void
req_stat_que(struct batch_request *preq)
{
char *name;
pbs_queue *pque;
struct batch_reply *preply;
int rc = 0;
int type = 0;
/*
* first, validate the name of the requested object, either
* a queue, or null for all queues
*/
name = preq->rq_ind.rq_status.rq_id;
if ((*name == '\0') || (*name =='@'))
type = 1;
else {
pque = find_queuebyname(name);
#ifdef NAS /* localmod 075 */
if (pque == NULL)
pque = find_resvqueuebyname(name);
#endif /* localmod 075 */
if (pque == NULL) {
req_reject(PBSE_UNKQUE, 0, preq);
return;
}
}
preply = &preq->rq_reply;
preply->brp_choice = BATCH_REPLY_CHOICE_Status;
CLEAR_HEAD(preply->brp_un.brp_status);
if (type == 0) { /* get status of the one named queue */
rc = status_que(pque, preq, &preply->brp_un.brp_status);
} else { /* get status of queues */
pque = (pbs_queue *)GET_NEXT(svr_queues);
while (pque) {
rc = status_que(pque, preq, &preply->brp_un.brp_status);
if (rc != 0) {
if (rc == PBSE_PERM)
rc = 0;
else
break;
}
pque = (pbs_queue *)GET_NEXT(pque->qu_link);
}
}
if (rc) {
(void)reply_free(preply);
req_reject(rc, bad, preq);
} else {
(void)reply_send(preq);
}
}
pbs_queue *
get_dfltque(void)
{
pbs_queue *pq = (pbs_queue *)0;
if (server.sv_attr[SRV_ATR_dflt_que].at_flags & ATR_VFLAG_SET)
pq = find_queuebyname(server.sv_attr[SRV_ATR_dflt_que].at_val.at_str);
return (pq);
}
pbs_queue *get_dfltque(void)
{
pbs_queue *pq = NULL;
if (server.sv_attr[SRV_ATR_dflt_que].at_flags & ATR_VFLAG_SET)
{
pq = find_queuebyname(server.sv_attr[SRV_ATR_dflt_que].at_val.at_str);
}
return(pq);
} /* END get_dfltque() */
pbs_queue *get_dfltque(void)
{
pbs_queue *pq = NULL;
char *dque = NULL;
if (get_svr_attr_str(SRV_ATR_dflt_que, &dque) == PBSE_NONE)
{
pq = find_queuebyname(dque);
}
return(pq);
} /* END get_dfltque() */
static int local_move(
job *jobp,
struct batch_request *req)
{
char *id = "local_move";
pbs_queue *qp;
char *destination = jobp->ji_qs.ji_destin;
int mtype;
/* search for destination queue */
if ((qp = find_queuebyname(destination)) == NULL)
{
sprintf(log_buffer, "queue %s does not exist\n",
destination);
log_err(-1, id, log_buffer);
pbs_errno = PBSE_UNKQUE;
return(ROUTE_PERM_FAILURE);
}
/*
* if being moved at specific request of administrator, then
* checks on queue availability, etc. are skipped;
* otherwise all checks are enforced.
*/
if (req == 0)
{
mtype = MOVE_TYPE_Route; /* route */
}
else if (req->rq_perm & (ATR_DFLAG_MGRD | ATR_DFLAG_MGWR))
{
mtype = MOVE_TYPE_MgrMv; /* privileged move */
}
else
{
mtype = MOVE_TYPE_Move; /* non-privileged move */
}
if ((pbs_errno = svr_chkque(
jobp,
qp,
get_variable(jobp, pbs_o_host), mtype, NULL)))
{
/* should this queue be retried? */
return(should_retry_route(pbs_errno));
}
/* dequeue job from present queue, update destination and */
/* queue_rank for new queue and enqueue into destination */
svr_dequejob(jobp);
strcpy(jobp->ji_qs.ji_queue, destination);
jobp->ji_wattr[JOB_ATR_qrank].at_val.at_long = ++queue_rank;
pbs_errno = svr_enquejob(jobp);
if (pbs_errno != 0)
{
return(ROUTE_PERM_FAILURE); /* should never ever get here */
}
jobp->ji_lastdest = 0; /* reset in case of another route */
job_save(jobp, SAVEJOB_FULL);
return(ROUTE_SUCCESS);
} /* END local_move() */
int req_stat_job(
struct batch_request *preq) /* ptr to the decoded request */
{
struct stat_cntl *cntl; /* see svrfunc.h */
char *name;
job *pjob = NULL;
pbs_queue *pque = NULL;
int rc = PBSE_NONE;
char log_buf[LOCAL_LOG_BUF_SIZE];
enum TJobStatTypeEnum type = tjstNONE;
/*
* first, validate the name of the requested object, either
* a job, a queue, or the whole server.
