void
checkFinish (void)
{
struct jobCard *jobCard, *nextJob;
for (jobCard = jobQueHead->forw; (jobCard != jobQueHead);
jobCard = nextJob) {
nextJob = jobCard->forw;
if (!(IS_FINISH(jobCard->jobSpecs.jStatus))
&& !(IS_POST_FINISH(jobCard->jobSpecs.jStatus) ) ) {
if ( (jobsig(jobCard, 0, FALSE) < 0)
|| ( (jobCard->jobSpecs.jAttrib & JOB_FORCE_KILL)
&& (jobCard->jobSpecs.termTime
< time(0)-MAX(6,jobTerminateInterval*3)) ) ) {
jobGone (jobCard);
}
}
if (jobCard->jobSpecs.actPid) {
if (killpg(jobCard->jobSpecs.actPid, SIGCONT) == 0)
continue;
if (kill(jobCard->jobSpecs.actPid, SIGCONT) == 0)
continue;
if (jobCard->cleanupPid > 0 &&
kill(jobCard->cleanupPid, SIGCONT) == 0)
continue;
sigActEnd(jobCard);
continue;
}
if (IS_FINISH(jobCard->jobSpecs.jStatus)
|| IS_POST_FINISH(jobCard->jobSpecs.jStatus)
|| (jobCard->jobSpecs.jStatus & JOB_STAT_PEND)) {
job_finish (jobCard, TRUE);
}
}
}
int
main (int argc, char **argv, char **environ)
{
char *queue = NULL, *host = NULL, *jobName = NULL, *user = NULL;
LS_LONG_INT jobId;
int options;
struct jobInfoEnt *jInfo;
char *outFile;
char fflag = FALSE;
int cc;
int rc;
rc = _i18n_init (I18N_CAT_MIN);
if (lsb_init (argv[0]) < 0)
{
lsb_perror ("lsb_init");
exit (-1);
}
while ((cc = getopt (argc, argv, "Vhfq:m:J:")) != EOF)
{
switch (cc)
{
case 'q':
if (queue || host || jobName)
oneOf (argv[0]);
queue = optarg;
break;
case 'm':
if (queue || host || jobName)
oneOf (argv[0]);
host = optarg;
break;
case 'J':
if (queue || host || jobName)
oneOf (argv[0]);
jobName = optarg;
break;
case 'V':
fputs (_LS_VERSION_, stderr);
exit (0);
case 'f':
fflag = TRUE;
break;
case 'h':
default:
usage (argv[0]);
}
}
jobId = 0;
options = LAST_JOB;
if (argc >= optind + 1)
{
if (queue || host || jobName)
{
oneOf (argv[0]);
}
else if ((argc > 2 && !fflag) || (argc > 3 && fflag))
usage (argv[0]);
if (getOneJobId (argv[optind], &jobId, 0))
{
usage (argv[0]);
}
options = 0;
}
if (lsb_openjobinfo (jobId, jobName, NULL, queue, host, options) < 0
|| (jInfo = lsb_readjobinfo (NULL)) == NULL)
{
if (jobId != 0 || jobName != NULL)
{
user = ALL_USERS;
if (lsb_openjobinfo (jobId, jobName, user, queue, host, options) < 0
|| (jInfo = lsb_readjobinfo (NULL)) == NULL)
{
jobInfoErr (jobId, jobName, NULL, queue, host, options);
exit (-1);
}
}
else
{
jobInfoErr (jobId, jobName, NULL, queue, host, options);
exit (-1);
}
}
lsb_closejobinfo ();
if (jobId && jInfo->jobId != jobId)
{
lsberrno = LSBE_JOB_ARRAY;
lsb_perror ("bpeek");
exit (-1);
}
if ((jInfo->submit.options & SUB_INTERACTIVE) &&
!(jInfo->submit.options & (SUB_OUT_FILE | SUB_ERR_FILE)))
{
fprintf (stderr, _i18n_msg_get (ls_catd, NL_SETN, 2456, "Job <%s> : Cannot bpeek an interactive job.\n"), /* catgets 2456 */
