diff --git a/META b/META
index 7b2a9a9df34139df7a2cc8ccfdd2d394aae01a5f..b9f60e14189d5b6a874166c039606c3e7cd303b2 100644
--- a/META
+++ b/META
@@ -3,9 +3,9 @@
Api_revision: 0
Major: 2
Meta: 1
- Micro: 6
+ Micro: 7
Minor: 1
Name: slurm
Release: 1
Release_tags: dist
- Version: 2.1.6
+ Version: 2.1.7
diff --git a/NEWS b/NEWS
index 0990fda7c72a04ba6d1a1876aee43fa3161ed06a..85f38030421d7f17e888909d5aa312bce6ced267 100644
--- a/NEWS
+++ b/NEWS
@@ -1,6 +1,49 @@
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 2.1.8
+========================
+
+* Changes in SLURM 2.1.7
+========================
+ -- Modify srun, salloc and sbatch parsing for the --signal option to accept
+ either a signal name in addition to the previously supported signal
+ numbers (e.g. "--signal=USR2@200").
+ -- BLUEGENE - Fixed sinfo --long --Node output for cpus on a single cnode.
+ -- In sched/wiki2 - Fix another logic bug in support of Moab being able to
+ identify preemptable jobs.
+ -- In sched/wiki2 - For BlueGene systems only: Fix bug preventing Moab from
+ being able to correctly change the node count of pending jobs.
+ -- In select/cons_res - Fix bug preventing job preemption with a configuration
+ of Shared=FORCE:1 and PreemptMode=GANG,SUSPEND.
+ -- In the TaskProlog, add support for an "unset" option to clear environment
+ variables for the user application. Also add support for embedded white-
+ space in the environment variables exported to the user application
+ (everything after the equal sign to the end of the line is included without
+ alteration).
+ -- Do not install /etc/init.d/slurm or /etc/init.d/slurmdbd on AIX systems.
+ -- BLUEGENE - fixed check for small blocks if a node card of a midplane is
+ in an error state other jobs can still run on the midplane on other
+ nodecards.
+ -- BLUEGENE - Check to make sure job killing is in the active job table in
+ DB2 when killing the job.
+ -- Correct logic to support ResvOverRun configuration parameter.
+ -- Get --acctg-freq option working for srun and salloc commands.
+ -- Fix sinfo display of drained nodes correctly with the summarize flag.
+ -- Fix minor memory leaks in slurmd and slurmstepd.
+ -- Better error messages for failed step launch.
+ -- Modify srun to insure compatability of the --relative option with the node
+ count requested.
+
+* Changes in SLURM 2.1.6-2
+==========================
+ -- In sched/wiki2 - Fix logic in support of Moab being able to identify
+ preemptable jobs.
+ -- Applied fixes to a debug4 message in priority_multifactor.c sent in by
+ Per Lundqvist
+ -- BLUEGENE - Fixed issue where incorrect nodecards could be picked when
+ looking at combining small blocks to make a larger small block.
+
* Changes in SLURM 2.1.6
========================
-- For newly submitted jobs, report expected start time in squeue --start as
@@ -4957,4 +5000,4 @@ documents those changes that are of interest to users and admins.
-- Change directory to /tmp in slurmd if daemonizing.
-- Logfiles are reopened on reconfigure.
-$Id: NEWS 19896 2010-03-26 21:49:42Z jette $
+$Id: NEWS 20049 2010-04-16 00:34:17Z da $
diff --git a/doc/html/accounting.shtml b/doc/html/accounting.shtml
index 4c01186639e25e6c4b3be2b40f6969a5e2bbaf05..f4952fcb55819f6f01883d71172c85bc08b5c018 100644
--- a/doc/html/accounting.shtml
+++ b/doc/html/accounting.shtml
@@ -559,7 +559,7 @@ sacctmgr add account chemistry,physics parent=science \
For example, to permit user <i>da</i> to execute jobs on all clusters
with a default account of <i>test</i> execute:</p>
<pre>
-sacctmgr add user da default=test
+sacctmgr add user da DefaultAccount=test
</pre>
<p>If <b>AccountingStorageEnforce=associations</b> is configured in
diff --git a/doc/man/man1/salloc.1 b/doc/man/man1/salloc.1
index af07a8434f7243cf74699f0b02afa5b8bbf66c02..f0dcb8c9744b9074e4e1436e330da2e1110d5777 100644
--- a/doc/man/man1/salloc.1
+++ b/doc/man/man1/salloc.1
@@ -135,7 +135,7 @@ For example: \fB\-\-constraint="opteron&video"\fR or
In the first example, only nodes having both the feature "opteron" AND
the feature "video" will be used.
There is no mechanism to specify that you want one node with feature
-"opteron" and another node with feature "video" in that case that no
+"opteron" and another node with feature "video" in case no
node has both features.
If only one of a set of possible options should be used for all allocated
nodes, then use the OR operator and enclose the options within square brackets.
@@ -757,8 +757,8 @@ When a job is within \fIsig_time\fR seconds of its end time,
send it the signal \fIsig_num\fR.
Due to the resolution of event handling by SLURM, the signal may
be sent up to 60 seconds earlier than specified.
-Both \fIsig_time\fR and \fIsig_num\fR must have integer values
-between zero and 65535.
+\fIsig_num\fR may either be a signal number or name (e.g. "10" or "USR1").
+\fIsig_time\fR must have integer value between zero and 65535.
By default, no signal is sent before the job's end time.