*/
if (LOGLEVEL >= 7)
{
sprintf(log_buf, "note");
log_event(PBSEVENT_JOB, PBS_EVENTCLASS_JOB, __func__, log_buf);
}
/* FORMAT: name = { <JOBID> | <QUEUEID> | '' } */
name = preq->rq_ind.rq_status.rq_id;
if (preq->rq_extend != NULL)
{
/* evaluate pbs_job_stat() 'extension' field */
if (!strncasecmp(preq->rq_extend, "truncated", strlen("truncated")))
{
/* truncate response by 'max_report' */
type = tjstTruncatedServer;
}
else if (!strncasecmp(preq->rq_extend, "summarize_arrays", strlen("summarize_arrays")))
{
type = tjstSummarizeArraysServer;
}
} /* END if (preq->rq_extend != NULL) */
if (isdigit((int)*name))
{
/* status a single job */
if (is_array(name))
{
if (type != tjstSummarizeArraysServer)
{
type = tjstArray;
}
}
else
{
type = tjstJob;
if ((pjob = svr_find_job(name, FALSE)) == NULL)
{
rc = PBSE_UNKJOBID;
}
else
unlock_ji_mutex(pjob, __func__, "1", LOGLEVEL);
}
}
else if (isalpha(name[0]))
{
if (type == tjstNONE)
type = tjstQueue;
else if (type == tjstSummarizeArraysServer)
type = tjstSummarizeArraysQueue;
else
type = tjstTruncatedQueue;
/* if found, this mutex is released later */
if ((pque = find_queuebyname(name)) == NULL)
{
rc = PBSE_UNKQUE;
}
}
else if ((*name == '\0') || (*name == '@'))
{
/* status all jobs at server */
if (type == tjstNONE)
type = tjstServer;
}
else
{
rc = PBSE_IVALREQ;
}
if (rc != 0)
{
/* is invalid - an error */
req_reject(rc, 0, preq, NULL, NULL);
return(rc);
}
preq->rq_reply.brp_choice = BATCH_REPLY_CHOICE_Status;
CLEAR_HEAD(preq->rq_reply.brp_un.brp_status);
cntl = (struct stat_cntl *)calloc(1, sizeof(struct stat_cntl));
if (cntl == NULL)
{
if (pque != NULL)
unlock_queue(pque, "req_stat_job", (char *)"no memory cntl", LOGLEVEL);
req_reject(PBSE_SYSTEM, 0, preq, NULL, NULL);
return(PBSE_SYSTEM);
}
if ((type == tjstTruncatedQueue) ||
(type == tjstTruncatedServer))
{
if (pque != NULL)
{
unlock_queue(pque, __func__, "", LOGLEVEL);
pque = NULL;
}
}
cntl->sc_type = (int)type;
cntl->sc_conn = -1;
cntl->sc_pque = pque;
cntl->sc_origrq = preq;
cntl->sc_post = req_stat_job_step2;
cntl->sc_jobid[0] = '\0'; /* cause "start from beginning" */
req_stat_job_step2(cntl); /* go to step 2, see if running is current */
if (pque != NULL)
unlock_queue(pque, "req_stat_job", (char *)"success", LOGLEVEL);
free(cntl);
return(PBSE_NONE);
} /* END req_stat_job() */
int req_stat_que(
batch_request *preq)
{
char *name;
pbs_queue *pque = NULL;
struct batch_reply *preply;
int rc = 0;
int type = 0;
char log_buf[LOCAL_LOG_BUF_SIZE+1];
/*
* first, validate the name of the requested object, either
* a queue, or null for all queues
*/
name = preq->rq_ind.rq_status.rq_id;
if ((*name == '\0') || (*name == '@'))
{
type = 1;
}
else
{
pque = find_queuebyname(name);
if (pque == NULL)
{
rc = PBSE_UNKQUE;
snprintf(log_buf, LOCAL_LOG_BUF_SIZE, "cannot locate queue %s", name);
req_reject(rc, 0, preq, NULL, log_buf);
return rc;
}
}
preply = &preq->rq_reply;
preply->brp_choice = BATCH_REPLY_CHOICE_Status;
CLEAR_HEAD(preply->brp_un.brp_status);
if (type == 0)
{
/* get status of the named queue */
mutex_mgr pque_mutex = mutex_mgr(pque->qu_mutex, true);
rc = status_que(pque, preq, &preply->brp_un.brp_status);
/* pque_qu_mutex will be unlocked in the destructor when we leave this scope */
}
else
{
/* pque == NULL before next_queue */
svr_queues.lock();
all_queues_iterator *iter = svr_queues.get_iterator();
svr_queues.unlock();
/* get status of all queues */
while ((pque = next_queue(&svr_queues,iter)) != NULL)
{
mutex_mgr pque_mutex = mutex_mgr(pque->qu_mutex, true);
rc = status_que(pque, preq, &preply->brp_un.brp_status);
if (rc != 0)
{
if (rc != PBSE_PERM)
{
break;
}
rc = 0;
}
}
delete iter;
}
if (rc != PBSE_NONE)
{
reply_free(preply);
req_reject(PBSE_NOATTR, rc, preq, NULL, "status_queue failed");
}
else
{
reply_send_svr(preq);
}
return rc;
} /* END req_stat_que() */
int req_stat_que(
struct batch_request *preq)
{
char *name;
pbs_queue *pque = NULL;
struct batch_reply *preply;
int rc = 0;
int type = 0;
char log_buf[LOCAL_LOG_BUF_SIZE+1];
/*
* first, validate the name of the requested object, either
* a queue, or null for all queues
*/
name = preq->rq_ind.rq_status.rq_id;
if ((*name == '\0') || (*name == '@'))
{
