SEXP
lsf_job_submit(SEXP sexp_debug, SEXP sexp_command, SEXP sexp_ncpus)
{
int jobId, debug, i;
struct submit submitRequest;
struct submitReply submitReply;
debug = INTEGER(sexp_debug)[0];
memset(&submitRequest, 0, sizeof(submitRequest));
for (i = 0; i < LSF_RLIM_NLIMITS; i++)
submitRequest.rLimits[i] = DEFAULT_RLIMIT;
submitRequest.command = CHAR(STRING_ELT(sexp_command, 0));
submitRequest.options |= SUB_OUT_FILE;
if (debug) {
submitRequest.outFile = "Rlsf_job_output.%J";
} else {
submitRequest.outFile = "/dev/null";
}
submitRequest.numProcessors = INTEGER(sexp_ncpus)[0];
submitRequest.maxNumProcessors = INTEGER(sexp_ncpus)[0];
jobId = lsb_submit(&submitRequest, &submitReply);
if (jobId == -1) {
Rprintf("lsf_job_submit: lsb_submit: %s\n", lsb_sysmsg());
return AsInt(0);
}
return AsInt(jobId);
}
void
sub_perror (char *usrMsg)
{
if (usrMsg) {
fputs(usrMsg, stderr);
fputs(": ", stderr);
}
fputs(lsb_sysmsg(), stderr);
}
SEXP
lsf_initialize(void)
{
if (lsb_init("R")) {
Rprintf("lsf_initialize: lsb_init: %s\n", lsb_sysmsg());
return AsInt(-1);
}
else {
return AsInt(0);
}
if (putenv("BSUB_QUIET=1")) {
return AsInt(0);
}
}
SEXP
lsf_resume_job(SEXP sexp_jobid)
{
int jobid, rc;
jobid = INTEGER(sexp_jobid)[0];
rc = lsb_signaljob(jobid, SIGCONT);
if (rc < 0) {
Rprintf("lsf_resume_job: lsb_signaljob: %s\n", lsb_sysmsg());
return AsInt(-1);
}
return AsInt(0);
}
char *
lsb_sperror(char *usrMsg)
{
char errmsg[256];
char *rtstr;
errmsg[0] = '\0';
if (usrMsg)
sprintf(errmsg, "%s: ", usrMsg);
strcat(errmsg, lsb_sysmsg());
if ((rtstr=(char *)malloc(sizeof(char)*(strlen(errmsg)+1))) == NULL){
lserrno = LSE_MALLOC;
return NULL;
}
strcpy(rtstr, errmsg);
return rtstr;
}
int
main(int argc, char **argv)
{
int cc;
struct job_group jg;
if (lsb_init(argv[0]) < 0) {
lsb_perror("lsb_init");
return -1;
}
while ((cc = getopt(argc, argv, "hV")) != EOF) {
switch (cc) {
case 'V':
fputs(_LS_VERSION_, stderr);
return 0;
case 'h':
usage();
exit(-1);
}
}
if (argc <= optind) {
usage();
return -1;
}
jg.group_name = argv[argc - 1];
cc = lsb_deljgrp(&jg);
if (cc != LSBE_NO_ERROR) {
fprintf(stderr, "bgdel: %s.\n", lsb_sysmsg());
return -1;
}
printf("Group %s removed successfully.\n", jg.group_name);
return 0;
}
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);
}
}
SEXP
lsf_job_submit2(SEXP ctrl) {
char *name;
char *val;
char *vals[100];
int jobId, debug, i, j;
SEXP elmt = R_NilValue;
SEXP names = getAttrib(ctrl, R_NamesSymbol);
struct submit submitRequest;
struct submitReply submitReply;
memset(&submitRequest, 0, sizeof(submitRequest));
for (i = 0; i < LSF_RLIM_NLIMITS; i++)
submitRequest.rLimits[i] = DEFAULT_RLIMIT;
for (i = 0; i < length(names); i++) {
name = CHAR(STRING_ELT(names, i));
elmt = VECTOR_ELT(ctrl, i);
val = CHAR(STRING_ELT(elmt, 0));
if (0) {
} else if (!strcmp(name, "jobName")) {
submitRequest.options |= SUB_JOB_NAME;
submitRequest.jobName = val;
} else if (!strcmp(name, "queue")) {
submitRequest.options |= SUB_QUEUE;
submitRequest.queue = val;
} else if (!strcmp(name, "askedHosts")){
submitRequest.options |= SUB_HOST;
char **chars;
chars = (char **)calloc(length(elmt), sizeof(char *));
for (j = 0; j < length(elmt); j++){
val=CHAR(STRING_ELT(elmt, j));
chars[j] = val;
}
