This commit was manufactured by cvs2svn to create tag

'slurm-0-3-9-1'.

META

@@ -9,8 +9,8 @@
   Name:		slurm
   Major:	0
   Minor:	3
-  Micro:        7
-  Version:	0.3.7
+  Micro:        9
+  Version:	0.3.9
   Release:	1
   API_CURRENT:	5	
   API_AGE:	4

NEWS

@@ -1,6 +1,23 @@
 This file describes changes in recent versions of SLURM. It primarily
 documents those changes that are of interest to users and admins. 
 
+* Changes in SLURM 0.3.9
+========================
+ -- Fixes for reported problems:
+   - slurm/512: Let job steps run on DRAINING nodes
+   - slurm/513: Gracefully deal with UIDs missing from passwd file
+ -- Create new allocation as needed for debugger in case old allocation
+    has been purged
+ -- Fix bug in scheduling jobs when a processor count is specified 
+    and FastSchedule=0 and the cluster is heterogeneous.
+ -- Fix srun bug when --input, --output and --error are all "none"
+ -- Allow single task id to be selected with --input, --output, and --error.
+ -- Create shared memory segment for Elan statistics when using the 
+    switch/elan plugin.
+
+* Changes in SLURM 0.3.8
+========================
+ -- More fixes necessary for TotalView.
 
 * Changes in SLURM 0.3.7
 ========================

doc/man/man1/scontrol.1

@@ -171,6 +171,7 @@ Possible values are"YES" and "NO".
 Set the job's required features on nodes specified value. Multiple values 
 may be comma separated if all features are required (AND operation) or 
 separated by "|" if any of the specified features are required (OR operation).
+Value may be cleared with blank data value, "Features=".
 .TP
 \fIJobId\fP=<id>
 Identify the job to be updated. This specification is required.
@@ -196,6 +197,7 @@ Set the job's priority to the specified value.
 \fIReqNodeList\fP=<nodes>
 Set the job's list of required node. Multiple node names may be specified using 
 simple node range expressions (e.g. "lx[10-20]"). 
+Value may be cleared with blank data value, "ReqNodeList=".
 .TP
 \fIReqNodes\fP=<count>
 Set the job's count of required nodes to the specified value.
@@ -232,8 +234,9 @@ changing its underlying state.
 \fBSPECIFICATIONS FOR UPDATE AND DELETE COMMANDS, PARTITIONS\fR
 .TP
 \fIAllowGroups\fP=<name>
-Identify the user groups which may use this partition. Multiple groups 
-may be specified in a comma separated list. 
+Identify the user groups which may use this partition. 
+Multiple groups may be specified in a comma separated list.
+To permit all groups to use the partition specify "AllowGroups=ALL" 
 .TP
 \fIDefault\fP=<yes|no>
 Specify if this partition is to be used by jobs which do not explicitly 
@@ -249,6 +252,7 @@ Possible values are"YES" and "NO".
 Identify the node(s) to be associated with this partition. Multiple node names 
 may be specified using simple node range expressions (e.g. "lx[10-20]"). 
 Note that jobs may only be associated with one partition at any time.
+Specify a blank data value to remove all nodes from a partition: "Nodes=".
 .TP
 \fIPartitionName\fP=<name>
 Identify the partition to be updated. This specification is required.

doc/man/man1/srun.1

@@ -223,9 +223,11 @@ the job. By default only errors are displayed.
 .TP
 \fB\-W\fR, \fB\-\-wait\fR=\fIseconds\fR
 Specify how long to wait after the first task terminates before terminating 
-all remaining tasks. The default value is unlimited. This can be useful to 
-insure that a job is terminated in a timely fashion in the event that one 
-or more tasks terminate prematurely.
+all remaining tasks. A value of 0 indicates an unlimited wait (a warning will
+be issued after 60 seconds). The default value is set by the WaitTime 
+parameter in the slurm configuration file (see \fBslurm.conf(5)\fR). This 
+option can be useful to insure that a job is terminated in a timely fashion 
+in the event that one or more tasks terminate prematurely.
 .TP
 \fB\-q\fR, \fB\-\-quit-on-interrupt\fR
 Quit immediately on single SIGINT (Ctrl-C). Use of this option

etc/init.d.slurm

@@ -16,6 +16,11 @@
 
 # $Id$
 
+BINDIR=/usr/bin
+CONFDIR=/etc/slurm
+LIBDIR=/usr/lib
+SBINDIR=/usr/sbin
+
 # Source function library.
 [ -f /etc/rc.d/init.d/functions ] || exit 0
 . /etc/rc.d/init.d/functions
@@ -28,14 +33,17 @@ else
     SLURMD_OPTIONS=""
 fi
 
