void
do_reboot(XDR * xdrs, int chfd, struct LSFHeader * reqHdr)
{
static char fname[] = "do_reboot()";
char reply_buf[MSGSIZE / 8];
XDR xdrs2;
sbdReplyType reply;
struct LSFHeader replyHdr;
if (logclass & LC_TRACE)
ls_syslog(LOG_DEBUG, "%s: Entering this routine...", fname);
reply = ERR_NO_ERROR;
xdrmem_create(&xdrs2, reply_buf, MSGSIZE / 8, XDR_ENCODE);
initLSFHeader_(&replyHdr);
if (reqHdr->opCode == CMD_SBD_REBOOT)
replyHdr.opCode = LSBE_NO_ERROR;
else
replyHdr.opCode = reply;
if (!xdr_encodeMsg(&xdrs2, (char *) 0, &replyHdr, 0, 0, NULL)) {
ls_syslog(LOG_ERR, I18N_FUNC_FAIL, fname, "xdr_encodeMsg");
xdr_destroy(&xdrs2);
relife();
return;
}
if (chanWrite_(chfd, reply_buf, XDR_GETPOS(&xdrs2)) <= 0) {
ls_syslog(LOG_ERR, I18N_FUNC_FAIL_M, fname, "chanWrite_");
xdr_destroy(&xdrs2);
relife();
return;
}
ls_syslog(LOG_NOTICE, _i18n_msg_get(ls_catd , NL_SETN, 5828,
"Slave batch daemon reboot command received")); /* catgets 5828 */
relife();
ls_syslog(LOG_ERR, _i18n_msg_get(ls_catd , NL_SETN, 5829,
"Unable to relife during rebooting: %m")); /* catgets 5829 */
xdr_destroy(&xdrs2);
}
void
do_shutdown(XDR * xdrs, int chfd, struct LSFHeader * reqHdr)
{
static char fname[] = "do_shutdown()";
char reply_buf[MSGSIZE / 8];
XDR xdrs2;
sbdReplyType reply;
struct LSFHeader replyHdr;
reply = ERR_NO_ERROR;
xdrmem_create(&xdrs2, reply_buf, MSGSIZE / 8, XDR_ENCODE);
initLSFHeader_(&replyHdr);
if (reqHdr->opCode == CMD_SBD_SHUTDOWN)
replyHdr.opCode = LSBE_NO_ERROR;
else
replyHdr.opCode = reply;
if (!xdr_encodeMsg(&xdrs2, (char *) 0, &replyHdr, 0, 0, NULL)) {
ls_syslog(LOG_ERR, I18N_FUNC_FAIL, fname, "xdr_encodeMsg");
xdr_destroy(&xdrs2);
relife();
return;
}
if (chanWrite_(chfd, reply_buf, XDR_GETPOS(&xdrs2)) <= 0) {
ls_syslog(LOG_ERR, _i18n_msg_get(ls_catd , NL_SETN, 5835,
"%s: Sending shutdown reply to master failed: %m"), /* catgets 5835 */
fname);
xdr_destroy(&xdrs2);
relife();
return;
}
xdr_destroy(&xdrs2);
ls_syslog(LOG_NOTICE, _i18n_msg_get(ls_catd , NL_SETN, 5836,
"Slave batch daemon shutdown command received")); /* catgets 5836 */
die(SLAVE_SHUTDOWN);
}
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;
}
static void
tryStop (char *myhostnm, struct hostLoad *myload)
{
static char fname[] = "tryStop";
struct jobCard *jobCard, *next;
int reasons, subreasons, stopmore = FALSE;
static int errCount = 0, lastTryStopTime = 0;
if (now - lastTryStopTime < sbdSleepTime) {
return;
}
lastTryStopTime = now;
for (jobCard = jobQueHead->forw; jobCard != jobQueHead; jobCard = next) {
next = jobCard->forw;
if (jobCard->jobSpecs.numToHosts == 1) {
if ((jobCard->jobSpecs.jStatus & JOB_STAT_RUN)
&& (now >= jobCard->jobSpecs.startTime + sbdSleepTime)
&& shouldStop (myload, jobCard, &reasons, &subreasons, 1, &stopmore)) {