type = 1;
}
else
{
pque = find_queuebyname(name);
if (pque == NULL)
{
rc = PBSE_UNKQUE;
snprintf(log_buf, LOCAL_LOG_BUF_SIZE, "can not locate queue %s", name);
req_reject(rc, 0, preq, NULL, log_buf);
return rc;
}
}
preply = &preq->rq_reply;
preply->brp_choice = BATCH_REPLY_CHOICE_Status;
CLEAR_HEAD(preply->brp_un.brp_status);
if (type == 0)
{
/* get status of the named queue */
rc = status_que(pque, preq, &preply->brp_un.brp_status);
unlock_queue(pque, "req_stat_que", "type == 0", LOGLEVEL);
}
else
{
/* pque == NULL before next_queue */
int iter = -1;
/* get status of all queues */
while ((pque = next_queue(&svr_queues,&iter)) != NULL)
{
rc = status_que(pque, preq, &preply->brp_un.brp_status);
if (rc != 0)
{
if (rc != PBSE_PERM)
{
unlock_queue(pque, "req_stat_que", "break", LOGLEVEL);
break;
}
rc = 0;
}
unlock_queue(pque, "req_stat_que", "end while", LOGLEVEL);
}
}
if (rc != PBSE_NONE)
{
reply_free(preply);
req_reject(PBSE_NOATTR, rc, preq, NULL, "status_queue failed");
}
else
{
reply_send_svr(preq);
}
return rc;
} /* END req_stat_que() */
void *queue_route(
void *vp)
{
pbs_queue *pque;
job *pjob = NULL;
char *queue_name;
char log_buf[LOCAL_LOG_BUF_SIZE];
int iter = -1;
time_t time_now = time(NULL);
queue_name = (char *)vp;
if (queue_name == NULL)
{
sprintf(log_buf, "NULL queue name");
log_err(-1, __func__, log_buf);
return(NULL);
}
if (LOGLEVEL >= 7)
{
snprintf(log_buf, sizeof(log_buf), "queue name: %s", queue_name);
log_event(PBSEVENT_SYSTEM, PBS_EVENTCLASS_QUEUE, __func__, log_buf);
}
pthread_mutex_lock(reroute_job_mutex);
pque = find_queuebyname(queue_name);
if (pque == NULL)
{
sprintf(log_buf, "Could not find queue %s", queue_name);
log_err(-1, __func__, log_buf);
free(queue_name);
pthread_mutex_unlock(reroute_job_mutex);
return(NULL);
}
while ((pjob = next_job(pque->qu_jobs,&iter)) != NULL)
{
/* the second condition says we only want to try if routing
* has been tried once - this is to let req_commit have the
* first crack at routing always */
unlock_queue(pque, __func__, (char *)NULL, 0);
if ((pjob->ji_qs.ji_un.ji_routet.ji_rteretry <= time_now - ROUTE_RETRY_TIME) &&
(pjob->ji_qs.ji_un.ji_routet.ji_rteretry != 0))
{
reroute_job(pjob, pque);
unlock_ji_mutex(pjob, __func__, (char *)"1", LOGLEVEL);
}
else
unlock_ji_mutex(pjob, __func__, (char *)"1", LOGLEVEL);
}
free(queue_name);
unlock_queue(pque, __func__, (char *)NULL, 0);
pthread_mutex_unlock(reroute_job_mutex);
return(NULL);
} /* END queue_route() */
void *queue_route(
void *vp)
{
pbs_queue *pque;
job *pjob = NULL;
char *queue_name;
char log_buf[LOCAL_LOG_BUF_SIZE];
all_jobs_iterator *iter = NULL;
queue_name = (char *)vp;
if (queue_name == NULL)
{
sprintf(log_buf, "NULL queue name");
log_err(-1, __func__, log_buf);
return(NULL);
}
while (1)
{
pthread_mutex_lock(reroute_job_mutex);
/* Before we attempt to service this queue, make sure we can find it. */
pque = find_queuebyname(queue_name);
if (pque == NULL)
{
sprintf(log_buf, "Could not find queue %s", queue_name);
log_err(-1, __func__, log_buf);
free(queue_name);
return(NULL);
}
mutex_mgr que_mutex(pque->qu_mutex, true);
pque->qu_jobs->lock();
iter = pque->qu_jobs->get_iterator();
pque->qu_jobs->unlock();
if (LOGLEVEL >= 7)
{
snprintf(log_buf, sizeof(log_buf), "routing any ready jobs in queue: %s", queue_name);
log_event(PBSEVENT_SYSTEM, PBS_EVENTCLASS_QUEUE, __func__, log_buf);
}
while ((pjob = next_job(pque->qu_jobs,iter)) != NULL)
{
/* We only want to try if routing has been tried at least once - this is to let
* req_commit have the first crack at routing always. */
if (pjob->ji_commit_done == 0) /* when req_commit is done it will set ji_commit_done to 1 */
{
unlock_ji_mutex(pjob, __func__, "1", LOGLEVEL);
continue;
}
/* queue must be unlocked when calling reroute_job */
que_mutex.unlock();
reroute_job(pjob);
unlock_ji_mutex(pjob, __func__, "2", LOGLEVEL);
/* need to relock queue when we go to call next_job */
pque = find_queuebyname(queue_name);
if (pque == NULL)
{
sprintf(log_buf, "Could not find queue %s", queue_name);
log_err(-1, __func__, log_buf);
free(queue_name);
delete iter;
return(NULL);
}
que_mutex.mark_as_locked();
}
/* we come out of the while loop with the queue locked.