If a \fIsig_num\fR is specified without any \fIsig_time\fR,
the default time will be 60 seconds.
diff --git a/doc/man/man1/sbatch.1 b/doc/man/man1/sbatch.1
index e5fe97639a4d5e92662ec44952687bca661b2d06..65925e84386c1bbbba208532d19c51544988696f 100644
--- a/doc/man/man1/sbatch.1
+++ b/doc/man/man1/sbatch.1
@@ -142,7 +142,7 @@ For example: \fB\-\-constraint="opteron&video"\fR or
In the first example, only nodes having both the feature "opteron" AND
the feature "video" will be used.
There is no mechanism to specify that you want one node with feature
-"opteron" and another node with feature "video" in that case that no
+"opteron" and another node with feature "video" in case no
node has both features.
If only one of a set of possible options should be used for all allocated
nodes, then use the OR operator and enclose the options within square brackets.
@@ -841,8 +841,8 @@ When a job is within \fIsig_time\fR seconds of its end time,
send it the signal \fIsig_num\fR.
Due to the resolution of event handling by SLURM, the signal may
be sent up to 60 seconds earlier than specified.
-Both \fIsig_time\fR and \fIsig_num\fR must have integer values
-between zero and 65535.
+\fIsig_num\fR may either be a signal number or name (e.g. "10" or "USR1").
+\fIsig_time\fR must have integer value between zero and 65535.
By default, no signal is sent before the job's end time.
If a \fIsig_num\fR is specified without any \fIsig_time\fR,
the default time will be 60 seconds.
diff --git a/doc/man/man1/scontrol.1 b/doc/man/man1/scontrol.1
index e29f6a70e29f97687ebf3a486989ce6cef6cb2c6..626466e3ffa2bcfee27e74ef427a6e98ff3336db 100644
--- a/doc/man/man1/scontrol.1
+++ b/doc/man/man1/scontrol.1
@@ -381,10 +381,27 @@ The second number of the signal that caused the process to terminate if
it was terminated by a signal.
.TP
\fIFeatures\fP=<features>
-Set the job's required node features. 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=".
+Set the job's required node features.
+The list of features may include multiple feature names separated
+by ampersand (AND) and/or vertical bar (OR) operators.
+For example: \fBFeatures="opteron&video"\fR or \fBFeatures="fast|faster"\fR.
+In the first example, only nodes having both the feature "opteron" AND
+the feature "video" will be used.
+There is no mechanism to specify that you want one node with feature
+"opteron" and another node with feature "video" in case no
+node has both features.
+If only one of a set of possible options should be used for all allocated
+nodes, then use the OR operator and enclose the options within square brackets.
+For example: "\fBFeatures=[rack1|rack2|rack3|rack4]"\fR might
+be used to specify that all nodes must be allocated on a single rack of
+the cluster, but any of those four racks can be used.
+A request can also specify the number of nodes needed with some feature
+by appending an asterisk and count after the feature name.
+For example "\fBFeatures=graphics*4"\fR
+indicates that at least four allocated nodes must have the feature "graphics."
+Constraints with node counts may only be combined with AND operators.
+Value may be cleared with blank data value, for example "Features=".
+
.TP
\fIGeometry\fP=<geo>
Reset the required job geometry.
diff --git a/doc/man/man1/sinfo.1 b/doc/man/man1/sinfo.1
index 7cf092586e8f3e19621ef1fd5b8b00ac106d70f9..24578808a5bd302136ca5a195942020692ae5569 100644
--- a/doc/man/man1/sinfo.1
+++ b/doc/man/man1/sinfo.1
@@ -82,7 +82,7 @@ when running with various options are
"%9P %5a %.10l %.5D %6t %N"
.TP
.I "\-\-summarize"
-"%9P %5a %.10l %15F %N"
+"%9P %5a %.10l %16F %N"
.TP
.I "\-\-long"
"%9P %5a %.10l %.8s %4r %5h %10g %.5D %11T %N"
diff --git a/doc/man/man1/srun.1 b/doc/man/man1/srun.1
index 3e58d440d82f93e5a8debc757b4cca2777fbe682..364c54043f23431b3f6532ee7c00a144b048e85a 100644
--- a/doc/man/man1/srun.1
+++ b/doc/man/man1/srun.1
@@ -131,7 +131,7 @@ For example: \fB\-\-constraint="opteron&video"\fR or
In the first example, only nodes having both the feature "opteron" AND
the feature "video" will be used.
There is no mechanism to specify that you want one node with feature
-"opteron" and another node with feature "video" in that case that no
+"opteron" and another node with feature "video" in case no
node has both features.
If only one of a set of possible options should be used for all allocated
nodes, then use the OR operator and enclose the options within square brackets.
@@ -953,8 +953,8 @@ When a job is within \fIsig_time\fR seconds of its end time,
send it the signal \fIsig_num\fR.
Due to the resolution of event handling by SLURM, the signal may
be sent up to 60 seconds earlier than specified.
-Both \fIsig_time\fR and \fIsig_num\fR must have integer values
-between zero and 65535.
+\fIsig_num\fR may either be a signal number or name (e.g. "10" or "USR1").
+\fIsig_time\fR must have integer value between zero and 65535.
By default, no signal is sent before the job's end time.
If a \fIsig_num\fR is specified without any \fIsig_time\fR,
the default time will be 60 seconds.
diff --git a/doc/man/man5/slurm.conf.5 b/doc/man/man5/slurm.conf.5
index a5f683cdb8a7a3d3e439aac851663f7ef9196d75..e6f25ed1fecaf0bbcb8447e35db2f57399df5dd3 100644
--- a/doc/man/man5/slurm.conf.5
+++ b/doc/man/man5/slurm.conf.5
@@ -1685,32 +1685,48 @@ Fully qualified pathname of a program to be execute as the slurm job's
owner prior to initiation of each task.