lsb_jobid2str (jInfo->jobId));
exit (-1);
}
if (IS_PEND (jInfo->status) || jInfo->execUsername[0] == '\0')
{
fprintf (stderr, _i18n_msg_get (ls_catd, NL_SETN, 2454, "Job <%s> : Not yet started.\n"), /* catgets 2454 */
lsb_jobid2str (jInfo->jobId));
exit (-1);
}
if (IS_FINISH (jInfo->status))
{
fprintf (stderr, _i18n_msg_get (ls_catd, NL_SETN, 2455, "Job <%s> : Already finished.\n"), /* catgets 2455 */
lsb_jobid2str (jInfo->jobId));
exit (-1);
}
if ((outFile = lsb_peekjob (jInfo->jobId)) == NULL)
{
char msg[50];
sprintf (msg, "%s <%s>", I18N_Job, lsb_jobid2str (jInfo->jobId));
lsb_perror (msg);
exit (-1);
}
displayOutput (outFile, jInfo, fflag, environ);
_i18n_end (ls_catd);
exit (0);
}
void
prtJobRusage(struct jobInfoEnt *job)
{
char prline[MAXLINELEN];
int i, j;
int linepos;
if (IS_FINISH(job->status))
return;
if (IS_PEND(job->status)) {
if (job->runRusage.utime || job->runRusage.stime) {
if (uf_format)
printf ("%s: Resource usage collected. The CPU time used is %d seconds.",
_i18n_ctime(ls_catd, CTIME_FORMAT_a_b_d_T, &job->jRusageUpdateTime),
job->runRusage.utime + job->runRusage.stime);
else {
sprintf(prline, "%s: %s.\n",
_i18n_ctime(ls_catd, CTIME_FORMAT_a_b_d_T,
&job->jRusageUpdateTime),
I18N(644, "Resource usage collected")); /* catgets 644 */
prtLine(prline);
sprintf(prline, (_i18n_msg_get(ls_catd,NL_SETN,645, " The CPU time used is %d seconds.\n")), /* catgets 645 */
job->runRusage.utime + job->runRusage.stime);
prtLine(prline);
}
}
return;
};
if (job->runRusage.utime > 0 || job->runRusage.stime > 0
|| job->runRusage.mem > 0 || job->runRusage.swap > 0
|| job->runRusage.npgids > 0 || job->runRusage.npids > 0) {
if (uf_format)
printf ("%s: Resource usage collected.",
_i18n_ctime(ls_catd, CTIME_FORMAT_a_b_d_T, &job->jRusageUpdateTime));
else {
sprintf(prline, "%s: %s.\n",
_i18n_ctime(ls_catd, CTIME_FORMAT_a_b_d_T,
&job->jRusageUpdateTime),
I18N(646, "Resource usage collected")); /* catgets 646 */
prtLine(prline);
}
} else
return;
if (job->runRusage.utime > 0 || job->runRusage.stime > 0) {
if (uf_format)
printf (" The CPU time used is %d seconds.",
job->runRusage.utime + job->runRusage.stime);
else {
sprintf(prline, (_i18n_msg_get(ls_catd,NL_SETN,647, " The CPU time used is %d seconds.\n")), /* catgets 647 */
job->runRusage.utime + job->runRusage.stime);
prtLine(prline);
}
}
if (job->runRusage.mem > 0) {
if (uf_format) {
if (job->runRusage.mem > 1024)
printf(" MEM: %d Mbytes;", job->runRusage.mem/1024);
else
printf(" MEM: %d Kbytes;", job->runRusage.mem);
}
else {
if (job->runRusage.mem > 1024)
sprintf(prline, (_i18n_msg_get(ls_catd,NL_SETN,648, " MEM: %d Mbytes")), job->runRusage.mem/1024); /* catgets 648 */
else
sprintf(prline, (_i18n_msg_get(ls_catd,NL_SETN,649, " MEM: %d Kbytes")), job->runRusage.mem); /* catgets 649 */