jobSuspendAction (jobCard, SIG_SUSP_LOAD, reasons, subreasons);
if (stopmore)
continue;
else
return;
}
} else {
struct hostLoad *load;
int numh;
struct nameList *hostList;
numh = jobCard->jobSpecs.numToHosts;
hostList = lsb_compressStrList(jobCard->jobSpecs.toHosts, numh);
numh = hostList->listSize;
if (hostList->listSize == 1) {
load = myload;
} else {
load = ls_loadofhosts ("-", &numh, EFFECTIVE, 0,
hostList->names, hostList->listSize);
}
if (load == NULL) {
if (errCount < 3)
ls_syslog(LOG_ERR, I18N_JOB_FAIL_S_MM, fname,
lsb_jobid2str(jobCard->jobSpecs.jobId),
"ls_loadofhosts");
errCount++;
if (lserrno == LSE_LIM_BADHOST)
relife();
if (lserrno == LSE_BAD_XDR)
relife();
if (lserrno == LSE_LIM_DOWN || lserrno == LSE_TIME_OUT)
myStatus |= NO_LIM;
continue;
} else {
errCount = 0;
myStatus = 0;
}
if ((jobCard->jobSpecs.jStatus & JOB_STAT_RUN)
&& now >= jobCard->jobSpecs.startTime + sbdSleepTime) {
if (shouldStop (load, jobCard, &reasons, &subreasons, numh, &stopmore)) {
jobSuspendAction (jobCard, SIG_SUSP_LOAD, reasons, subreasons);
if (stopmore)
break;
else
return;
}
}
}
}
return;
}
static void
tryResume (struct hostLoad *myload)
{
char fname[] = "tryResume";
struct jobCard *jobCard, *next;
static int errCount = 0, lastTryResumeTime = 0;
if (now - lastTryResumeTime < sbdSleepTime) {
return;
}
lastTryResumeTime = now;
for (jobCard = jobQueHead->back; jobCard != jobQueHead; jobCard = next) {
next = jobCard->back;
if (!(jobCard->jobSpecs.jStatus & JOB_STAT_SSUSP) ||
jobCard->jobSpecs.actPid)
continue;
if (jobCard->jobSpecs.numToHosts == 1) {
if (shouldResume (myload, jobCard, 1)) {
if (jobResumeAction(jobCard, SIG_RESUME_LOAD, LOAD_REASONS) < 0)
continue;
else
return;
}
} else {
int numh;
struct hostLoad *load;
struct nameList *hostList;
numh = jobCard->jobSpecs.numToHosts;
hostList = lsb_compressStrList(jobCard->jobSpecs.toHosts, numh);
numh = hostList->listSize;
load = ls_loadofhosts ("-", &numh, EFFECTIVE, 0,
hostList->names, hostList->listSize);
if (load == NULL) {
if (errCount < 3)
ls_syslog(LOG_ERR, I18N_JOB_FAIL_S_M, fname,
lsb_jobid2str(jobCard->jobSpecs.jobId),
"ls_loadofhosts");
errCount++;
if (lserrno == LSE_LIM_BADHOST)
relife();
if (lserrno == LSE_BAD_XDR)
relife();
if (lserrno == LSE_LIM_DOWN || lserrno == LSE_TIME_OUT)
myStatus |= NO_LIM;
continue;
} else {
myStatus = 0;
errCount = 0;
}
if (!shouldResume (load, jobCard, numh))
continue;
if (jobResumeAction(jobCard, SIG_RESUME_LOAD, LOAD_REASONS) < 0)
continue;
else
return;
}
}
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;
}
void
do_probe(XDR * xdrs, int chfd, struct LSFHeader * reqHdr)
{
static char fname[] = "do_probe()";
char reply_buf[MSGSIZE];
XDR xdrs2;