Besides the normal environment variables, this has SLURM_TASK_PID
available to identify the process ID of the task being started.
-Standard output from this program of the form
-"export NAME=value" will be used to set environment variables
-for the task being spawned.
-Standard output from this program of the form
-"print ..." will cause that line (without the leading "print ")
+Standard output from this program can be used to control the environment
+variables and output for the user program.
+.RS
+.TP 20
+\fBexport NAME=value\fR
+Will set environment variables for the task being spawned.
+Everything after the equal sign to the end of the
+line will be used as the value for the environment variable.
+Exporting of functions is not currently supported.
+.TP
+\fBprint ...\fR
+Will cause that line (without the leading "print ")
to be printed to the job's standard output.
+.TP
+\fBunset NAME\fR
+Will clear environment variables for the task being spawned.
+.TP
The order of task prolog/epilog execution is as follows:
-.RS
.TP
-\fB1. pre_launch()\fR: function in TaskPlugin
+\fB1. pre_launch()\fR
+Function in TaskPlugin
.TP
-\fB2. TaskProlog\fR: system\-wide per task program defined in slurm.conf
+\fB2. TaskProlog\fR
+System\-wide per task program defined in slurm.conf
.TP
-\fB3. user prolog\fR: job step specific task program defined using
+\fB3. user prolog\fR
+Job step specific task program defined using
\fBsrun\fR's \fB\-\-task\-prolog\fR option or \fBSLURM_TASK_PROLOG\fR
environment variable
.TP
\fB4.\fR Execute the job step's task
.TP
-\fB5. user epilog\fR: job step specific task program defined using
+\fB5. user epilog\fR
+Job step specific task program defined using
\fBsrun\fR's \fB\-\-task\-epilog\fR option or \fBSLURM_TASK_EPILOG\fR
environment variable
.TP
-\fB6. TaskEpilog\fR: system\-wide per task program defined in slurm.conf
+\fB6. TaskEpilog\fR
+System\-wide per task program defined in slurm.conf
.TP
-\fB7. post_term()\fR: function in TaskPlugin
+\fB7. post_term()\fR
+Function in TaskPlugin
.RE
.TP
diff --git a/etc/init.d.slurm b/etc/init.d.slurm
index 90cc7a1f2cea62e04949608baceb51ec5885f550..d7ab6af2085521e8bc655f87b6a9b860b4aed0f2 100644
--- a/etc/init.d.slurm
+++ b/etc/init.d.slurm
@@ -38,7 +38,10 @@ if [ -f /etc/rc.status ]; then
rc_reset
else
- [ -f /etc/rc.d/init.d/functions ] || exit 0
+ if [ ! -f /etc/rc.d/init.d/functions ]; then
+ echo "Could not find /etc/rc.d/init.d/functions. Is some other daemon launch mechanism used?"
+ exit 1
+ fi
. /etc/rc.d/init.d/functions
SUSE=0
STARTPROC=daemon
@@ -52,6 +55,11 @@ else
RETVAL=0
fi
+if [ ! -x $BINDIR/scontrol ]; then
+ echo "Could not find $BINDIR/scontrol. Bad path?"
+ exit 1
+fi
+
# We can not use a starter program without losing environment
# variables that are critical on Blue Gene systems
if [ -d /bgl/BlueLight/ppcfloor ]; then
@@ -66,7 +74,10 @@ else
SLURMD_OPTIONS=""
fi
-[ -f $CONFDIR/slurm.conf ] || exit 1
+if [ ! -f $CONFDIR/slurm.conf ]; then
+ echo "Could not find $CONFDIR/slurm.conf. Bad path?"
+ exit 1
+fi
# setup library paths for slurm and munge support
export LD_LIBRARY_PATH="$LIBDIR:$LD_LIBRARY_PATH"
diff --git a/etc/init.d.slurmdbd b/etc/init.d.slurmdbd
index e8df33d0b3eb3ff1456c5cd87ffa772e372ababc..18d75f54f964fcf9ff97a0119b02af65da1bd831 100755
--- a/etc/init.d.slurmdbd
+++ b/etc/init.d.slurmdbd
@@ -33,7 +33,10 @@ if [ -f /etc/rc.status ]; then
rc_reset
else
- [ -f /etc/rc.d/init.d/functions ] || exit 0
+ if [ ! -f /etc/rc.d/init.d/functions ]; then
+ echo "Could not find /etc/rc.d/init.d/functions. Is some other daemon launch mechanism used?"
+ exit 1
+ fi
. /etc/rc.d/init.d/functions
SUSE=0
STARTPROC=daemon
@@ -60,7 +63,10 @@ else
SLURMDBD_OPTIONS=""
fi
-[ -f $CONFDIR/slurmdbd.conf ] || exit 1
+if [ ! -f $CONFDIR/slurmdbd.conf ]; then
+ echo "Could not find $CONFDIR/slurmdbd.conf. Bad path?"