prtLine(prline);
}
}
if (job->runRusage.swap > 0) {
char *space;
if (job->runRusage.mem > 0)
space = "; ";
else
space = " ";
if (uf_format) {
if (job->runRusage.swap > 1024)
printf(" SWAP: %d Mbytes;", job->runRusage.swap/1024);
else
printf(" SWAP: %d Kbytes;", job->runRusage.swap);
}
else {
if (job->runRusage.swap > 1024)
sprintf(prline, (_i18n_msg_get(ls_catd,NL_SETN,650, "%sSWAP: %d Mbytes\n")), space, /* catgets 650 */
job->runRusage.swap/1024);
else
sprintf(prline, (_i18n_msg_get(ls_catd,NL_SETN,651, "%sSWAP: %d Kbytes\n")), space, job->runRusage.swap); /* catgets 651 */
prtLine(prline);
}
} else {
if (job->runRusage.mem > 0 && !uf_format) {
sprintf(prline, "\n");
prtLine(prline);
}
}
if (job->runRusage.npgids <= 0)
return;
for (i=0; i < job->runRusage.npgids; i++) {
if (uf_format)
printf (" PGID: %d; ", job->runRusage.pgid[i]);
else {
sprintf(prline, (_i18n_msg_get(ls_catd,NL_SETN,652, " PGID: %d; ")), job->runRusage.pgid[i]); /* catgets 652 */
linepos = strlen(prline);
prtLine(prline);
}
sprintf(prline, (_i18n_msg_get(ls_catd,NL_SETN,653, "PIDs: "))); /* catgets 653 */
linepos += 6;
prtLineWUF(prline);
for (j=0; j < job->runRusage.npids; j++) {
if (job->runRusage.pgid[i] == job->runRusage.pidInfo[j].pgid) {
sprintf(prline, "%d ", job->runRusage.pidInfo[j].pid);
if (uf_format)
printf ("%d%s", job->runRusage.pidInfo[j].pid, j==job->runRusage.npids-1?"":" ");
else {
linepos += strlen(prline);
if (linepos >= 80) {
char *newline ="\n ";
prtLine(newline);
prtLine(prline);
linepos = strlen(prline) + 21;
}
else
prtLine(prline);
}
}
}
if (uf_format)
printf(";");
else {
sprintf(prline, "\n");
prtLine(prline);
}
}
sprintf(prline, "\n");
prtLineWUF(prline);
if (uf_format && job->runRusage.mem > 0) {
printf ("\n MEMORY USAGE:\n");
printf (" MAX MEM: N/A MBytes; AVG MEM: N/A MBytes\n");
}
}
int
status_job (mbdReqType reqType,
struct jobCard *jp, int newStatus, sbdReplyType err)
{
static char fname[] = "status_job()";
static int seq = 1;
static char lastHost[MAXHOSTNAMELEN];
int reply;
char *request_buf;
char *reply_buf = NULL;
XDR xdrs;
struct LSFHeader hdr;
int cc;
struct statusReq statusReq;
int flags;
int i;
int len;
struct lsfAuth *auth = NULL;
if ((logclass & LC_TRACE) && (logclass & LC_SIGNAL))
ls_syslog (LOG_DEBUG, "%s: Entering ... regType %d jobId %s",
fname, reqType, lsb_jobid2str (jp->jobSpecs.jobId));
if (newStatus == JOB_STAT_EXIT)
{
jp->userJobSucc = FALSE;
}
if (MASK_STATUS (newStatus) == JOB_STAT_DONE)
{
jp->userJobSucc = TRUE;
}
if (IS_POST_FINISH (newStatus))
{
if (jp->userJobSucc != TRUE)
{
return 0;
}
}
if (masterHost == NULL)
return -1;
if (jp->notReported < 0)
{
jp->notReported = -INFINIT_INT;