We don't want it locked while we sleep */
que_mutex.unlock();
pthread_mutex_unlock(reroute_job_mutex);
delete iter;
sleep(route_retry_interval);
}
free(queue_name);
return(NULL);
} /* END queue_route() */
int local_move(
job *pjob,
int *my_err,
struct batch_request *req)
{
pbs_queue *dest_que = NULL;
char *destination = pjob->ji_qs.ji_destin;
int mtype;
char log_buf[LOCAL_LOG_BUF_SIZE];
char job_id[PBS_MAXSVRJOBID+1];
int rc;
bool reservation = false;
/* Sometimes multiple threads are trying to route the same job. Protect against this
* by making sure that the destionation queue and the current queue are different.
* If they are the same then consider it done correctly */
if (!strcmp(pjob->ji_qs.ji_queue, pjob->ji_qs.ji_destin))
return(PBSE_NONE);
if (LOGLEVEL >= 8)
{
sprintf(log_buf, "%s", pjob->ji_qs.ji_jobid);
log_event(PBSEVENT_JOB, PBS_EVENTCLASS_JOB, __func__, log_buf);
}
/*
* if being moved at specific request of administrator, then
* checks on queue availability, etc. are skipped;
* otherwise all checks are enforced.
*/
if (req == 0)
{
mtype = MOVE_TYPE_Route; /* route */
}
else if (req->rq_perm & (ATR_DFLAG_MGRD | ATR_DFLAG_MGWR))
{
mtype = MOVE_TYPE_MgrMv; /* privileged move */
}
else
{
mtype = MOVE_TYPE_Move; /* non-privileged move */
}
strcpy(job_id, pjob->ji_qs.ji_jobid);
unlock_ji_mutex(pjob, __func__, NULL, LOGLEVEL);
dest_que = find_queuebyname(destination);
if (dest_que == NULL)
{
/* this should never happen */
sprintf(log_buf, "queue %s does not exist\n", pjob->ji_qs.ji_queue);
log_err(-1, __func__, log_buf);
*my_err = PBSE_UNKQUE;
return(-1);
}
mutex_mgr dest_que_mutex = mutex_mgr(dest_que->qu_mutex, true);
if ((pjob = svr_find_job(job_id, TRUE)) == NULL)
{
/* job disappeared while locking queue */
return(PBSE_JOB_RECYCLED);
}
/* check the destination */
if ((*my_err = svr_chkque(pjob, dest_que, get_variable(pjob, pbs_o_host), mtype, NULL)))
{
/* should this queue be retried? */
return(should_retry_route(*my_err));
}
reservation = have_reservation(pjob, dest_que);
/* dequeue job from present queue, update destination and */
/* queue_rank for new queue and enqueue into destination */
dest_que_mutex.unlock();
rc = svr_dequejob(pjob, FALSE);
if (rc)
return(rc);
snprintf(pjob->ji_qs.ji_queue, sizeof(pjob->ji_qs.ji_queue), "%s", destination);
pjob->ji_wattr[JOB_ATR_qrank].at_val.at_long = ++queue_rank;
if ((*my_err = svr_enquejob(pjob, FALSE, NULL, reservation, false)) == PBSE_JOB_RECYCLED)
return(-1);
if (*my_err != PBSE_NONE)
{
return(-1); /* should never ever get here */
}
if (pjob != NULL)
{
pjob->ji_lastdest = 0; /* reset in case of another route */
job_save(pjob, SAVEJOB_FULL, 0);
}
return(PBSE_NONE);
} /* END local_move() */
void req_stat_job(struct batch_request *preq)
{
int at_least_one_success = 0;
int dosubjobs = 0;
int dohistjobs = 0;
char *name;
job *pjob = NULL;
pbs_queue *pque = NULL;
struct batch_reply *preply;
int rc = 0;
int type = 0;
char *pnxtjid = NULL;
/* check for any extended flag in the batch request. 't' for
* the sub jobs. If 'x' is there, then check if the server is
* configured for history job info. If not set or set to FALSE,
* return with PBSE_JOBHISTNOTSET error. Otherwise select history
* jobs.
*/
if (preq->rq_extend) {
if (strchr(preq->rq_extend, (int)'t'))
dosubjobs = 1; /* status sub jobs of an Array Job */
if (strchr(preq->rq_extend, (int)'x')) {
if (svr_history_enable == 0) {
req_reject(PBSE_JOBHISTNOTSET, 0, preq);
return;
}
dohistjobs = 1; /* status history jobs */
}
}
/*
* first, validate the name of the requested object, either
* a job, a queue, or the whole server.
* type = 1 for a job, Array job, subjob or range of subjobs, or
* a comma separated list of the above.