struct LSFHeader replyHdr;
struct sbdPackage sbdPackage;
struct jobSpecs *jobSpecs;
int i;
struct lsfAuth *auth = NULL;
if (reqHdr->length == 0)
return;
initLSFHeader_(&replyHdr);
replyHdr.opCode = ERR_NO_ERROR;
jobSpecs = NULL;
if (!xdr_sbdPackage(xdrs, &sbdPackage, reqHdr)) {
ls_syslog(LOG_ERR, I18N_FUNC_FAIL, fname, "xdr_sbdPackage");
relife();
} else {
if (sbdPackage.numJobs) {
jobSpecs = my_calloc(sbdPackage.numJobs,
sizeof(struct jobSpecs), fname);
for (i = 0; i < sbdPackage.numJobs; i++) {
if (!xdr_arrayElement(xdrs, (char *) &(jobSpecs[i]),
reqHdr, xdr_jobSpecs)) {
replyHdr.opCode = ERR_BAD_REQ;
ls_syslog(LOG_ERR, _i18n_msg_get(ls_catd , NL_SETN, 5815,
"%s: %s(%d) failed for %d jobs"), /* catgets 5815 */
fname, "xdr_arrayElement", i, sbdPackage.numJobs);
break;
}
refreshJob(&(jobSpecs[i]));
xdr_lsffree(xdr_jobSpecs, (char *)&jobSpecs[i], reqHdr);
}
}
}
if (replyHdr.opCode == ERR_NO_ERROR)
if (!xdr_sbdPackage1(xdrs, &sbdPackage, reqHdr)) {
ls_syslog(LOG_ERR, I18N_FUNC_FAIL, fname, "xdr_sbdPackage1");
relife();
}
if (replyHdr.opCode == ERR_NO_ERROR) {
if (myStatus & NO_LIM) {
replyHdr.opCode = ERR_NO_LIM;
}
}
xdrmem_create(&xdrs2, reply_buf, MSGSIZE, XDR_ENCODE);
if (!xdr_encodeMsg(&xdrs2, NULL, &replyHdr, NULL, 0, auth)) {
ls_syslog(LOG_ERR, I18N_FUNC_FAIL, fname, "xdr_encodeMsg");
relife();
}
if (chanWrite_(chfd, reply_buf, XDR_GETPOS(&xdrs2)) <= 0) {
ls_syslog(LOG_ERR, I18N_FUNC_FAIL_M, fname, "chanWrite_");
}
xdr_destroy(&xdrs2);
if (jobSpecs != NULL)
free(jobSpecs);
getManagerId(&sbdPackage);
mbdPid = sbdPackage.mbdPid;
sbdSleepTime = sbdPackage.sbdSleepTime;
retryIntvl = sbdPackage.retryIntvl;
preemPeriod = sbdPackage.preemPeriod;
pgSuspIdleT = sbdPackage.pgSuspIdleT;
maxJobs = sbdPackage.maxJobs;
uJobLimit = sbdPackage.uJobLimit;
rusageUpdateRate = sbdPackage.rusageUpdateRate;
rusageUpdatePercent = sbdPackage.rusageUpdatePercent;
jobTerminateInterval = sbdPackage.jobTerminateInterval;
for (i = 0; i < sbdPackage.nAdmins; i++)
FREEUP(sbdPackage.admins[i]);
FREEUP(sbdPackage.admins);
return;
}
void
do_modifyjob(XDR * xdrs, int chfd, struct LSFHeader * reqHdr)
{
static char fname[] = "do_switchjob()";
char reply_buf[MSGSIZE];
XDR xdrs2;
struct jobSpecs jobSpecs;
struct jobReply jobReply;
sbdReplyType reply;
char found = FALSE;
struct LSFHeader replyHdr;
char *replyStruct;
struct jobCard *jp;
struct lsfAuth *auth = NULL;
memset(&jobReply, 0, sizeof(struct jobReply));
if (!xdr_jobSpecs(xdrs, &jobSpecs, reqHdr)) {
reply = ERR_BAD_REQ;
ls_syslog(LOG_ERR, I18N_FUNC_FAIL, fname, "xdr_jobSpecs");
goto sendReply;
}
for (jp = jobQueHead->back; jp != jobQueHead; jp = jp->back)
if (jp->jobSpecs.jobId == jobSpecs.jobId) {
found = TRUE;
break;
}
if (!found) {