return (0);
}
statusReq.jobId = jp->jobSpecs.jobId;
statusReq.actPid = jp->jobSpecs.actPid;
statusReq.jobPid = jp->jobSpecs.jobPid;
statusReq.jobPGid = jp->jobSpecs.jobPGid;
statusReq.newStatus = newStatus;
statusReq.reason = jp->jobSpecs.reasons;
statusReq.subreasons = jp->jobSpecs.subreasons;
statusReq.sbdReply = err;
statusReq.lsfRusage = jp->lsfRusage;
statusReq.execUid = jp->jobSpecs.execUid;
statusReq.numExecHosts = 0;
statusReq.execHosts = NULL;
statusReq.exitStatus = jp->w_status;
statusReq.execCwd = jp->jobSpecs.execCwd;
statusReq.execHome = jp->jobSpecs.execHome;
statusReq.execUsername = jp->execUsername;
statusReq.queuePostCmd = "";
statusReq.queuePreCmd = "";
statusReq.msgId = jp->delieveredMsgId;
if (IS_FINISH (newStatus))
{
if (jp->maxRusage.mem > jp->runRusage.mem)
jp->runRusage.mem = jp->maxRusage.mem;
if (jp->maxRusage.swap > jp->runRusage.swap)
jp->runRusage.swap = jp->maxRusage.swap;
if (jp->maxRusage.stime > jp->runRusage.stime)
jp->runRusage.stime = jp->maxRusage.stime;
if (jp->maxRusage.utime > jp->runRusage.utime)
jp->runRusage.utime = jp->maxRusage.utime;
}
statusReq.runRusage.mem = jp->runRusage.mem;
statusReq.runRusage.swap = jp->runRusage.swap;
statusReq.runRusage.utime = jp->runRusage.utime;
statusReq.runRusage.stime = jp->runRusage.stime;
statusReq.runRusage.npids = jp->runRusage.npids;
statusReq.runRusage.pidInfo = jp->runRusage.pidInfo;
statusReq.runRusage.npgids = jp->runRusage.npgids;
statusReq.runRusage.pgid = jp->runRusage.pgid;
statusReq.actStatus = jp->actStatus;
statusReq.sigValue = jp->jobSpecs.actValue;
statusReq.seq = seq;
seq++;
if (seq >= MAX_SEQ_NUM)
seq = 1;
len = 1024 + ALIGNWORD_ (sizeof (struct statusReq));
len += ALIGNWORD_ (strlen (statusReq.execHome)) + 4 +
ALIGNWORD_ (strlen (statusReq.execCwd)) + 4 +
ALIGNWORD_ (strlen (statusReq.execUsername)) + 4;
for (i = 0; i < statusReq.runRusage.npids; i++)
len += ALIGNWORD_ (sizeof (struct pidInfo)) + 4;
for (i = 0; i < statusReq.runRusage.npgids; i++)
len += ALIGNWORD_ (sizeof (int)) + 4;
if (logclass & (LC_TRACE | LC_COMM))
ls_syslog (LOG_DEBUG, "%s: The length of the job message is: <%d>", fname,
len);
if ((request_buf = malloc (len)) == NULL)
{
ls_syslog (LOG_ERR, I18N_FUNC_FAIL_M, fname, "malloc");
return (-1);
}
xdrmem_create (&xdrs, request_buf, len, XDR_ENCODE);
initLSFHeader_ (&hdr);
hdr.opCode = reqType;
if (!xdr_encodeMsg (&xdrs, (char *) &statusReq, &hdr, xdr_statusReq, 0,
auth))
{
ls_syslog (LOG_ERR, I18N_JOB_FAIL_S_M,
fname, lsb_jobid2str (jp->jobSpecs.jobId), "xdr_statusReq");
lsb_merr2 (I18N_FUNC_FAIL, fname, "xdr_statusReq");
xdr_destroy (&xdrs);
FREEUP (request_buf);
relife ();
}
flags = CALL_SERVER_NO_HANDSHAKE;
if (statusChan >= 0)
flags |= CALL_SERVER_USE_SOCKET;