* 2 for jobs in a queue,
* 3 for jobs in the server, or
*/
name = preq->rq_ind.rq_status.rq_id;
if ( isdigit((int)*name) ) {
/* a single job id */
type = 1;
rc = PBSE_UNKJOBID;
} else if ( isalpha((int)*name) ) {
pque = find_queuebyname(name) /* status jobs in a queue */;
#ifdef NAS /* localmod 075 */
if (pque == NULL)
pque = find_resvqueuebyname(name);
#endif /* localmod 075 */
if (pque)
type = 2;
else
rc = PBSE_UNKQUE;
} else if ((*name == '\0') || (*name == '@')) {
type = 3; /* status all jobs at server */
} else
rc = PBSE_IVALREQ;
if (type == 0) { /* is invalid - an error */
req_reject(rc, 0, preq);
return;
}
preply = &preq->rq_reply;
preply->brp_choice = BATCH_REPLY_CHOICE_Status;
CLEAR_HEAD(preply->brp_un.brp_status);
rc = PBSE_NONE;
if (type == 1) {
/*
* If there is more than one job id, any status for any
* one job is returned, then no error is given.
* If a single job id is requested and there is an error
* the error is returned.
*/
pnxtjid = name;
while ((name = parse_comma_string_r(&pnxtjid)) != NULL) {
if ((rc = stat_a_jobidname(preq, name, dohistjobs, dosubjobs)) == PBSE_NONE)
at_least_one_success = 1;
}
if (at_least_one_success == 1)
reply_send(preq);
else
req_reject(rc, 0, preq);
return;
} else if (type == 2) {
pjob = (job *)GET_NEXT(pque->qu_jobs);
while (pjob && (rc == PBSE_NONE)) {
rc = do_stat_of_a_job(preq, pjob, dohistjobs, dosubjobs);
pjob = (job *)GET_NEXT(pjob->ji_jobque);
}
} else {
pjob = (job *)GET_NEXT(svr_alljobs);
while (pjob && (rc == PBSE_NONE)) {
rc = do_stat_of_a_job(preq, pjob, dohistjobs, dosubjobs);
pjob = (job *)GET_NEXT(pjob->ji_alljobs);
}
}
if (rc && (rc != PBSE_PERM))
req_reject(rc, bad, preq);
else
reply_send(preq);
}
/**
* @brief
* Move a job to another queue in this Server.
*
* @par
* Check the destination to see if it can accept the job.
* If the job can enter the new queue, dequeue from the existing queue and
* enqueue into the new queue
*
* @par
* Note - the destination is specified by the queue's name in the
* ji_qs.ji_destin element of the job structure.
*
* param[in] jobp - pointer to job to move
* param[in] req - client request from a qmove client, null if a route
*
* @return int
* @retval 0 : success
* @retval -1 : permanent failure or rejection, see pbs_errno
* @retval 1 : failed but try again later
*/
int
local_move(job *jobp, struct batch_request *req)
{
pbs_queue *qp;
char *destination = jobp->ji_qs.ji_destin;
int mtype;
attribute *pattr;
long newtype = -1;
/* search for destination queue */
if ((qp = find_queuebyname(destination)) == NULL) {
sprintf(log_buffer,
"queue %s does not exist",
destination);
log_err(-1, __func__, log_buffer);
pbs_errno = PBSE_UNKQUE;
return -1;
}
/*
* if being moved at specific request of administrator, then
* checks on queue availability, etc. are skipped;
* otherwise all checks are enforced.
*/
if (req == NULL) {
mtype = MOVE_TYPE_Route; /* route */
} else if (req->rq_perm & (ATR_DFLAG_MGRD | ATR_DFLAG_MGWR)) {
mtype = MOVE_TYPE_MgrMv; /* privileged move */
} else {
mtype = MOVE_TYPE_Move; /* non-privileged move */
}
pbs_errno = svr_chkque(jobp, qp,
get_hostPart(jobp->ji_wattr[(int)JOB_ATR_job_owner].at_val.at_str),
mtype);
if (pbs_errno) {
/* should this queue be retried? */
return (should_retry_route(pbs_errno));
}
/* dequeue job from present queue, update destination and */
/* queue_rank for new queue and enqueue into destination */
svr_dequejob(jobp);
jobp->ji_myResv = NULL;
strncpy(jobp->ji_qs.ji_queue, qp->qu_qs.qu_name, PBS_MAXQUEUENAME);
jobp->ji_qs.ji_queue[PBS_MAXQUEUENAME] = '\0';
jobp->ji_wattr[(int)JOB_ATR_qrank].at_val.at_long = ++queue_rank;
jobp->ji_wattr[(int)JOB_ATR_qrank].at_flags |= ATR_VFLAG_MODCACHE;
pattr = &jobp->ji_wattr[(int)JOB_ATR_reserve_ID];
if (qp->qu_resvp) {
job_attr_def[(int)JOB_ATR_reserve_ID].at_decode(pattr,
(char *)0, (char *)0, qp->qu_resvp->ri_qs.ri_resvID);
jobp->ji_myResv = qp->qu_resvp;
} else {
job_attr_def[(int)JOB_ATR_reserve_ID].at_decode(pattr,
(char *)0, (char *)0, (char*)0);
}
if (server.sv_attr[(int)SRV_ATR_EligibleTimeEnable].at_val.at_long == 1) {
newtype = determine_accruetype(jobp);
if (newtype == -1)
/* unable to determine accruetype, set it to NEW */
(void)update_eligible_time(JOB_INITIAL, jobp);
else
/* found suiting accruetype, update to this */
(void)update_eligible_time(newtype, jobp);
}
if ((pbs_errno = svr_enquejob(jobp)) != 0)
return -1; /* should never ever get here */
jobp->ji_lastdest = 0; /* reset in case of another route */
(void)job_save(jobp, SAVEJOB_FULL);
/* If a scheduling cycle is in progress, then this moved job may have
* had changes resulting from the move that would impact scheduling or
* placement, add job to list of jobs which cannot be run in this cycle.