submitRequest.numAskedHosts = length(elmt);
submitRequest.askedHosts = chars;
} else if (!strcmp(name, "resReq")){
submitRequest.options |= SUB_RES_REQ;
submitRequest.resReq = val;
} else if (!strcmp(name, "rlimit_cpu")) {
submitRequest.rLimits[LSF_RLIMIT_CPU] = atoi(val);
} else if (!strcmp(name, "rlimit_fsize")) {
submitRequest.rLimits[LSF_RLIMIT_FSIZE] = atoi(val);
} else if (!strcmp(name, "rlimit_data")) {
submitRequest.rLimits[LSF_RLIMIT_DATA] = atoi(val);
} else if (!strcmp(name, "rlimit_stack")) {
submitRequest.rLimits[LSF_RLIMIT_STACK] = atoi(val);
} else if (!strcmp(name, "rlimit_core")) {
submitRequest.rLimits[LSF_RLIMIT_CORE] = atoi(val);
} else if (!strcmp(name, "rlimit_rss")) {
submitRequest.rLimits[LSF_RLIMIT_RSS] = atoi(val);
} else if (!strcmp(name, "rlimit_nofile")) {
submitRequest.rLimits[LSF_RLIMIT_NOFILE] = atoi(val);
} else if (!strcmp(name, "rlimit_open_max")) {
submitRequest.rLimits[LSF_RLIMIT_OPEN_MAX] = atoi(val);
} else if (!strcmp(name, "rlimit_swap")) {
submitRequest.rLimits[LSF_RLIMIT_SWAP] = atoi(val);
} else if (!strcmp(name, "rlimit_run")) {
submitRequest.rLimits[LSF_RLIMIT_RUN] = atoi(val);
} else if (!strcmp(name, "rlimit_process")) {
submitRequest.rLimits[LSF_RLIMIT_PROCESS] = atoi(val);
} else if (!strcmp(name, "hostSpec")) {
submitRequest.options |= SUB_HOST_SPEC;
submitRequest.hostSpec = val;
} else if (!strcmp(name, "numProcessors")){
submitRequest.numProcessors = atoi(val);
} else if (!strcmp(name, "dependCond")){
submitRequest.options |= SUB_DEPEND_COND;
submitRequest.dependCond = val;
} else if (!strcmp(name, "beginTime")) {
submitRequest.beginTime = atoi(val);
} else if (!strcmp(name, "termTime")) {
submitRequest.termTime = atoi(val);
} else if (!strcmp(name, "sigValue")) {
submitRequest.options |= SUB_WINDOW_SIG;
submitRequest.sigValue= atoi(val);
} else if (!strcmp(name, "command")){
submitRequest.command = val;
} else if (!strcmp(name, "inFile")){
submitRequest.options |= SUB_IN_FILE;
submitRequest.inFile = val;
} else if (!strcmp(name, "outFile")){
submitRequest.options |= SUB_OUT_FILE;
submitRequest.outFile = val;
} else if (!strcmp(name, "errFile")){
submitRequest.options |= SUB_ERR_FILE;
submitRequest.errFile = val;
} else if (!strcmp(name, "chkpntPeriod")){
submitRequest.options |= SUB_CHKPNTABLE;
submitRequest.options |= SUB_CHKPNT_PERIOD;
submitRequest.chkpntPeriod = atoi(val);
} else if (!strcmp(name, "chkpntDir")){
submitRequest.options |= SUB_CHKPNTABLE;
submitRequest.options |= SUB_CHKPNT_DIR;
submitRequest.chkpntDir = val;
} else if (!strcmp(name, "xFile")){
/*not implemented*/
/* SUB_OTHER_FILES probably should be set */
/* submitRequest.nxf should be assigned */
} else if (!strcmp(name, "preExecCmd")){
submitRequest.options |= SUB_PRE_EXEC;
submitRequest.preExecCmd = val;
} else if (!strcmp(name, "mailUser")){
submitRequest.options |= SUB_MAIL_USER;
submitRequest.mailUser = val;
} else if (!strcmp(name, "delOptions")){
submitRequest.delOptions = atoi(val);
} else if (!strcmp(name, "projectName")){
submitRequest.options |= SUB_PROJECT_NAME;
submitRequest.projectName = val;
} else if (!strcmp(name, "maxNumProcessors")){
submitRequest.maxNumProcessors = atoi(val);
} else if (!strcmp(name, "loginShell")){