if (reqType == BATCH_RUSAGE_JOB)
flags |= CALL_SERVER_NO_WAIT_REPLY;
if (logclass & LC_COMM)
ls_syslog (LOG_DEBUG1, "%s: before call_server statusChan=%d flags=%d",
fname, statusChan, flags);
cc = call_server (masterHost,
mbd_port,
request_buf,
XDR_GETPOS (&xdrs),
&reply_buf,
&hdr,
connTimeout, readTimeout, &statusChan, NULL, NULL, flags);
if (cc < 0)
{
statusChan = -1;
if (!equalHost_ (masterHost, lastHost))
{
if (errno != EINTR)
ls_syslog (LOG_DEBUG,
"%s: Failed to reach mbatchd on host <%s> for job <%s>: %s",
fname, masterHost, lsb_jobid2str (jp->jobSpecs.jobId),
lsb_sysmsg ());
strcpy (lastHost, masterHost);
}
xdr_destroy (&xdrs);
FREEUP (request_buf);
failcnt++;
return (-1);
}
else if (cc == 0)
{
}
failcnt = 0;
lastHost[0] = '\0';
xdr_destroy (&xdrs);
FREEUP (request_buf);
if (cc)
free (reply_buf);
if (flags & CALL_SERVER_NO_WAIT_REPLY)
{
struct timeval timeval;
timeval.tv_sec = 0;
timeval.tv_usec = 0;
if (rd_select_ (chanSock_ (statusChan), &timeval) == 0)
{
jp->needReportRU = FALSE;
jp->lastStatusMbdTime = now;
return 0;
}
CLOSECD (statusChan);
if (logclass & LC_COMM)
ls_syslog (LOG_DEBUG1,
"%s: Job <%s> rd_select() failed, assume connection broken",
fname, lsb_jobid2str (jp->jobSpecs.jobId));
return (-1);
}
reply = hdr.opCode;
switch (reply)
{
case LSBE_NO_ERROR:
case LSBE_LOCK_JOB:
jp->needReportRU = FALSE;
jp->lastStatusMbdTime = now;
if (reply == LSBE_LOCK_JOB)
{
if (IS_SUSP (jp->jobSpecs.jStatus))
jp->jobSpecs.reasons |= SUSP_MBD_LOCK;
else
ls_syslog (LOG_ERR, _i18n_msg_get (ls_catd, NL_SETN, 5204, "%s: Job <%s> is in status <%x> and mbatchd wants to lock it, ignored."), /* catgets 5204 */
fname,
lsb_jobid2str (jp->jobSpecs.jobId),
jp->jobSpecs.jStatus);
}
return (0);
case LSBE_NO_JOB:
if (!IS_POST_FINISH (jp->jobSpecs.jStatus))
{
ls_syslog (LOG_ERR, _i18n_msg_get (ls_catd, NL_SETN, 5205, "%s: Job <%s> is forgotten by mbatchd on host <%s>, ignored."), fname, lsb_jobid2str (jp->jobSpecs.jobId), masterHost); /* catgets 5205 */
}
jp->notReported = -INFINIT_INT;
return (0);
case LSBE_STOP_JOB:
if (jobsig (jp, SIGSTOP, TRUE) < 0)
SET_STATE (jp->jobSpecs.jStatus, JOB_STAT_EXIT);
else
{
SET_STATE (jp->jobSpecs.jStatus, JOB_STAT_USUSP);
jp->jobSpecs.reasons |= SUSP_USER_STOP;
}
return (-1);
case LSBE_SBATCHD:
ls_syslog (LOG_ERR, _i18n_msg_get (ls_catd, NL_SETN, 5206, "%s: mbatchd on host <%s> doesn't think I'm configured as a batch server when I report the status for job <%s>"), /* catgets 5206 */
fname, masterHost, lsb_jobid2str (jp->jobSpecs.jobId));
return (-1);
default:
ls_syslog (LOG_ERR, _i18n_msg_get (ls_catd, NL_SETN, 5207, "%s: Illegal reply code <%d> from mbatchd on host <%s> for job <%s>"), /* catgets 5207 */
fname,