*/
if ((req == NULL || (req->rq_conn != scheduler_sock)) && (scheduler_jobs_stat))
am_jobs_add(jobp);
return 0;
}
static void req_stat_job_step2(
struct stat_cntl *cntl) /* I/O (freed on return) */
{
svrattrl *pal;
job *pjob = NULL;
struct batch_request *preq;
struct batch_reply *preply;
int rc = 0;
enum TJobStatTypeEnum type;
pbs_queue *pque = NULL;
int exec_only = 0;
int IsTruncated = 0;
long DTime; /* delta time - only report full attribute list if J->MTime > DTime */
static svrattrl *dpal = NULL;
int job_array_index = 0;
job_array *pa = NULL;
preq = cntl->sc_origrq;
type = (enum TJobStatTypeEnum)cntl->sc_type;
preply = &preq->rq_reply;
/* See pbs_server_attributes(1B) for details on "poll_jobs" behaviour */
/* NOTE: If IsTruncated is true, should walk all queues and walk jobs in each queue
until max_reported is reached (NYI) */
if (dpal == NULL)
{
/* build 'delta' attribute list */
svrattrl *tpal;
tlist_head dalist;
int aindex;
int atrlist[] =
{
JOB_ATR_jobname,
JOB_ATR_resc_used,
JOB_ATR_LAST
};
CLEAR_LINK(dalist);
for (aindex = 0;atrlist[aindex] != JOB_ATR_LAST;aindex++)
{
if ((tpal = attrlist_create("", "", 23)) == NULL)
{
return;
}
tpal->al_valln = atrlist[aindex];
if (dpal == NULL)
dpal = tpal;
append_link(&dalist, &tpal->al_link, tpal);
}
} /* END if (dpal == NULL) */
if (type == tjstArray)
{
pa = get_array(preq->rq_ind.rq_status.rq_id);
}
if (!server.sv_attr[(int)SRV_ATR_PollJobs].at_val.at_long)
{
/* polljobs not set - indicates we may need to obtain fresh data from
MOM */
if (cntl->sc_jobid[0] == '\0')
pjob = NULL;
else
pjob = find_job(cntl->sc_jobid);
while (1)
{
if (pjob == NULL)
{
/* start from the first job */
if (type == tjstJob)
{
pjob = find_job(preq->rq_ind.rq_status.rq_id);
}
else if (type == tjstQueue)
{
pjob = (job *)GET_NEXT(cntl->sc_pque->qu_jobs);
}
else if (type == tjstArray)
{
job_array_index = 0;
/* increment job_array_index until we find a non-null pointer or hit the end */
while (job_array_index < pa->ai_qs.array_size && (pjob = pa->jobs[job_array_index]) == NULL)
job_array_index++;
}
else
{
if ((type == tjstTruncatedServer) || (type == tjstTruncatedQueue))
IsTruncated = TRUE;
pjob = (job *)GET_NEXT(svr_alljobs);
}
} /* END if (pjob == NULL) */
else
{
/* get next job */
if (type == tjstJob)
break;
if (type == tjstQueue)
pjob = (job *)GET_NEXT(pjob->ji_jobque);
else
pjob = (job *)GET_NEXT(pjob->ji_alljobs);
if (type == tjstArray)
{
pjob = NULL;
/* increment job_array_index until we find a non-null pointer or hit the end */
while (++job_array_index < pa->ai_qs.array_size && (pjob = pa->jobs[job_array_index]) == NULL)
;
}
}
if (pjob == NULL)
break;
/* PBS_RESTAT_JOB defaults to 30 seconds */
if ((pjob->ji_qs.ji_substate == JOB_SUBSTATE_RUNNING) &&
((time_now - pjob->ji_momstat) > JobStatRate))
{
/* go to MOM for status */
strcpy(cntl->sc_jobid, pjob->ji_qs.ji_jobid);
if ((rc = stat_to_mom(pjob, cntl)) == PBSE_SYSTEM)
{
break;
}
if (rc != 0)
{
rc = 0;
continue;
}
return; /* will pick up after mom replies */
}
} /* END while(1) */
if (cntl->sc_conn >= 0)
svr_disconnect(cntl->sc_conn); /* close connection to MOM */
if (rc != 0)
{
free(cntl);
reply_free(preply);
req_reject(rc, 0, preq, NULL, "cannot get update from mom");
return;
}
} /* END if (!server.sv_attr[(int)SRV_ATR_PollJobs].at_val.at_long) */
/*
* now ready for part 3, building the status reply,
* loop through again
*/
if (type == tjstSummarizeArraysQueue || type == tjstSummarizeArraysServer)
{
update_array_statuses();
}
if (type == tjstJob)
pjob = find_job(preq->rq_ind.rq_status.rq_id);
else if (type == tjstQueue)
pjob = (job *)GET_NEXT(cntl->sc_pque->qu_jobs);
else if (type == tjstSummarizeArraysQueue)
pjob = (job *)GET_NEXT(cntl->sc_pque->qu_jobs_array_sum);
else if (type == tjstSummarizeArraysServer)
pjob = (job *)GET_NEXT(svr_jobs_array_sum);
else if (type == tjstArray)
{
job_array_index = 0;
pjob = NULL;