submitRequest.options |= SUB_LOGIN_SHELL;
submitRequest.loginShell = val; }
else if (!strcmp(name, "userGroup")){
submitRequest.options |= SUB_USER_GROUP;
submitRequest.loginShell = val;
} else if (!strcmp(name, "exceptList")){
submitRequest.exceptList = val;
} else if (!strcmp(name, "exclusive")){
submitRequest.options |= SUB_EXCLUSIVE;
} else if (!strcmp(name, "notifyBegin")){
submitRequest.options |= SUB_NOTIFY_BEGIN;
} else if (!strcmp(name, "notifyEnd")){
submitRequest.options |= SUB_NOTIFY_END;
} else if (!strcmp(name, "restart")){
submitRequest.options |= SUB_RESTART;
} else if (!strcmp(name, "restartForce")){
submitRequest.options |= SUB_RESTART_FORCE;
} else if (!strcmp(name, "rerunnable")){
submitRequest.options |= SUB_RERUNNABLE;
/*} else if (!strcmp(name, "chkpnt_copy")){
submitRequest.options |= SUB_CHKPNT_COPY;*/
/*} else if (!strcmp(name, "chkpnt_force")){
submitRequest.options |= SUB_CHKPNT_FORCE;*/
} else if (!strcmp(name, "interactive")){
submitRequest.options |= SUB_INTERACTIVE;
} else if (!strcmp(name, "pty")){
submitRequest.options |= SUB_PTY;
} else if (!strcmp(name, "pty_shell")){
submitRequest.options |= SUB_PTY_SHELL;
} else if (!strcmp(name, "hold")){
submitRequest.options2 |= SUB2_HOLD;
} else if (!strcmp(name, "wait")){
/* This will require forking */
/*submitRequest.options2 |= SUB2_BSUB_BLOCK;*/
} else if (!strcmp(name, "debug")){
/* is debug used anymore? */
if (val) debug = 1;
}
val = "";
}
jobId = lsb_submit(&submitRequest, &submitReply);
if (jobId == -1) {
Rprintf("lsf_job_submit: lsb_submit: %s\n", lsb_sysmsg());
return AsInt(0);
}
return AsInt(jobId);
}
SEXP
lsf_job_status(SEXP sexp_jobid)
{
int jobid, numrec;
struct jobInfoEnt *jInfo;
SEXP status;
jobid = INTEGER(sexp_jobid)[0];
if ((numrec = lsb_openjobinfo(jobid, NULL, NULL, NULL, NULL, ALL_JOB)) < 0) {
Rprintf("lsf_job_status: lsb_openjobinfo: %s\n", lsb_sysmsg());
return R_NilValue;
}
jInfo = lsb_readjobinfo(&numrec);
if (jInfo == NULL) {
Rprintf("lsf_job_status: lsb_readjobinfo: %s\n", lsb_sysmsg());
lsb_closejobinfo();
return R_NilValue;
}
lsb_closejobinfo();
PROTECT(status = allocVector(STRSXP, 1));
switch(jInfo->status) {
case JOB_STAT_NULL:
SET_STRING_ELT(status, 0, mkChar("NULL"));
break;
case JOB_STAT_PEND:
SET_STRING_ELT(status, 0, mkChar("PEND"));
break;
case JOB_STAT_PSUSP:
SET_STRING_ELT(status, 0, mkChar("PSUSP"));
break;
case JOB_STAT_RUN:
SET_STRING_ELT(status, 0, mkChar("RUN"));
break;
case JOB_STAT_RUN|JOB_STAT_WAIT:
SET_STRING_ELT(status, 0, mkChar("WAIT"));
break;
case JOB_STAT_SSUSP:
SET_STRING_ELT(status, 0, mkChar("SSUSP"));
break;
case JOB_STAT_USUSP:
SET_STRING_ELT(status, 0, mkChar("USUSP"));
break;
case JOB_STAT_EXIT:
if (jInfo->reasons & EXIT_ZOMBIE)
SET_STRING_ELT(status, 0, mkChar("ZOMBI"));
else
SET_STRING_ELT(status, 0, mkChar("EXIT"));
break;
case JOB_STAT_DONE:
case JOB_STAT_DONE|JOB_STAT_PDONE:
case JOB_STAT_DONE|JOB_STAT_PERR:
case JOB_STAT_DONE|JOB_STAT_WAIT:
SET_STRING_ELT(status, 0, mkChar("DONE"));
break;
case JOB_STAT_UNKWN:
SET_STRING_ELT(status, 0, mkChar("UNKWN"));
break;
default:
Rprintf("lsf_job_status: job state <%d> is unknown.\n", jInfo->status);
SET_STRING_ELT(status, 0, mkChar("ERROR"));
break;
}
UNPROTECT(1);
return status;
}