+ exit 1
+fi
# setup library paths for slurm and munge support
export LD_LIBRARY_PATH="$LIBDIR:$LD_LIBRARY_PATH"
diff --git a/slurm.spec b/slurm.spec
index 2e4a03b821c3b125c56dfdbb5b9ac7bfba55d6de..abcaec43da0bdf2cb4450e058fb33f5df3415471 100644
--- a/slurm.spec
+++ b/slurm.spec
@@ -1,4 +1,4 @@
-# $Id: slurm.spec 19555 2010-02-19 22:35:07Z da $
+# $Id: slurm.spec 19976 2010-04-07 15:55:58Z jette $
#
# Note that this package is not relocatable
@@ -83,14 +83,14 @@
%endif
Name: slurm
-Version: 2.1.6
+Version: 2.1.7
Release: 1%{?dist}
Summary: Simple Linux Utility for Resource Management
License: GPL
Group: System Environment/Base
-Source: slurm-2.1.6.tar.bz2
+Source: slurm-2.1.7.tar.bz2
BuildRoot: %{_tmppath}/%{name}-%{version}-%{release}
URL: https://computing.llnl.gov/linux/slurm/
@@ -352,7 +352,7 @@ Gives the ability for SLURM to use Berkeley Lab Checkpoint/Restart
#############################################################################
%prep
-%setup -n slurm-2.1.6
+%setup -n slurm-2.1.7
%build
%configure --program-prefix=%{?_program_prefix:%{_program_prefix}} \
@@ -382,13 +382,13 @@ DESTDIR="$RPM_BUILD_ROOT" make install
DESTDIR="$RPM_BUILD_ROOT" make install-contrib
%ifos aix5.3
-mv ${RPM_BUILD_ROOT}%{_bindir}/srun ${RPM_BUILD_ROOT}%{_sbindir}
+ mv ${RPM_BUILD_ROOT}%{_bindir}/srun ${RPM_BUILD_ROOT}%{_sbindir}
+%else
+ if [ -d /etc/init.d ]; then
+ install -D -m755 etc/init.d.slurm $RPM_BUILD_ROOT/etc/init.d/slurm
+ install -D -m755 etc/init.d.slurmdbd $RPM_BUILD_ROOT/etc/init.d/slurmdbd
+ fi
%endif
-
-if [ -d /etc/init.d ]; then
- install -D -m755 etc/init.d.slurm $RPM_BUILD_ROOT/etc/init.d/slurm
- install -D -m755 etc/init.d.slurmdbd $RPM_BUILD_ROOT/etc/init.d/slurmdbd
-fi
install -D -m644 etc/slurm.conf.example ${RPM_BUILD_ROOT}%{_sysconfdir}/slurm.conf.example
install -D -m644 etc/slurmdbd.conf.example ${RPM_BUILD_ROOT}%{_sysconfdir}/slurmdbd.conf.example
install -D -m755 etc/slurm.epilog.clean ${RPM_BUILD_ROOT}%{_sysconfdir}/slurm.epilog.clean
@@ -421,9 +421,8 @@ rm -f ${RPM_BUILD_ROOT}%{_mandir}/man1/srun_cr* ${RPM_BUILD_ROOT}%{_bindir}/srun
# Build conditional file list for main package
LIST=./slurm.files
touch $LIST
-if [ -d /etc/init.d ]; then
- echo "/etc/init.d/slurm" >> $LIST
-fi
+test -f $RPM_BUILD_ROOT/etc/init.d/slurm &&
+ echo /etc/init.d/slurm >> $LIST
%if %{slurm_with aix}
install -D -m644 etc/federation.conf.example ${RPM_BUILD_ROOT}%{_sysconfdir}/federation.conf.example
@@ -446,9 +445,8 @@ test -f $RPM_BUILD_ROOT/%{_libdir}/slurm/switch_federation.so &&
LIST=./slurmdbd.files
touch $LIST
-if [ -d /etc/init.d ]; then
- echo "/etc/init.d/slurmdbd" >> $LIST
-fi
+test -f $RPM_BUILD_ROOT/etc/init.d/slurm &&
+ echo /etc/init.d/slurmdbd >> $LIST
LIST=./plugins.files
touch $LIST
diff --git a/src/common/proc_args.c b/src/common/proc_args.c
index 55781b9a49857320b7ec2565712b4beee3942719..d68811a5c04b39826a43e7cecaa9e2750c0895f4 100644
--- a/src/common/proc_args.c
+++ b/src/common/proc_args.c
@@ -751,17 +751,20 @@ int get_signal_opts(char *optarg, uint16_t *warn_signal, uint16_t *warn_time)
if (optarg == NULL)
return -1;
- num = strtol(optarg, &endptr, 10);
- if ((num < 0) || (num > 0x0ffff))
+ endptr = strchr(optarg, '@');
+ if (endptr)
+ endptr[0] = '\0';
+ num = (uint16_t) sig_name2num(optarg);
+ if (endptr)
+ endptr[0] = '@';
+ if ((num < 1) || (num > 0x0ffff))
return -1;
*warn_signal = (uint16_t) num;
- if (endptr[0] == '\0') {
+ if (!endptr) {
*warn_time = 60;
return 0;
}
- if (endptr[0] != '@')
- return -1;
num = strtol(endptr+1, &endptr, 10);
if ((num < 0) || (num > 0x0ffff))
@@ -771,3 +774,46 @@ int get_signal_opts(char *optarg, uint16_t *warn_signal, uint16_t *warn_time)
return 0;
return -1;
}
+
+/* Convert a signal name to it's numeric equivalent.