/* increment job_array_index until we find a non-null pointer or hit the end */
while (job_array_index < pa->ai_qs.array_size && (pjob = pa->jobs[job_array_index]) == NULL)
job_array_index++;
}
else
pjob = (job *)GET_NEXT(svr_alljobs);
DTime = 0;
if (preq->rq_extend != NULL)
{
char *ptr;
/* FORMAT: { EXECQONLY | DELTA:<EPOCHTIME> } */
if (strstr(preq->rq_extend, EXECQUEONLY))
exec_only = 1;
ptr = strstr(preq->rq_extend, "DELTA:");
if (ptr != NULL)
{
ptr += strlen("delta:");
DTime = strtol(ptr, NULL, 10);
}
}
free(cntl);
if ((type == tjstTruncatedServer) || (type == tjstTruncatedQueue))
{
long sentJobCounter;
long qjcounter;
long qmaxreport;
/* loop through all queues */
for (pque = (pbs_queue *)GET_NEXT(svr_queues);
pque != NULL;
pque = (pbs_queue *)GET_NEXT(pque->qu_link))
{
qjcounter = 0;
if ((exec_only == 1) &&
(pque->qu_qs.qu_type != QTYPE_Execution))
{
/* ignore routing queues */
continue;
}
if (((pque->qu_attr[QA_ATR_MaxReport].at_flags & ATR_VFLAG_SET) != 0) &&
(pque->qu_attr[QA_ATR_MaxReport].at_val.at_long >= 0))
{
qmaxreport = pque->qu_attr[QA_ATR_MaxReport].at_val.at_long;
}
else
{
qmaxreport = TMAX_JOB;
}
if (LOGLEVEL >= 5)
{
sprintf(log_buffer,"giving scheduler up to %ld idle jobs in queue %s\n",
qmaxreport,
pque->qu_qs.qu_name);
log_event(
PBSEVENT_SYSTEM,
PBS_EVENTCLASS_QUEUE,
pque->qu_qs.qu_name,
log_buffer);
}
sentJobCounter = 0;
/* loop through jobs in queue */
for (pjob = (job *)GET_NEXT(pque->qu_jobs);
pjob != NULL;
pjob = (job *)GET_NEXT(pjob->ji_jobque))
{
if ((qjcounter >= qmaxreport) &&
(pjob->ji_qs.ji_state == JOB_STATE_QUEUED))
{
/* max_report of queued jobs reached for queue */
continue;
}
pal = (svrattrl *)GET_NEXT(preq->rq_ind.rq_status.rq_attr);
rc = status_job(
pjob,
preq,
(pjob->ji_wattr[(int)JOB_ATR_mtime].at_val.at_long >= DTime) ? pal : dpal,
&preply->brp_un.brp_status,
&bad);
if ((rc != 0) && (rc != PBSE_PERM))
{
req_reject(rc, bad, preq, NULL, NULL);
return;
}
sentJobCounter++;
if (pjob->ji_qs.ji_state == JOB_STATE_QUEUED)
qjcounter++;
} /* END for (pjob) */
if (LOGLEVEL >= 5)
{
sprintf(log_buffer,"sent scheduler %ld total jobs for queue %s\n",
sentJobCounter,
pque->qu_qs.qu_name);
log_event(
PBSEVENT_SYSTEM,
PBS_EVENTCLASS_QUEUE,
pque->qu_qs.qu_name,
log_buffer);
}
} /* END for (pque) */
reply_send(preq);
return;
} /* END if ((type == tjstTruncatedServer) || ...) */
while (pjob != NULL)
{
/* go ahead and build the status reply for this job */
if (exec_only)
{
pque = find_queuebyname(pjob->ji_qs.ji_queue);
if (pque->qu_qs.qu_type != QTYPE_Execution)
goto nextjob;
}
pal = (svrattrl *)GET_NEXT(preq->rq_ind.rq_status.rq_attr);
rc = status_job(
pjob,
preq,
pal,
&preply->brp_un.brp_status,
&bad);
if ((rc != 0) && (rc != PBSE_PERM))
{
req_reject(rc, bad, preq, NULL, NULL);
return;
}
/* get next job */
nextjob:
if (type == tjstJob)
break;
if (type == tjstQueue)
pjob = (job *)GET_NEXT(pjob->ji_jobque);
else if (type == tjstSummarizeArraysQueue)
pjob = (job *)GET_NEXT(pjob->ji_jobque_array_sum);
else if (type == tjstSummarizeArraysServer)
pjob = (job *)GET_NEXT(pjob->ji_jobs_array_sum);
else if (type == tjstArray)
{
pjob = NULL;
/* increment job_array_index until we find a non-null pointer or hit the end */
while (++job_array_index < pa->ai_qs.array_size && (pjob = pa->jobs[job_array_index]) == NULL)
;
}
else
pjob = (job *)GET_NEXT(pjob->ji_alljobs);
rc = 0;
} /* END while (pjob != NULL) */
reply_send(preq);
if (LOGLEVEL >= 7)
{
log_event(PBSEVENT_SYSTEM,
PBS_EVENTCLASS_JOB,
"req_statjob",
"Successfully returned the status of queued jobs\n");
}
return;
} /* END req_stat_job_step2() */
void req_stat_job(
struct batch_request *preq) /* ptr to the decoded request */
{
struct stat_cntl *cntl; /* see svrfunc.h */
char *name;
job *pjob = NULL;
pbs_queue *pque = NULL;
int rc = 0;
enum TJobStatTypeEnum type = tjstNONE;
/*
* first, validate the name of the requested object, either
* a job, a queue, or the whole server.