reply, masterHost, lsb_jobid2str (jp->jobSpecs.jobId));
return (-1);
}
}
void
job_checking (void)
{
static char fname[] = "job_checking";
struct jobCard *jobCard, *nextJob;
struct hostLoad *myload, savedLoad;
char *myhostnm;
static time_t last_check;
char preempted = FALSE;
int i;
if (last_check == 0)
last_check = now;
if (jobcnt <= 0) {
last_check = now;
return;
}
checkFinish ();
for (jobCard = jobQueHead->forw; (jobCard != jobQueHead);
jobCard = nextJob) {
nextJob = jobCard->forw;
if (IS_FINISH(jobCard->jobSpecs.jStatus)
|| (jobCard->jobSpecs.jStatus & JOB_STAT_PEND))
continue;
ruLimits(jobCard);
if (IS_RUN_JOB_CMD(jobCard->jobSpecs.jStatus)) {
jobCard->runTime += (int) (now - last_check);
}
if (jobCard->runTime >
jobCard->jobSpecs.lsfLimits[LSF_RLIMIT_RUN].rlim_curl) {
if ((jobCard->jobSpecs.terminateActCmd == NULL)
|| (jobCard->jobSpecs.terminateActCmd[0] == '\0')) {
if (jobCard->runTime >
jobCard->jobSpecs.lsfLimits[LSF_RLIMIT_RUN].rlim_curl
+ WARN_TIME && jobCard->timeExpire) {
if ((IS_SUSP (jobCard->jobSpecs.jStatus))
&& (jobCard->jobSpecs.reasons & SUSP_RES_LIMIT)
&& (jobCard->jobSpecs.subreasons & SUB_REASON_RUNLIMIT))
continue;
else if (jobCard->jobSpecs.jStatus & JOB_STAT_KILL)
continue;
else {
ls_syslog(LOG_INFO, \
"%s: warning period expired killing the job=%d",
fname, jobCard->jobSpecs.jobId);
jobSigStart (jobCard, SIG_TERM_RUNLIMIT, 0, 0, SIGLOG);
sbdlog_newstatus(jobCard);
jobCard->jobSpecs.jStatus |= JOB_STAT_KILL;
}
} else if (!jobCard->timeExpire) {
ls_syslog(LOG_INFO, I18N(5704,
"%s: sending warning signal to job=%d"), /* catgets 5704 */
fname, jobCard->jobSpecs.jobId);
jobsig(jobCard, SIGUSR2, FALSE);
jobCard->timeExpire = TRUE;
}
} else {
if (jobCard->runTime >
jobCard->jobSpecs.lsfLimits[LSF_RLIMIT_RUN].rlim_curl) {
if ((IS_SUSP (jobCard->jobSpecs.jStatus))
&& (jobCard->jobSpecs.reasons & SUSP_RES_LIMIT)
&& (jobCard->jobSpecs.subreasons & SUB_REASON_RUNLIMIT))
continue;
else {
jobSigStart (jobCard, SIG_TERM_RUNLIMIT, 0, 0, SIGLOG);
sbdlog_newstatus(jobCard);
}
}
}
continue;
}
if (jobCard->jobSpecs.termTime && now > jobCard->jobSpecs.termTime
&& !(jobCard->jobSpecs.jAttrib & JOB_FORCE_KILL)) {
if ((jobCard->jobSpecs.terminateActCmd == NULL)
|| (jobCard->jobSpecs.terminateActCmd[0] == '\0')) {
if (now > jobCard->jobSpecs.termTime + WARN_TIME
&& jobCard->timeExpire) {
if ((IS_SUSP (jobCard->jobSpecs.jStatus))
&& (jobCard->jobSpecs.reasons & SUSP_RES_LIMIT)
&& (jobCard->jobSpecs.subreasons & SUB_REASON_DEADLINE))
continue;
else if (jobCard->jobSpecs.jStatus & JOB_STAT_KILL)
continue;
else {
jobSigStart (jobCard, SIG_TERM_DEADLINE, 0, 0, SIGLOG);
sbdlog_newstatus(jobCard);
jobCard->jobSpecs.jStatus |= JOB_STAT_KILL;
}
} else
if (!jobCard->timeExpire) {
jobsig(jobCard, SIGUSR2, FALSE);