+ * Return -1 on failure */
+int sig_name2num(char *signal_name)
+{
+ char *sig_name[] = {"HUP", "INT", "QUIT", "KILL", "TERM",
+ "USR1", "USR2", "CONT", NULL};
+ int sig_num[] = {SIGHUP, SIGINT, SIGQUIT, SIGKILL, SIGTERM,
+ SIGUSR1, SIGUSR2, SIGCONT};
+ char *ptr;
+ long tmp;
+ int sig;
+ int i;
+
+ tmp = strtol(signal_name, &ptr, 10);
+ if (ptr != signal_name) { /* found a number */
+ if (xstring_is_whitespace(ptr))
+ sig = (int)tmp;
+ else
+ return 0;
+ } else {
+ ptr = (char *)signal_name;
+ while (isspace(*ptr))
+ ptr++;
+ if (strncasecmp(ptr, "SIG", 3) == 0)
+ ptr += 3;
+ for (i = 0; ; i++) {
+ if (sig_name[i] == NULL)
+ return 0;
+ if (strncasecmp(ptr, sig_name[i],
+ strlen(sig_name[i])) == 0) {
+ /* found the signal name */
+ if (!xstring_is_whitespace(ptr +
+ strlen(sig_name[i])))
+ return 0;
+ sig = sig_num[i];
+ break;
+ }
+ }
+ }
+
+ return sig;
+}
diff --git a/src/common/proc_args.h b/src/common/proc_args.h
index 2e75bd816325ac612664ce2160f459d4721f6536..201377a0e19f0b931008f333ee5444afe1e71d12 100644
--- a/src/common/proc_args.h
+++ b/src/common/proc_args.h
@@ -132,4 +132,8 @@ char *print_geometry(const uint16_t *geometry);
* RET 0 on success, -1 on failure */
int get_signal_opts(char *optarg, uint16_t *warn_signal, uint16_t *warn_time);
+/* Convert a signal name to it's numeric equivalent.
+ * Return 0 on failure */
+int sig_name2num(char *signal_name);
+
#endif /* !_PROC_ARGS_H */
diff --git a/src/plugins/priority/multifactor/priority_multifactor.c b/src/plugins/priority/multifactor/priority_multifactor.c
index 1e114f52018b39071f5733a68d8a4f40bb64864b..aa3bda1aee6bcba4909c29b87cb818e729f209bd 100644
--- a/src/plugins/priority/multifactor/priority_multifactor.c
+++ b/src/plugins/priority/multifactor/priority_multifactor.c
@@ -794,10 +794,10 @@ static void *_decay_thread(void *no_data)
debug4("adding %f new usage to "
"assoc %u (user='%s' acct='%s') "
"raw usage is now %Lf. Group "
- "wall added %d making it %d.",
+ "wall added %d making it %f.",
real_decay, assoc->id,
assoc->user, assoc->acct,
- assoc->usage_raw, run_delta,
+ assoc->usage_raw, run_decay,
assoc->grp_used_wall);
assoc = assoc->parent_assoc_ptr;
diff --git a/src/plugins/sched/wiki2/job_modify.c b/src/plugins/sched/wiki2/job_modify.c
index 44d6aabcfca6ad24412719b44eea7eb1140f4e8e..6f77e3606009bc79295f85438be627fc95a9b069 100644
--- a/src/plugins/sched/wiki2/job_modify.c
+++ b/src/plugins/sched/wiki2/job_modify.c
@@ -329,15 +329,47 @@ host_fini: if (rc) {
if (IS_JOB_PENDING(job_ptr) && job_ptr->details) {
job_ptr->details->min_nodes = job_desc.min_nodes;
- if (job_ptr->details->max_nodes
- && (job_ptr->details->max_nodes < job_desc.min_nodes))
+ if (job_ptr->details->max_nodes &&
+ (job_ptr->details->max_nodes < job_desc.min_nodes))
job_ptr->details->max_nodes =
job_desc.min_nodes;
info("wiki: change job %u min_nodes to %u",
jobid, new_node_cnt);
#ifdef HAVE_BG
+{
+ static uint16_t cpus_per_node = 0;
+ /* Set the block type to NAV here since if the
+ job size changes from a small block to a
+ regular size block the block types won't
+ jive.
+ */
+ uint16_t conn_type = (uint16_t)SELECT_NAV;
+ /* Also, if geometry is set we need to 0 out
+ the first part of it so the bluegene plugin
+ doesn't look at it any more.