*/
/* FORMAT: name = { <JOBID> | <QUEUEID> | '' } */
name = preq->rq_ind.rq_status.rq_id;
if (preq->rq_extend != NULL)
{
/* evaluate pbs_job_stat() 'extension' field */
if (!strncasecmp(preq->rq_extend, "truncated", strlen("truncated")))
{
/* truncate response by 'max_report' */
type = tjstTruncatedServer;
}
else if (!strncasecmp(preq->rq_extend, "summarize_arrays", strlen("summarize_arrays")))
{
type = tjstSummarizeArraysServer;
}
} /* END if (preq->rq_extend != NULL) */
if (isdigit((int)*name))
{
/* status a single job */
if (is_array(name))
{
if (type != tjstSummarizeArraysServer)
{
type = tjstArray;
}
pjob = find_array_template(name);
}
else
{
type = tjstJob;
if ((pjob = find_job(name)) == NULL)
{
rc = PBSE_UNKJOBID;
}
}
}
else if (isalpha(name[0]))
{
if (type == tjstNONE)
type = tjstQueue;
else if (type == tjstSummarizeArraysServer)
type = tjstSummarizeArraysQueue;
else
type = tjstTruncatedQueue;
if ((pque = find_queuebyname(name)) == NULL)
{
rc = PBSE_UNKQUE;
}
}
else if ((*name == '\0') || (*name == '@'))
{
/* status all jobs at server */
if (type == tjstNONE)
type = tjstServer;
}
else
{
rc = PBSE_IVALREQ;
}
if (rc != 0)
{
/* is invalid - an error */
req_reject(rc, 0, preq, NULL, NULL);
return;
}
preq->rq_reply.brp_choice = BATCH_REPLY_CHOICE_Status;
CLEAR_HEAD(preq->rq_reply.brp_un.brp_status);
cntl = (struct stat_cntl *)malloc(sizeof(struct stat_cntl));
if (cntl == NULL)
{
req_reject(PBSE_SYSTEM, 0, preq, NULL, NULL);
return;
}
cntl->sc_type = (int)type;
cntl->sc_conn = -1;
cntl->sc_pque = pque;
cntl->sc_origrq = preq;
cntl->sc_post = req_stat_job_step2;
cntl->sc_jobid[0] = '\0'; /* cause "start from beginning" */
if (server.sv_attr[(int)SRV_ATR_PollJobs].at_val.at_long)
cntl->sc_post = 0; /* we're not going to make clients wait */
req_stat_job_step2(cntl); /* go to step 2, see if running is current */
return;
} /* END req_stat_job() */
void req_stat_que(
struct batch_request *preq) /* ptr to the decoded request */
{
char *name;
pbs_queue *pque = NULL;
struct batch_reply *preply;
int rc = 0;
int type = 0;
/*
* first, validate the name of the requested object, either
* a queue, or null for all queues
*/
name = preq->rq_ind.rq_status.rq_id;
if ((*name == '\0') || (*name == '@'))
{
type = 1;
}
else
{
pque = find_queuebyname(name);
if (pque == NULL)
{
req_reject(PBSE_UNKQUE, 0, preq, NULL, "cannot locate queue");
return;
}
}
preply = &preq->rq_reply;
preply->brp_choice = BATCH_REPLY_CHOICE_Status;
CLEAR_HEAD(preply->brp_un.brp_status);
if (type == 0)
{
/* get status of the named queue */
rc = status_que(pque, preq, &preply->brp_un.brp_status);
}
else
{
/* get status of all queues */
pque = (pbs_queue *)GET_NEXT(svr_queues);
while (pque != NULL)
{
rc = status_que(pque, preq, &preply->brp_un.brp_status);
if (rc != 0)
{
if (rc != PBSE_PERM)
break;
rc = 0;
}
pque = (pbs_queue *)GET_NEXT(pque->qu_link);
}
}
if (rc != 0)
{
reply_free(preply);
req_reject(rc, bad, preq, NULL, "status_queue failed");
}
else
{
reply_send(preq);
}
return;
} /* END req_stat_que() */