-[ -f /etc/slurm/slurm.conf ] || exit 1
+[ -f $CONFDIR/slurm.conf ] || exit 1
+
+# setup library paths for slurm and munge support
+export LD_LIBRARY_PATH="$LIBDIR:$LD_LIBRARY_PATH"
 
 RETVAL=0
 
 start() {
     echo -n "starting $1: " 
     unset HOME MAIL USER USERNAME 
-    daemon /usr/sbin/$1 $2
+    daemon $SBINDIR/$1 $2
     RETVAL=$?
     echo
     touch /var/lock/subsys/slurm
@@ -52,7 +60,7 @@ stop() {
 }
 
 startall() {
-    for prog in `scontrol show daemons`; do 
+    for prog in `$BINDIR/scontrol show daemons`; do 
         optvar=`echo ${prog}_OPTIONS | tr "a-z" "A-Z"`
         start $prog ${!optvar} 
     done
@@ -68,7 +76,7 @@ slurmstatus() {
     local rpid
     local pidfile
 
-    pidfile=`grep -i ${base}pid /etc/slurm/slurm.conf | grep -v '^ *#'`
+    pidfile=`grep -i ${base}pid $CONFDIR/slurm.conf | grep -v '^ *#'`
     if [ $? = 0 ]; then
         pidfile=${pidfile##*=}
         pidfile=${pidfile%#*}
@@ -123,33 +131,34 @@ case "$1" in
         startall
         ;;
     stop)
-	for prog in `scontrol show daemons`; do
+	for prog in `$BINDIR/scontrol show daemons`; do
 	    stop $prog
 	done
         ;;
     status)
-	for prog in `scontrol show daemons`; do
+	for prog in `$BINDIR/scontrol show daemons`; do
 	   slurmstatus $prog
 	done
         ;;
     restart)
-        $0 stop && $0 start
+        $0 stop
+	$0 start
         ;;
     condrestart)
         if [ -f /var/lock/subsys/slurm ]; then
-            for prog in `scontrol show daemons`; do
+            for prog in `$BINDIR/scontrol show daemons`; do
                  stop $prog
                  start $prog
             done
         fi
         ;;
     reconfig)
-	for prog in `scontrol show daemons`; do
+	for prog in `$BINDIR/scontrol show daemons`; do
 	    killproc $prog -HUP
 	done
 	;;
     test)
-	for prog in `scontrol show daemons`; do   
+	for prog in `$BINDIR/scontrol show daemons`; do   
 	    echo "$prog runs here"
 	done
 	;;

slurm.spec.in

@@ -24,8 +24,12 @@ Requires: openssl >= 0.9.6
 #
 %if %{?_with_debug:1}%{!?_with_debug:0}
   %define _enable_debug --enable-debug
-  %define __os_install_post /usr/lib/rpm/brp-compress
 %endif
+#
+# Never allow rpm to strip binaries as this will break
+#  parallel debugging capability
+#
+%define __os_install_post /usr/lib/rpm/brp-compress
 
 
 %package devel
@@ -111,13 +115,13 @@ DESTDIR="$RPM_BUILD_ROOT" make install
 install -D -m755 etc/init.d.slurm $RPM_BUILD_ROOT/etc/rc.d/init.d/slurm
 install -D -m644 etc/slurm.conf.example $RPM_BUILD_ROOT/etc/slurm/slurm.conf
 # Delete unpackaged files:
-rm -f $RPM_BUILD_ROOT/usr/lib/slurm/*.{a,la}
+rm -f $RPM_BUILD_ROOT/%{_libdir}/slurm/*.{a,la}
 