jobCard->timeExpire = TRUE;
}
} else {
if (now > jobCard->jobSpecs.termTime) {
if ((IS_SUSP (jobCard->jobSpecs.jStatus))
&& (jobCard->jobSpecs.reasons & SUSP_RES_LIMIT)
&& (jobCard->jobSpecs.subreasons & SUB_REASON_DEADLINE))
continue;
else {
jobSigStart (jobCard, SIG_TERM_DEADLINE, 0, 0, SIGLOG);
sbdlog_newstatus(jobCard);
}
}
}
continue;
}
if (! window_ok (jobCard)
&& !(jobCard->jobSpecs.jAttrib & JOB_URGENT_NOSTOP)) {
if (! (jobCard->jobSpecs.options & SUB_WINDOW_SIG)
|| ((jobCard->jobSpecs.options & SUB_WINDOW_SIG)
&& now - jobCard->windWarnTime >= WARN_TIME)) {
jobSuspendAction(jobCard, SIG_SUSP_WINDOW, SUSP_QUEUE_WINDOW, 0);
continue;
}
} else {
jobResumeAction(jobCard, SIG_RESUME_WINDOW, SUSP_QUEUE_WINDOW);
continue;
}
}
if ((myhostnm = ls_getmyhostname()) == NULL) {
ls_syslog(LOG_ERR, I18N_FUNC_FAIL_MM, fname, "ls_getmyhostname");
die(SLAVE_FATAL);
}
myload = ls_loadofhosts (NULL, 0, EXACT|EFFECTIVE, 0, &myhostnm, 1);
if (myload == NULL) {
if (myStatus != NO_LIM)
ls_syslog(LOG_INFO, I18N_FUNC_FAIL_MM, fname, "ls_loadofhosts");
if (lserrno == LSE_LIM_BADHOST)
relife();
if (lserrno == LSE_BAD_XDR)
relife();
if (lserrno == LSE_LIM_DOWN || lserrno == LSE_TIME_OUT) {
myStatus |= NO_LIM;
tryChkpntMig();
}
last_check = now;
return;
} else
myStatus = 0;
memcpy ((char *)&savedLoad, (char *)myload, sizeof (struct hostLoad));
savedLoad.li = (float *) my_malloc (allLsInfo->numIndx * sizeof (float),
"job_checking");
savedLoad.status = (int *) my_malloc
((1 + GET_INTNUM(allLsInfo->numIndx)) * sizeof (int), "job_checking");
for (i = 0; i < allLsInfo->numIndx; i++)
savedLoad.li[i] = myload->li[i];
for (i = 0; i < 1 + GET_INTNUM(allLsInfo->numIndx); i++)
savedLoad.status[i] = myload->status[i];
tryResume (&savedLoad);
if (!preempted)
tryStop (myhostnm, &savedLoad);
tryChkpntMig();
FREEUP(savedLoad.li);
FREEUP(savedLoad.status);
last_check = now;
return;
}
void
do_sigjob(XDR * xdrs, int chfd, struct LSFHeader * reqHdr)
{
static char fname[] = "do_sigjob()";
char reply_buf[MSGSIZE];
XDR xdrs2;
struct jobSig jobSig;
sbdReplyType reply;
struct jobReply jobReply;
struct LSFHeader replyHdr;
char *replyStruct;
struct jobCard *jp = NULL;
char found = FALSE;
int cc;
int sigValue;
int savedActReasons;
int savedActSubReasons;
struct lsfAuth *auth = NULL;
memset(&jobReply, 0, sizeof(struct jobReply));
if (!xdr_jobSig(xdrs, &jobSig, reqHdr)) {
reply = ERR_BAD_REQ;
ls_syslog(LOG_ERR, I18N_FUNC_FAIL, fname, "xdr_jobSig");
goto Reply1;
}
jobSig.sigValue = sig_decode(jobSig.sigValue);
sigValue = jobSig.sigValue;
if (logclass & LC_SIGNAL)
ls_syslog(LOG_DEBUG, "do_sigJob: sigValue =%d", sigValue);
for (jp = jobQueHead->forw; (jp != jobQueHead); jp = jp->forw) {
