int issue_to_svr(
char *servern, /* I */
struct batch_request *preq, /* I */
void (*replyfunc) (struct work_task *)) /* I */
{
int do_retry = 0;
int handle;
pbs_net_t svraddr;
char *svrname;
unsigned int port = pbs_server_port_dis;
struct work_task *pwt;
strcpy(preq->rq_host, servern);
preq->rq_fromsvr = 1;
preq->rq_perm = ATR_DFLAG_MGRD | ATR_DFLAG_MGWR | ATR_DFLAG_SvWR;
svrname = parse_servername(servern, &port);
svraddr = get_hostaddr(svrname);
if (svraddr == (pbs_net_t)0)
{
if (pbs_errno == PBS_NET_RC_RETRY)
{
/* Non fatal error - retry */
do_retry = 1;
}
}
else
{
handle = svr_connect(svraddr, port, process_Dreply, ToServerDIS);
if (handle >= 0)
{
return(issue_Drequest(handle, preq, replyfunc, NULL));
}
else if (handle == PBS_NET_RC_RETRY)
{
do_retry = 1;
}
}
/* if reached here, it didn`t go, do we retry? */
if (do_retry)
{
pwt = set_task(
WORK_Timed,
(long)(time_now + PBS_NET_RETRY_TIME),
reissue_to_svr,
(void *)preq);
pwt->wt_parmfunc = replyfunc;
return(0);
}
/* FAILURE */
return(-1);
} /* END issue_to_svr() */
int finalize_rerunjob(
batch_request *preq,
job *pjob,
int rc)
{
int Force;
char log_buf[LOCAL_LOG_BUF_SIZE];
if (pjob == NULL)
return(PBSE_BAD_PARAMETER);
mutex_mgr pjob_mutex = mutex_mgr(pjob->ji_mutex, true);
if (preq->rq_extend && !strncasecmp(preq->rq_extend, RERUNFORCE, strlen(RERUNFORCE)))
Force = 1;
else
Force = 0;
switch (rc)
{
case -1:
/* completed job was requeued */
/* clear out job completion time if there is one */
break;
case 0:
/* requeue request successful */
pjob->ji_qs.ji_substate = JOB_SUBSTATE_RERUN;
break;
case PBSE_SYSTEM: /* This may not be accurate...*/
rc = PBSE_MEM_MALLOC;
snprintf(log_buf, LOCAL_LOG_BUF_SIZE, "Can not allocate memory");
req_reject(rc, 0, preq, NULL, log_buf);
return rc;
break;
default:
if (Force == 0)
{
rc = PBSE_MOMREJECT;
snprintf(log_buf, LOCAL_LOG_BUF_SIZE, "Rejected by mom");
req_reject(rc, 0, preq, NULL, log_buf);
return rc;
}
else
{
int newstate;
int newsubst;
unsigned int dummy;
char *tmp;
long cray_enabled = FALSE;
get_svr_attr_l(SRV_ATR_CrayEnabled, &cray_enabled);
if ((cray_enabled == TRUE) &&
(pjob->ji_wattr[JOB_ATR_login_node_id].at_val.at_str != NULL))
tmp = parse_servername(pjob->ji_wattr[JOB_ATR_login_node_id].at_val.at_str, &dummy);
else
tmp = parse_servername(pjob->ji_wattr[JOB_ATR_exec_host].at_val.at_str, &dummy);
/* Cannot communicate with MOM, forcibly requeue job.
This is a relatively disgusting thing to do */
sprintf(log_buf, "rerun req to %s failed (rc=%d), forcibly requeueing job",
tmp, rc);
free(tmp);
log_event(
PBSEVENT_ERROR | PBSEVENT_ADMIN | PBSEVENT_JOB,
PBS_EVENTCLASS_JOB,
pjob->ji_qs.ji_jobid,
log_buf);
log_err(-1, __func__, log_buf);
strcat(log_buf, ", previous output files may be lost");
svr_mailowner(pjob, MAIL_OTHER, MAIL_FORCE, log_buf);
svr_setjobstate(pjob, JOB_STATE_EXITING, JOB_SUBSTATE_RERUN3, FALSE);
rel_resc(pjob); /* free resc assigned to job */
pjob->ji_modified = 1; /* force full job save */
pjob->ji_momhandle = -1;
pjob->ji_qs.ji_svrflags &= ~JOB_SVFLG_StagedIn;
svr_evaljobstate(*pjob, newstate, newsubst, 0);
svr_setjobstate(pjob, newstate, newsubst, FALSE);
}
break;
} /* END switch (rc) */
pjob->ji_qs.ji_svrflags = (pjob->ji_qs.ji_svrflags &
~(JOB_SVFLG_CHECKPOINT_FILE |JOB_SVFLG_CHECKPOINT_MIGRATEABLE |
JOB_SVFLG_CHECKPOINT_COPIED)) | JOB_SVFLG_HASRUN;
sprintf(log_buf, msg_manager, msg_jobrerun, preq->rq_user, preq->rq_host);
log_event(PBSEVENT_JOB,PBS_EVENTCLASS_JOB,pjob->ji_qs.ji_jobid,log_buf);
reply_ack(preq);
/* note in accounting file */
account_record(PBS_ACCT_RERUN, pjob, NULL);
return rc;
} /* END req_rerunjob() */