 # Build file lists for optional plugin packages
 for plugin in auth_munge auth_authd switch_elan; do
    LIST=./${plugin}.files
    touch $LIST
-   test -f $RPM_BUILD_ROOT/usr/lib/slurm/${plugin}.so &&
+   test -f $RPM_BUILD_ROOT/%{_libdir}/slurm/${plugin}.so &&
      echo %{_libdir}/slurm/${plugin}.so > $LIST
 done
 
@@ -197,17 +201,12 @@ rm -rf $RPM_BUILD_ROOT
 
 %post
 /sbin/ldconfig %{_libdir}
-if [ -x /etc/rc.d/init.d/slurm ]; then
-    [ -x /sbin/chkconfig ] && /sbin/chkconfig --del slurm
+if [ $1 = 1 ]; then
     [ -x /sbin/chkconfig ] && /sbin/chkconfig --add slurm
-#    if ! /etc/rc.d/init.d/slurm status | grep -q running \
-#      || ! /etc/rc.d/init.d/slurm test | grep -q  slurmctld; then
-#        /etc/rc.d/init.d/slurm start
-#    fi 
 fi
 
 %preun
-if [ "$1" = 0 ]; then
+if [ $1 = 0 ]; then
     if [ -x /etc/rc.d/init.d/slurm ]; then
         [ -x /sbin/chkconfig ] && /sbin/chkconfig --del slurm
         if /etc/rc.d/init.d/slurm status | grep -q running; then
@@ -224,6 +223,11 @@ fi
 
 
 %changelog
+* Fri Oct 01 2004 Mark Grondona <mgrondona@llnl.gov>
+- don't delete and add service in %post
+* Wed Aug 04 2004 Mark Grondona <mgrondona@llnl.gov>
+- don't allow rpm to strip binaries since this breaks interface to parallel
+  debuggers.
 * Thu Jul 29 2004 Morris Jette <jette1@llnl.gov>
 - added checkpoint_none.so and jobcomp_script.so plugins
 * Fri Mar 07 2004 Mark Grondona <mgrondona@llnl.gov>

src/common/slurm_protocol_api.c

@@ -902,6 +902,9 @@ int slurm_send_rc_msg(slurm_msg_t *msg, int rc)
 {
         slurm_msg_t resp_msg;
         return_code_msg_t rc_msg;
+
+        if (msg->conn_fd < 0)
+                return (ENOTCONN);
 
         rc_msg.return_code = rc;
 

src/plugins/auth/munge/auth_munge.c

@@ -452,7 +452,7 @@ _decode_cred(char *m, slurm_auth_credential_t *c)
 		error ("Munge decode failed: %s %s", 
 			munge_ctx_strerror(ctx), retry ? "(retrying ...)": "");
 
-		if ((e = EMUNGE_SOCKET) && retry--)
+		if ((e == EMUNGE_SOCKET) && retry--)
 			goto again;
 
 		/*

src/plugins/jobcomp/filetxt/jobcomp_filetxt.c

@@ -162,7 +162,7 @@ _get_user_name(uint32_t user_id, char *user_name, int buf_size)
 		else
 			snprintf(user_name, buf_size, "Unknown");
 		cache_uid = user_id;
-		snprintf(cache_name, sizeof(cache_name), user_info->pw_name);
+		snprintf(cache_name, sizeof(cache_name), user_name);
 	}
 }
 
@@ -218,7 +218,7 @@ int slurm_jobcomp_log_record ( struct job_record *job_ptr )
 			job_state_string(job_state), 
 			job_ptr->partition, lim_str, start_str, 
 			end_str, job_ptr->nodes);
-	tot_size = (strlen(job_rec) + 1);
+	tot_size = strlen(job_rec);
 