reply = ERR_NO_JOB;
ls_syslog(LOG_ERR, _i18n_msg_get(ls_catd , NL_SETN, 5808,
"%s: mbatchd trying to modify a non-existent job <%s>"), fname, lsb_jobid2str(jobSpecs.jobId)); /* catgets 5808 */
goto sendReply;
}
if (jp->jobSpecs.jStatus & (JOB_STAT_DONE | JOB_STAT_EXIT)) {
reply = ERR_JOB_FINISH;
goto sendReply;
}
if ((lsbJobCpuLimit != 1) &&
((jp->jobSpecs.lsfLimits[LSF_RLIMIT_CPU].rlim_maxl
!= jobSpecs.lsfLimits[LSF_RLIMIT_CPU].rlim_maxl) ||
(jp->jobSpecs.lsfLimits[LSF_RLIMIT_CPU].rlim_maxh
!= jobSpecs.lsfLimits[LSF_RLIMIT_CPU].rlim_maxh) ||
(jp->jobSpecs.lsfLimits[LSF_RLIMIT_CPU].rlim_curl
!= jobSpecs.lsfLimits[LSF_RLIMIT_CPU].rlim_curl) ||
(jp->jobSpecs.lsfLimits[LSF_RLIMIT_CPU].rlim_curh
!= jobSpecs.lsfLimits[LSF_RLIMIT_CPU].rlim_curh)
)) {
ls_syslog(LOG_ERR, _i18n_msg_get(ls_catd, NL_SETN, 5809, "%s, LSB_JOB_CPULIMIT is not set for the host, job <%s>, CPU limit not modified"), fname, lsb_jobid2str(jobSpecs.jobId));
} else {
memcpy((char *) &jp->jobSpecs.lsfLimits[LSF_RLIMIT_CPU],
(char *) &jobSpecs.lsfLimits[LSF_RLIMIT_CPU],
sizeof(struct lsfLimit));
}
if ((lsbJobMemLimit != 1) &&
((jp->jobSpecs.lsfLimits[LSF_RLIMIT_RSS].rlim_maxl
!= jobSpecs.lsfLimits[LSF_RLIMIT_RSS].rlim_maxl) ||
(jp->jobSpecs.lsfLimits[LSF_RLIMIT_RSS].rlim_maxh
!= jobSpecs.lsfLimits[LSF_RLIMIT_RSS].rlim_maxh) ||
(jp->jobSpecs.lsfLimits[LSF_RLIMIT_RSS].rlim_curl
!= jobSpecs.lsfLimits[LSF_RLIMIT_RSS].rlim_curl) ||
(jp->jobSpecs.lsfLimits[LSF_RLIMIT_RSS].rlim_curh
!= jobSpecs.lsfLimits[LSF_RLIMIT_RSS].rlim_curh)
)) {
ls_syslog(LOG_ERR, _i18n_msg_get(ls_catd, NL_SETN, 5810, "%s, LSB_JOB_MEMLIMIT is not set for the host, job <%s>, memory limit not modified"), fname, lsb_jobid2str(jobSpecs.jobId));
} else {
memcpy((char *) &jp->jobSpecs.lsfLimits[LSF_RLIMIT_RSS],
(char *) &jobSpecs.lsfLimits[LSF_RLIMIT_RSS],
sizeof(struct lsfLimit));
}
memcpy((char *) &jp->jobSpecs.lsfLimits[LSF_RLIMIT_RUN],
(char *) &jobSpecs.lsfLimits[LSF_RLIMIT_RUN],
sizeof(struct lsfLimit));
setRunLimit(jp, FALSE);
if (strcmp(jp->jobSpecs.outFile, jobSpecs.outFile) ||
!(strcmp(jobSpecs.outFile, "/dev/null")))
{
strcpy(jp->jobSpecs.outFile, jobSpecs.outFile);
if (strcmp(jobSpecs.outFile, "/dev/null") ||
(jobSpecs.options & SUB_OUT_FILE)) {
jp->jobSpecs.options |= SUB_OUT_FILE;
}
else {
jp->jobSpecs.options &= ~SUB_OUT_FILE;
}
}
if (strcmp(jp->jobSpecs.errFile, jobSpecs.errFile))
{
strcpy(jp->jobSpecs.errFile, jobSpecs.errFile);
if (!strcmp(jp->jobSpecs.errFile, "/dev/null")
&& !(jobSpecs.options & SUB_ERR_FILE)) {
jp->jobSpecs.options &= ~SUB_ERR_FILE;
}
}
if (jobSpecs.options & SUB_RERUNNABLE) {
jp->jobSpecs.options |= SUB_RERUNNABLE;
} else {
jp->jobSpecs.options &= ~SUB_RERUNNABLE;
}
sendReply:
xdr_lsffree(xdr_jobSpecs, (char *)&jobSpecs, reqHdr);
xdrmem_create(&xdrs2, reply_buf, MSGSIZE, XDR_ENCODE);
initLSFHeader_(&replyHdr);
replyHdr.opCode = reply;
if (reply == ERR_NO_ERROR)
replyStruct = (char *) &jobReply;
else {
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_JOB_FAIL_S_M, fname,
lsb_jobid2str(jp->jobSpecs.jobId), "chanWrite_");
}
xdr_destroy(&xdrs2);
return;
}
void
do_newjob(XDR *xdrs, int chfd, struct LSFHeader *reqHdr)
{
static char fname[] = "do_newjob()";
char reply_buf[MSGSIZE];
XDR xdrs2;
struct jobSpecs jobSpecs;
struct jobReply jobReply;
struct jobCard *jp;
sbdReplyType reply;
struct LSFHeader replyHdr;
char *replyStruct;
struct lsfAuth *auth = NULL;
memset(&jobReply, 0, sizeof(struct jobReply));
if (!xdr_jobSpecs(xdrs, &jobSpecs, reqHdr)) {
reply = ERR_BAD_REQ;
ls_syslog(LOG_ERR, I18N_FUNC_FAIL, fname, "xdr_jobSpecs");
goto sendReply;
}
for (jp = jobQueHead->forw; (jp != jobQueHead); jp = jp->forw) {
if (jp->jobSpecs.jobId == jobSpecs.jobId) {
jobReply.jobId = jp->jobSpecs.jobId;
jobReply.jobPid = jp->jobSpecs.jobPid;
jobReply.jobPGid = jp->jobSpecs.jobPGid;
jobReply.jStatus = jp->jobSpecs.jStatus;
reply = ERR_NO_ERROR;
goto sendReply;
}
}
jp = calloc(1, sizeof(struct jobCard));
if (jp == NULL) {
ls_syslog(LOG_ERR, I18N_JOB_FAIL_S_M, fname,
lsb_jobid2str(jobSpecs.jobId), "calloc");
reply = ERR_MEM;
goto sendReply;
}
memcpy((char *) &jp->jobSpecs, (char *) &jobSpecs,
sizeof(struct jobSpecs));
jp->jobSpecs.jStatus &= ~JOB_STAT_MIG;
jp->jobSpecs.startTime = now;
jp->jobSpecs.reasons = 0;
jp->jobSpecs.subreasons = 0;
/* Initialize the core number
*/
jp->core_num = -1;
if (jp->jobSpecs.jAttrib & Q_ATTRIB_EXCLUSIVE) {
if (lockHosts (jp) < 0) {
ls_syslog(LOG_ERR, I18N_JOB_FAIL_S, fname,
lsb_jobid2str(jp->jobSpecs.jobId), "lockHosts");
unlockHosts (jp, jp->jobSpecs.numToHosts);
reply = ERR_LOCK_FAIL;
freeWeek(jp->week);
FREEUP(jp);
goto sendReply;
}
}
jp->runTime = 0;
if (initJobCard(jp, &jobSpecs, (int *)&reply) < 0) {
if (jp->jobSpecs.jAttrib & Q_ATTRIB_EXCLUSIVE) {
unlockHosts (jp, jp->jobSpecs.numToHosts);
}
FREEUP(jp);
goto sendReply;
}
jp->execJobFlag = 0;
if (jp->runTime < 0) {
jp->runTime = 0;
}
jp->execGid = 0;
jp->execUsername[0] = '\0';
jp->jobSpecs.execUid = -1;
jp->jobSpecs.execUsername[0] = '\0';
if (jp->jobSpecs.jobSpoolDir[0] != '\0') {
char *tmp;
if ((tmp = getUnixSpoolDir (jp->jobSpecs.jobSpoolDir)) == NULL) {
jp->jobSpecs.jobSpoolDir[0] = '\0';
}
}
if ((logclass & LC_TRACE) && jp->jobSpecs.jobSpoolDir[0] != 0) {
ls_syslog(LOG_DEBUG,
"%s: the SpoolDir for job <%s> is %s \n",
fname, lsb_jobid2str(jp->jobSpecs.jobId),
jp->jobSpecs.jobSpoolDir);
}
if (jp->jobSpecs.options & SUB_PRE_EXEC)
SBD_SET_STATE(jp, (JOB_STAT_RUN | JOB_STAT_PRE_EXEC))