+ */
+ uint16_t req_geometry[SYSTEM_DIMENSIONS] = { 0 };
+
job_ptr->num_procs = job_desc.num_procs;
job_ptr->details->job_min_cpus = job_desc.job_min_cpus;
+
+ if (!cpus_per_node) {
+ select_g_alter_node_cnt(SELECT_GET_NODE_CPU_CNT,
+ &cpus_per_node);
+ }
+ new_node_cnt = job_ptr->num_procs;
+ if (cpus_per_node)
+ new_node_cnt /= cpus_per_node;
+ select_g_select_jobinfo_set(job_ptr->select_jobinfo,
+ SELECT_JOBDATA_NODE_CNT,
+ &new_node_cnt);
+ select_g_select_jobinfo_set(job_ptr->select_jobinfo,
+ SELECT_JOBDATA_CONN_TYPE,
+ &conn_type);
+ select_g_select_jobinfo_set(job_ptr->select_jobinfo,
+ SELECT_JOBDATA_GEOMETRY,
+ &req_geometry);
+}
#endif
last_job_update = now;
update_accounting = true;
diff --git a/src/plugins/sched/wiki2/job_will_run.c b/src/plugins/sched/wiki2/job_will_run.c
index 62ad4fe6e39829df5fade6c489691b041ea8df83..70b83434aa16f76a09ea9f98ec7a7ae857a6e027 100644
--- a/src/plugins/sched/wiki2/job_will_run.c
+++ b/src/plugins/sched/wiki2/job_will_run.c
@@ -48,7 +48,6 @@
#define MAX_JOB_QUEUE 20
-static void _preempt_list_del(void *x);
static char * _will_run_test(uint32_t jobid, time_t start_time,
char *node_list, int *err_code, char **err_msg);
static char * _will_run_test2(uint32_t jobid, time_t start_time,
@@ -410,7 +409,7 @@ static char * _will_run_test2(uint32_t jobid, time_t start_time,
uint32_t *preemptee, int preemptee_cnt,
int *err_code, char **err_msg)
{
- struct job_record *job_ptr = NULL, *pre_ptr, **pre_pptr;
+ struct job_record *job_ptr = NULL, *pre_ptr;
struct part_record *part_ptr;
bitstr_t *avail_bitmap = NULL, *resv_bitmap = NULL;
time_t start_res;
@@ -531,15 +530,10 @@ static char * _will_run_test2(uint32_t jobid, time_t start_time,
}
if (preemptee_cnt) {
- preemptee_candidates = list_create(_preempt_list_del);
+ preemptee_candidates = list_create(NULL);
for (i=0; i<preemptee_cnt; i++) {
- pre_ptr = find_job_record(preemptee[i]);
- if (pre_ptr) {
- pre_pptr = xmalloc(sizeof(struct
- job_record *));
- pre_pptr[0] = pre_ptr;
- list_append(preemptee_candidates, pre_pptr);
- }
+ if ((pre_ptr = find_job_record(preemptee[i])))
+ list_append(preemptee_candidates, pre_ptr);
}
}
@@ -570,15 +564,14 @@ static char * _will_run_test2(uint32_t jobid, time_t start_time,
xfree(hostlist);
if (preempted_jobs) {
- while ((pre_pptr = list_pop(preempted_jobs))) {
+ while ((pre_ptr = list_pop(preempted_jobs))) {
if (pre_cnt++)
sep = ",";
else
sep = " PREEMPT=";
snprintf(tmp_str, sizeof(tmp_str), "%s%u",
- sep, pre_pptr[0]->job_id);
+ sep, pre_ptr->job_id);
xstrcat(reply_msg, tmp_str);
- xfree(pre_pptr);
}
list_destroy(preempted_jobs);
}
@@ -594,11 +587,6 @@ static char * _will_run_test2(uint32_t jobid, time_t start_time,
return reply_msg;
}
-static void _preempt_list_del(void *x)
-{
- xfree(x);
-}
-
/*
* bitmap2wiki_node_name - given a bitmap, build a list of colon separated
* node names (if we can't use node range expressions), or the
diff --git a/src/plugins/select/bluegene/block_allocator/block_allocator.c b/src/plugins/select/bluegene/block_allocator/block_allocator.c
index 45c1f15815442ce1d13b65940200f22a8d842ab0..1bf7220d29f2aa3f7ebe66623a431d32775755fb 100644
--- a/src/plugins/select/bluegene/block_allocator/block_allocator.c
+++ b/src/plugins/select/bluegene/block_allocator/block_allocator.c
@@ -1,7 +1,7 @@
/*****************************************************************************\
* block_allocator.c - Assorted functions for layout of bluegene blocks,
* wiring, mapping for smap, etc.
- * $Id: block_allocator.c 19402 2010-02-04 01:02:41Z jette $
+ * $Id: block_allocator.c 20043 2010-04-16 00:29:44Z da $
*****************************************************************************
* Copyright (C) 2004-2007 The Regents of the University of California.
* Copyright (C) 2008-2009 Lawrence Livermore National Security.
@@ -1348,7 +1348,7 @@ extern int allocate_block(ba_request_t* ba_request, List results)
* Admin wants to remove a previous allocation.
* will allow Admin to delete a previous allocation retrival by letter code.
*/
-extern int remove_block(List nodes, int new_count)
+extern int remove_block(List nodes, int new_count, int conn_type)
{
int dim;
ba_node_t* curr_ba_node = NULL;
@@ -1371,6 +1371,10 @@ extern int remove_block(List nodes, int new_count)
ba_node->used = false;
ba_node->color = 7;
ba_node->letter = '.';
+ /* Small blocks don't use wires, and only have 1 node,
+ so just break. */
+ if(conn_type == SELECT_SMALL)
+ break;
for(dim=0;dim<BA_SYSTEM_DIMENSIONS;dim++) {
curr_switch = &ba_node->axis_switch[dim];
if(curr_switch->int_wire[0].used) {
@@ -1440,7 +1444,7 @@ extern int redo_block(List nodes, int *geo, int conn_type, int new_count)
if(!ba_node)
return SLURM_ERROR;
- remove_block(nodes, new_count);
+ remove_block(nodes, new_count, conn_type);
list_delete_all(nodes, &empty_null_destroy_list, "");
name = set_bg_block(nodes, ba_node->coord, geo, conn_type);
@@ -1646,7 +1650,7 @@ extern char *set_bg_block(List results, int *start,
if(!found) {
debug2("trying less efficient code");
- remove_block(results, color_count);
+ remove_block(results, color_count, conn_type);
list_delete_all(results, &empty_null_destroy_list, "");
list_append(results, ba_node);
found = _find_x_path(results, ba_node,
@@ -3440,7 +3444,10 @@ static int _reset_the_path(ba_switch_t *curr_switch, int source,
}
/*set the switch to not be used */
if(!curr_switch->int_wire[source].used) {
- debug("I reached the end, the source isn't used");
+ /* This means something overlapping the removing block
+ already cleared this, or the path just never was
+ complete in the first place. */
+ debug3("I reached the end, the source isn't used");
return 1;
}
curr_switch->int_wire[source].used = 0;
@@ -3704,7 +3711,8 @@ start_again:
}
if(results) {
- remove_block(results, color_count);
+ remove_block(results, color_count,
+ ba_request->conn_type);
list_delete_all(results,
&empty_null_destroy_list, "");
}
diff --git a/src/plugins/select/bluegene/block_allocator/block_allocator.h b/src/plugins/select/bluegene/block_allocator/block_allocator.h
index beeb861b778fad13f5bc55966f5cc5b911ec4e47..b5af1bf23aa53e089a48ec9d60fc873aa06e60bc 100644
--- a/src/plugins/select/bluegene/block_allocator/block_allocator.h
+++ b/src/plugins/select/bluegene/block_allocator/block_allocator.h
@@ -409,7 +409,7 @@ extern int allocate_block(ba_request_t* ba_request, List results);
* Admin wants to remove a previous allocation.