 	while ( offset < tot_size ) {
 		wrote = write(job_comp_fd, job_rec + offset,

src/plugins/switch/elan/qsw.c

@@ -35,6 +35,8 @@
 #include <sys/param.h>
 #include <sys/types.h>
 #include <sys/wait.h>
+#include <sys/ipc.h>
+#include <sys/shm.h>
 #include <syslog.h>
 #include <errno.h>
 #include <string.h>
@@ -56,8 +58,17 @@
  *  want to include here since we are using the new
  *  version-nonspecific libelanctrl.
  *  (XXX: What is the equivalent in libelanctrl?)
+ *
+ * slurm/482: the elan USER context range is now split
+ *  into two segments, regular user context and RMS 
+ *  context ranges. Do not allow a context range
+ *  (lowcontext -- highcontext) to span these two segments,
+ *  as this will generate and elan initialization error
+ *  when MPI tries to attach to the capability. For now,
+ *  restrict SLURM's range to the RMS one (starting at 0x400)
+ * 
  */
-# define ELAN_USER_BASE_CONTEXT_NUM    0x020
+# define ELAN_USER_BASE_CONTEXT_NUM    0x400 /* act. RMS_BASE_CONTEXT_NUM */
 # define ELAN_USER_TOP_CONTEXT_NUM     0x7ff
 
 # define Version      cap_version
@@ -108,6 +119,7 @@
 #define QSW_CTX_END		ELAN_USER_TOP_CONTEXT_NUM - 1 
 #define QSW_CTX_INVAL		(-1)
 
+
 /* 
  * We are going to some trouble to keep these defs private so slurm
  * hackers not interested in the interconnect details can just pass around
@@ -152,6 +164,8 @@ static qsw_libstate_t qsw_internal_state = NULL;
 static pthread_mutex_t qsw_lock = PTHREAD_MUTEX_INITIALIZER;
 static elanhost_config_t elanconf = NULL;
 
+static int shmid = -1;
+
 
 /*
  * Allocate a qsw_libstate_t.
@@ -664,6 +678,10 @@ qsw_setup_jobinfo(qsw_jobinfo_t j, int nprocs, bitstr_t *nodeset,
 int
 qsw_prgdestroy(qsw_jobinfo_t jobinfo)
 {
+
+	if (shmid >= 0)
+		shmctl (shmid, IPC_RMID, NULL);
+
 	if (rms_prgdestroy(jobinfo->j_prognum) < 0) {
 		/* translate errno values to more descriptive ones */
 		switch (errno) {
@@ -698,6 +716,50 @@ qsw_prog_fini(qsw_jobinfo_t jobinfo)
 #endif
 }
 
+/* Key for Elan stats shared memory segment is the
+ *  rms.o program description number, left shifted 9 less 1
+ *  to avoid conflicts with MPI shared memory
+ */
+static int elan_statkey (int prgid)
+{
+	return ((prgid << 9) - 1);
+}
+
+/*
+ * Return the statkey to caller if shared memory was created
+ */
+int qsw_statkey (qsw_jobinfo_t jobinfo)
+{
+	return (shmid > 0 ? elan_statkey (jobinfo->j_prognum) : -1);
+}
+
+/*
+ * Create shared memory segment for Elan stats use
+ *  (ELAN_STATKEY env var is set in switch_elan.c)
+ */
+static int
+_qsw_shmem_create (qsw_jobinfo_t jobinfo, uid_t uid)
+{
+	struct shmid_ds shm;
+	ELAN_CAPABILITY *cap = &jobinfo->j_cap;
+	key_t key = elan_statkey (jobinfo->j_prognum);
+	int maxLocal = cap->HighContext - cap->LowContext + 2;
+	int pgsize = getpagesize ();
+
+	if ((shmid = shmget (key, pgsize * (maxLocal + 1), IPC_CREAT)) < 0)
+		return (error ("Failed to create Elan state shmem: %m"));
+
+	/* Ensure permissions on segment allow user read/write access
+	 */
+	shm.shm_perm.uid  = uid;
+	shm.shm_perm.mode = 0600;
+
+	if (shmctl (shmid, IPC_SET, &shm) < 0)
+		return (error ("Failed to set perms on Elan state shm: %m"));
+
+	return (0);
+}
+
 /*
  * Process 2: Create the context and make capability available to children.
  */
@@ -785,6 +847,14 @@ qsw_prog_init(qsw_jobinfo_t jobinfo, uid_t uid)
 		goto fail;
 	}
 
+
+	/*
+ 	 * Create shared memory for libelan state
+	 *  Failure to create shared memory is not a fatal error.
+	 */
+	_qsw_shmem_create (jobinfo, uid);
+
+
 	/* note: _elan3_fini() destroys context and makes capability unavail */
 	/* do it in qsw_prog_fini() after app terminates */
 	return 0;