if (jp->jobSpecs.jobId != jobSig.jobId)
continue;
found = TRUE;
break;
}
if (found == FALSE) {
reply = ERR_NO_JOB;
jp = NULL;
goto Reply1;
}
if (jobSig.reasons & SUSP_MBD_LOCK) {
jp->jobSpecs.reasons = jobSig.reasons;
jp->jobSpecs.subreasons = jobSig.subReasons;
savedActReasons = jp->actReasons;
savedActSubReasons = jp->actSubReasons;
jp->actReasons = jobSig.reasons;
jp->actSubReasons = jobSig.subReasons;
}
if (jp->postJobStarted) {
reply = ERR_NO_ERROR;
goto Reply1;
}
if (IS_FINISH(jp->jobSpecs.jStatus)) {
reply = ERR_NO_ERROR;
goto Reply1;
}
if (jp->jobSpecs.jobPGid == -1) {
SBD_SET_STATE(jp, JOB_STAT_EXIT);
reply = ERR_NO_ERROR;
goto Reply;
}
if (!JOB_STARTED(jp)) {
if (isSigTerm(sigValue) == TRUE) {
if ((cc = jobSigStart (jp, sigValue, jobSig.actFlags, jobSig.chkPeriod, NO_SIGLOG)) < 0)
reply = ERR_SIG_RETRY;
else
reply = ERR_NO_ERROR;
goto Reply;
}
reply = ERR_SIG_RETRY;
if (logclass & LC_EXEC)
ls_syslog(LOG_DEBUG, "%s: Retry signal %s for job <%s>",
fname, getLsbSigSymbol(sigValue),
lsb_jobid2str(jp->jobSpecs.jobId));
goto Reply1;
}
if (IS_PEND(jp->jobSpecs.jStatus)) {
reply = ERR_SIG_RETRY;
goto Reply1;
}
if (jp->jobSpecs.actPid || (jp->jobSpecs.jStatus & JOB_STAT_MIG)) {
if ((cc = jobSigStart(jp,
sigValue,
jobSig.actFlags,
jobSig.chkPeriod,
NO_SIGLOG)) < 0)
reply = ERR_SIG_RETRY;
else {
jp->jobSpecs.jStatus &= ~JOB_STAT_MIG;
reply = ERR_NO_ERROR;
}
goto Reply;
}
if ((cc = jobSigStart(jp,
sigValue,
jobSig.actFlags,
jobSig.chkPeriod,
NO_SIGLOG)) < 0)
reply = ERR_SIG_RETRY;
else
reply = ERR_NO_ERROR;
Reply:
sbdlog_newstatus(jp);
Reply1:
xdrmem_create(&xdrs2, reply_buf, MSGSIZE, XDR_ENCODE);
initLSFHeader_(&replyHdr);
replyHdr.opCode = reply;
if (reply == ERR_NO_ERROR) {
jobReply.jobPid = jp->jobSpecs.jobPid;
jobReply.actPid = jp->jobSpecs.actPid;
jobReply.jobId = jp->jobSpecs.jobId;
jobReply.jobPGid = jp->jobSpecs.jobPGid;
jobReply.jStatus = jp->jobSpecs.jStatus;
jobReply.reasons = jp->jobSpecs.reasons;
jobReply.actStatus = jp->actStatus;
replyStruct = (char *) &jobReply;
} else {
if (reply != ERR_NO_JOB)
if ((jp != NULL) && (jobSig.reasons & SUSP_MBD_LOCK)) {
jp->actReasons = savedActReasons;
jp->actSubReasons = savedActSubReasons;
}
replyStruct = (char *) 0;
}
if (!xdr_encodeMsg(&xdrs2, replyStruct, &replyHdr, xdr_jobReply, 0, auth)) {
ls_syslog(LOG_ERR, I18N_JOB_FAIL_S_M, fname,
lsb_jobid2str(jp->jobSpecs.jobId), "xdr_jobReply");
relife();
}
if (chanWrite_(chfd, reply_buf, XDR_GETPOS(&xdrs2)) <= 0) {
ls_syslog(LOG_ERR, _i18n_msg_get(ls_catd , NL_SETN, 5821,
"%s: Sending jobReply (len=%d) to master failed for job <%s>: %m"), fname, XDR_GETPOS(&xdrs2), lsb_jobid2str(jobSig.jobId)); /* catgets 5821 */
}
if (jp != NULL)
jp->actStatus = ACT_NO;
xdr_destroy(&xdrs2);
return;
}