else
SBD_SET_STATE(jp, JOB_STAT_RUN);
reply = job_exec(jp, chfd);
if (reply != ERR_NO_ERROR) {
ls_syslog(LOG_ERR, I18N_JOB_FAIL_S, fname,
lsb_jobid2str(jp->jobSpecs.jobId), "job_exec");
if (jp->jobSpecs.jAttrib & Q_ATTRIB_EXCLUSIVE) {
unlockHosts (jp, jp->jobSpecs.numToHosts);
}
deallocJobCard(jp);
} else {
jobReply.jobId = jp->jobSpecs.jobId;
jobReply.jobPid = jp->jobSpecs.jobPid;
jobReply.jobPGid = jp->jobSpecs.jobPGid;
jobReply.jStatus = jp->jobSpecs.jStatus;
}
sendReply:
xdr_lsffree(xdr_jobSpecs, (char *)&jobSpecs, reqHdr);
xdrmem_create(&xdrs2, reply_buf, MSGSIZE, XDR_ENCODE);
initLSFHeader_(&replyHdr);
replyHdr.opCode = reply;
replyStruct = (reply == ERR_NO_ERROR) ? (char *) &jobReply : (char *) NULL;
if (!xdr_encodeMsg(&xdrs2, replyStruct, &replyHdr, xdr_jobReply, 0, auth)) {
ls_syslog(LOG_ERR, I18N_FUNC_FAIL, fname, "xdr_jobReply");
lsb_merr(_i18n_msg_get(ls_catd , NL_SETN, 5804,
"Fatal error: xdr_jobReply() failed; sbatchd relifing")); /* catgets 5804 */
relife();
}
if (chanWrite_(chfd, reply_buf, XDR_GETPOS(&xdrs2)) <= 0) {
ls_syslog(LOG_ERR, _i18n_msg_get(ls_catd , NL_SETN, 5805,
"%s: Sending jobReply (len=%d) to master failed: %m"), /* catgets 5805 */
fname, XDR_GETPOS(&xdrs2));
}
xdr_destroy(&xdrs2);
if (reply == ERR_NO_ERROR && !daemonParams[LSB_BSUBI_OLD].paramValue &&
PURE_INTERACTIVE(&jp->jobSpecs)) {
if (status_job (BATCH_STATUS_JOB, jp, jp->jobSpecs.jStatus,
ERR_NO_ERROR) < 0) {
jp->notReported++;
}
}
}
void
do_switchjob(XDR * xdrs, int chfd, struct LSFHeader * reqHdr)
{
static char fname[] = "do_switchjob()";
char reply_buf[MSGSIZE];
XDR xdrs2;
struct jobSpecs jobSpecs;
struct jobReply jobReply;
int i;
sbdReplyType reply;
char *cp;
char *word;
char found = FALSE;
struct LSFHeader replyHdr;
char *replyStruct;
struct jobCard *jp;
struct lsfAuth *auth = NULL;
memset(&jobReply, 0, sizeof(struct jobReply));
if (!xdr_jobSpecs(xdrs, &jobSpecs, reqHdr)) {
reply = ERR_BAD_REQ;
ls_syslog(LOG_ERR, I18N_FUNC_FAIL, fname, "xdr_jobSpecs");
goto sendReply;
}
for (jp = jobQueHead->back; jp != jobQueHead; jp = jp->back) {
if (jp->jobSpecs.jobId == jobSpecs.jobId) {
found = TRUE;
break;
}
}
if (!found) {
reply = ERR_NO_JOB;
ls_syslog(LOG_ERR, _i18n_msg_get(ls_catd , NL_SETN, 5807,
"%s: mbatchd trying to switch a non-existent job <%s>"), fname, lsb_jobid2str(jobSpecs.jobId)); /* catgets 5807 */
goto sendReply;
}
if (jp->jobSpecs.jStatus & (JOB_STAT_DONE | JOB_STAT_EXIT)) {
reply = ERR_JOB_FINISH;
goto sendReply;
}
cp = jobSpecs.windows;
freeWeek(jp->week);
while ((word = getNextWord_(&cp)) != NULL) {
if (addWindow(word, jp->week, "switchJob jobSpecs") < 0) {
ls_syslog(LOG_ERR, I18N_JOB_FAIL_S_S_M, fname,
lsb_jobid2str(jp->jobSpecs.jobId), "addWindow", word);