* will allow Admin to delete a previous allocation retrival by letter code.
*/
-extern int remove_block(List nodes, int new_count);
+extern int remove_block(List nodes, int new_count, int conn_type);
/*
* Admin wants to change something about a previous allocation.
diff --git a/src/plugins/select/bluegene/plugin/bg_job_place.c b/src/plugins/select/bluegene/plugin/bg_job_place.c
index 2b4476a9914b993f11b668b89466936a35ca00f7..acd5c679d349e306f2677bcfc8034878b53f2529 100644
--- a/src/plugins/select/bluegene/plugin/bg_job_place.c
+++ b/src/plugins/select/bluegene/plugin/bg_job_place.c
@@ -943,7 +943,7 @@ static int _find_best_block_match(List block_list,
request.avail_node_bitmap = slurm_block_bitmap;
select_g_select_jobinfo_get(job_ptr->select_jobinfo,
- SELECT_JOBDATA_MAX_CPUS, &max_cpus);
+ SELECT_JOBDATA_MAX_CPUS, &max_cpus);
/* since we only look at procs after this and not nodes we
* need to set a max_cpus if given
*/
@@ -994,12 +994,19 @@ static int _find_best_block_match(List block_list,
bg_record->bg_block_id, 1)
!= SLURM_SUCCESS) {
/* check_block_bp_states will
- already set things in an
- error state, so we don't
- have to do that here.
+ set this block in the main
+ list to an error state, but
+ we aren't looking
+ at the main list, so we
+ need to set this copy of
+ the block in an
+ error state as well.
*/
+ bg_record->job_running =
+ BLOCK_ERROR_STATE;
+ bg_record->state = RM_PARTITION_ERROR;
error("_find_best_block_match: Picked "
- "block had some issues with "
+ "block (%s) had some issues with "
"hardware, trying a different "
"one.",
bg_record->bg_block_id);
diff --git a/src/plugins/select/bluegene/plugin/bg_job_run.c b/src/plugins/select/bluegene/plugin/bg_job_run.c
index b702ca459fe9a58b3fa92d97e46fcc95c7803bfd..d69e4addda3c0623c4e21230326786b0bf04f6e8 100644
--- a/src/plugins/select/bluegene/plugin/bg_job_run.c
+++ b/src/plugins/select/bluegene/plugin/bg_job_run.c
@@ -2,7 +2,7 @@
* bg_job_run.c - blue gene job execution (e.g. initiation and termination)
* functions.
*
- * $Id: bg_job_run.c 19700 2010-03-08 18:50:55Z da $
+ * $Id: bg_job_run.c 20011 2010-04-13 19:10:21Z da $
*****************************************************************************
* Copyright (C) 2004-2006 The Regents of the University of California.
* Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER).
@@ -96,7 +96,7 @@ static pthread_cond_t agent_cond = PTHREAD_COND_INITIALIZER;
static int agent_cnt = 0;
#ifdef HAVE_BG_FILES
-static int _remove_job(db_job_id_t job_id);
+static int _remove_job(db_job_id_t job_id, char *block_id);
#endif
static void _bg_list_del(void *x);
@@ -113,14 +113,15 @@ static void _term_agent(bg_update_t *bg_update_ptr);
#ifdef HAVE_BG_FILES
/* Kill a job and remove its record from MMCS */
-static int _remove_job(db_job_id_t job_id)
+static int _remove_job(db_job_id_t job_id, char *block_id)
{
int rc;
int count = 0;
rm_job_t *job_rec = NULL;
rm_job_state_t job_state;
+ bool is_history = false;
- debug("removing job %d from MMCS", job_id);
+ debug("removing job %d from MMCS on block %s", job_id, block_id);
while(1) {
if (count)
sleep(POLL_INTERVAL);
@@ -152,22 +153,46 @@ static int _remove_job(db_job_id_t job_id)
continue;
}
+ /* If this job is in the history table we
+ should just exit here since it is marked
+ incorrectly */
+ if ((rc = bridge_get_data(job_rec, RM_JobInHist,
+ &is_history))
+ != STATUS_OK) {
+ (void) bridge_free_job(job_rec);
+ if (rc == JOB_NOT_FOUND) {
+ debug("job %d removed from MMCS", job_id);
+ return STATUS_OK;
+ }
+
+ error("bridge_get_data(RM_JobInHist) for jobid=%d "
+ "%s", job_id, bg_err_str(rc));
+ continue;
+ }
+
if ((rc = bridge_free_job(job_rec)) != STATUS_OK)
error("bridge_free_job: %s", bg_err_str(rc));
- debug2("job %d is in state %d", job_id, job_state);
+ debug2("job %d on block %s is in state %d history %d",
+ job_id, block_id, job_state, is_history);
/* check the state and process accordingly */
- if(job_state == RM_JOB_TERMINATED)
+ if(is_history) {
+ debug2("Job %d on block %s isn't in the "
+ "active job table anymore, final state was %d",
+ job_id, block_id, job_state);
+ return STATUS_OK;
+ } else if(job_state == RM_JOB_TERMINATED)
return STATUS_OK;
else if(job_state == RM_JOB_DYING) {
if(count > MAX_POLL_RETRIES)
- error("Job %d isn't dying, trying for "
- "%d seconds", job_id,
- count*POLL_INTERVAL);
+ error("Job %d on block %s isn't dying, "
+ "trying for %d seconds", job_id,
+ block_id, count*POLL_INTERVAL);
continue;
} else if(job_state == RM_JOB_ERROR) {
- error("job %d is in a error state.", job_id);
+ error("job %d on block %s is in a error state.",
+ job_id, block_id);
//free_bg_block();
return STATUS_OK;
@@ -179,7 +204,10 @@ static int _remove_job(db_job_id_t job_id)
* process gets killed with a SIGTERM. In the case of
* bridge_cancel_job it always gets killed with a
* SIGKILL. From IBM's point of view that is a bad
- * deally, so we are going to use signal ;).
