *pj;
char *pn;
char namebuf[MAXPATHLEN];
int qs_upgrade;
#ifndef PBS_MOM
char parent_id[PBS_MAXSVRJOBID + 1];
job_array *pa;
#endif
qs_upgrade = FALSE;
pj = job_alloc(); /* allocate & initialize job structure space */
if (pj == NULL)
{
/* FAILURE - cannot alloc memory */
return(NULL);
}
strcpy(namebuf, path_jobs); /* job directory path */
strcat(namebuf, filename);
fds = open(namebuf, O_RDONLY, 0);
if (fds < 0)
{
sprintf(log_buffer, "unable to open %s",
namebuf);
log_err(errno, "job_recov", log_buffer);
free((char *)pj);
/* FAILURE - cannot open job file */
return(NULL);
}
/* read in job quick save sub-structure */
if (read(fds, (char *)&pj->ji_qs, quicksize) != (ssize_t)quicksize &&
pj->ji_qs.qs_version == PBS_QS_VERSION)
{
sprintf(log_buffer, "Unable to read %s",
namebuf);
log_err(errno, "job_recov", log_buffer);
free((char *)pj);
close(fds);
return(NULL);
}
/* is ji_qs the version we expect? */
if (pj->ji_qs.qs_version != PBS_QS_VERSION)
{
/* ji_qs is older version */
sprintf(log_buffer,
"%s appears to be from an old version. Attempting to convert.\n",
namebuf);
log_err(-1, "job_recov", log_buffer);
if (job_qs_upgrade(pj, fds, namebuf, pj->ji_qs.qs_version) != 0)
{
sprintf(log_buffer, "unable to upgrade %s\n", namebuf);
log_err(-1, "job_recov", log_buffer);
free((char *)pj);
close(fds);
return(NULL);
}
qs_upgrade = TRUE;
} /* END if (pj->ji_qs.qs_version != PBS_QS_VERSION) */
/* Does file name match the internal name? */
/* This detects ghost files */
pn = strrchr(namebuf, (int)'/') + 1;
if (strncmp(pn, pj->ji_qs.ji_fileprefix, strlen(pj->ji_qs.ji_fileprefix)) != 0)
{
/* mismatch, discard job */
sprintf(log_buffer, "Job Id %s does not match file name for %s",
pj->ji_qs.ji_jobid,
namebuf);
log_err(-1, "job_recov", log_buffer);
free((char *)pj);
close(fds);
return(NULL);
}
/* read in working attributes */
if (recov_attr(
fds,
pj,
job_attr_def,
pj->ji_wattr,
(int)JOB_ATR_LAST,
(int)JOB_ATR_UNKN,
TRUE) != 0)
{
sprintf(log_buffer, "unable to recover %s (file is likely corrupted)",
namebuf);
log_err(-1, "job_recov", log_buffer);
job_free(pj);
close(fds);
return(NULL);
}
#ifdef PBS_MOM
/* read in tm sockets and ips */
if (recov_tmsock(fds, pj) != 0)
{
sprintf(log_buffer, "warning: tmsockets not recovered from %s (written by an older pbs_mom?)",
namebuf);
log_err(-1, "job_recov", log_buffer);
}
#else /* PBS_MOM */
if (pj->ji_wattr[(int)JOB_ATR_job_array_request].at_flags & ATR_VFLAG_SET)
{
/* job is part of an array. We need to put a link back to the server
job array struct for this array. We also have to link this job into
the linked list of jobs belonging to the array. */
array_get_parent_id(pj->ji_qs.ji_jobid, parent_id);
pa = get_array(parent_id);
if (pa == NULL)
{
job_abt(&pj, "Array job missing array struct, aborting job");
close(fds);
return NULL;
}
if (strcmp(parent_id, pj->ji_qs.ji_jobid) == 0)
{
pj->ji_is_array_template = TRUE;
pj->ji_arraystruct = pa;
}
else
{
pa->jobs[(int)pj->ji_wattr[JOB_ATR_job_array_id].at_val.at_long] = (void *)pj;
pj->ji_arraystruct = pa;
pa->jobs_recovered++;
/* This is a bit of a kluge, but for some reason if an array job was
on hold when the server went down the ji_wattr[JOB_ATR_hold].at_val.at_long
value is 0 on recovery even though pj->ji_qs.ji_state is JOB_STATE_HELD and
the substate is JOB_SUBSTATE_HELD
*/
if ((pj->ji_qs.ji_state == JOB_STATE_HELD) && (pj->ji_qs.ji_substate == JOB_SUBSTATE_HELD))
{
pj->ji_wattr[JOB_ATR_hold].at_val.at_long = HOLD_l;
pj->ji_wattr[JOB_ATR_hold].at_flags = ATR_VFLAG_SET;
}
}
}
#endif
close(fds);
/* all done recovering the job */
job_save(pj, SAVEJOB_FULL);
return(pj);
} /* END job_recov() */