freeWeek(jp->week);
reply = ERR_BAD_REQ;
goto sendReply;
}
}
jp->windEdge = now;
if ((jp->jobSpecs.jAttrib & Q_ATTRIB_EXCLUSIVE)
&& !(jobSpecs.jAttrib & Q_ATTRIB_EXCLUSIVE))
for (i = 0; i < jp->jobSpecs.numToHosts; i++)
if (unlockHost_(jp->jobSpecs.toHosts[i]) < 0
&& lserrno != LSE_LIM_NLOCKED)
ls_syslog(LOG_ERR, I18N_JOB_FAIL_S_S_MM, fname,
lsb_jobid2str(jp->jobSpecs.jobId), "unlockHost_", jp->jobSpecs.toHosts[i]);
strcpy(jp->jobSpecs.queue, jobSpecs.queue);
strcpy(jp->jobSpecs.windows, jobSpecs.windows);
jp->jobSpecs.priority = jobSpecs.priority;
jp->jobSpecs.nice = jobSpecs.nice;
jp->jobSpecs.jAttrib = jobSpecs.jAttrib;
freeThresholds (&jp->jobSpecs.thresholds);
saveThresholds (&jp->jobSpecs, &jobSpecs.thresholds);
memcpy((char *) &jp->jobSpecs.lsfLimits[LSF_RLIMIT_RUN],
(char *) &jobSpecs.lsfLimits[LSF_RLIMIT_RUN],
sizeof(struct lsfLimit));
strcpy (jp->jobSpecs.requeueEValues, jobSpecs.requeueEValues);
strcpy (jp->jobSpecs.resumeCond, jobSpecs.resumeCond);
strcpy (jp->jobSpecs.stopCond, jobSpecs.stopCond);
lsbFreeResVal (&jp->resumeCondVal);
if (jobSpecs.resumeCond && jobSpecs.resumeCond[0] != '\0') {
if ((jp->resumeCondVal = checkThresholdCond (jobSpecs.resumeCond))
== NULL)
ls_syslog(LOG_ERR, I18N_JOB_FAIL_S_S, fname,
lsb_jobid2str(jp->jobSpecs.jobId),
"checkThresholdCond", jobSpecs.resumeCond);
}
lsbFreeResVal (&jp->stopCondVal);
if (jobSpecs.stopCond && jobSpecs.stopCond[0] != '\0') {
if ((jp->stopCondVal = checkThresholdCond (jobSpecs.stopCond))
== NULL)
ls_syslog(LOG_ERR, I18N_JOB_FAIL_S_S, fname,
lsb_jobid2str(jp->jobSpecs.jobId),
"checkThresholdCond", jobSpecs.stopCond);
}
if (jobSpecs.options & SUB_LOGIN_SHELL) {
FREEUP (jp->jobSpecs.loginShell);
jp->jobSpecs.loginShell = safeSave (jobSpecs.loginShell);
}
strcpy (jp->jobSpecs.suspendActCmd, jobSpecs.suspendActCmd);
strcpy (jp->jobSpecs.resumeActCmd, jobSpecs.resumeActCmd);
strcpy (jp->jobSpecs.terminateActCmd, jobSpecs.terminateActCmd);
setRunLimit (jp, FALSE);
offList ((struct listEntry *)jp);
inJobLink (jp);
if (reniceJob(jp) < 0)
ls_syslog(LOG_DEBUG, "%s: renice job <%s> failed",
fname, lsb_jobid2str(jp->jobSpecs.jobId));
reply = ERR_NO_ERROR;
jobReply.jobId = jp->jobSpecs.jobId;
jobReply.jobPid = jp->jobSpecs.jobPid;
jobReply.jobPGid = jp->jobSpecs.jobPGid;
jobReply.jStatus = jp->jobSpecs.jStatus;
sendReply:
xdr_lsffree(xdr_jobSpecs, (char *)&jobSpecs, reqHdr);
xdrmem_create(&xdrs2, reply_buf, MSGSIZE, XDR_ENCODE);
initLSFHeader_(&replyHdr);
replyHdr.opCode = reply;
if (reply == ERR_NO_ERROR)
replyStruct = (char *) &jobReply;
else {
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_JOB_FAIL_S_M, fname,
lsb_jobid2str(jp->jobSpecs.jobId), "chanWrite_");
}
xdr_destroy(&xdrs2);
return;
}