+ * deally, so we are going to use signal ;). Sending
+ * a SIGKILL will kill the mpirun front end process,
+ * and if you kill that jobs will never get cleaned up and
+ * you end up with ciod unreacahble on the next job.
*/
// rc = bridge_cancel_job(job_id);
@@ -187,16 +215,23 @@ static int _remove_job(db_job_id_t job_id)
if (rc != STATUS_OK) {
if (rc == JOB_NOT_FOUND) {
- debug("job %d removed from MMCS", job_id);
+ debug("job %d on block %s removed from MMCS",
+ job_id, block_id);
return STATUS_OK;
}
if(rc == INCOMPATIBLE_STATE)
- debug("job %d is in an INCOMPATIBLE_STATE",
- job_id);
+ debug("job %d on block %s is in an "
+ "INCOMPATIBLE_STATE",
+ job_id, block_id);
else
error("bridge_signal_job(%d): %s", job_id,
bg_err_str(rc));
- }
+ } else if(count > MAX_POLL_RETRIES)
+ error("Job %d on block %s is in state %d and "
+ "isn't dying, and doesn't appear to be "
+ "responding to SIGTERM, trying for %d seconds",
+ job_id, block_id, job_state, count*POLL_INTERVAL);
+
}
error("Failed to remove job %d from MMCS", job_id);
@@ -378,7 +413,7 @@ static void _remove_jobs_on_block_and_reset(rm_job_list_t *job_list,
continue;
}
debug2("got job_id %d",job_id);
- if((rc = _remove_job(job_id)) == INTERNAL_ERROR) {
+ if((rc = _remove_job(job_id, block_id)) == INTERNAL_ERROR) {
job_remove_failed = 1;
break;
}
diff --git a/src/plugins/select/bluegene/plugin/bg_record_functions.c b/src/plugins/select/bluegene/plugin/bg_record_functions.c
index 774370597760d7eda4063c0a8f54c160efa5fa73..fe79d6e5f42c6b6917e80fde1786f3cb362db2ae 100644
--- a/src/plugins/select/bluegene/plugin/bg_record_functions.c
+++ b/src/plugins/select/bluegene/plugin/bg_record_functions.c
@@ -493,10 +493,15 @@ extern int bg_record_cmpf_inc(bg_record_t* rec_a, bg_record_t* rec_b)
int size_a = rec_a->node_cnt;
int size_b = rec_b->node_cnt;
- if (size_a < size_b)
- return -1;
- else if (size_a > size_b)
- return 1;
+ /* We only look at this if we are ordering blocks larger than
+ * a midplane, order of ionodes is how we order otherwise. */
+ if((size_a >= bg_conf->bp_node_cnt)
+ || (size_b >= bg_conf->bp_node_cnt)) {
+ if (size_a < size_b)
+ return -1;
+ else if (size_a > size_b)
+ return 1;
+ }
if(rec_a->nodes && rec_b->nodes) {
size_a = strcmp(rec_a->nodes, rec_b->nodes);
diff --git a/src/plugins/select/bluegene/plugin/block_sys.c b/src/plugins/select/bluegene/plugin/block_sys.c
index 264756d2434cee551323f232bc11d7a4a684b8b6..093a300035dad01b45e963b340b698b5accb4deb 100755
--- a/src/plugins/select/bluegene/plugin/block_sys.c
+++ b/src/plugins/select/bluegene/plugin/block_sys.c
@@ -1,7 +1,7 @@
/*****************************************************************************\
* block_sys.c - component used for wiring up the blocks
*
- * $Id: block_sys.c 19849 2010-03-23 19:10:13Z da $
+ * $Id: block_sys.c 19995 2010-04-09 20:55:46Z da $
*****************************************************************************
* Copyright (C) 2004 The Regents of the University of California.
* Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER).
@@ -267,7 +267,8 @@ static int _post_allocate(bg_record_t *bg_record)
#ifdef HAVE_BG_FILES
#ifdef HAVE_BGL
-static int _find_nodecard(rm_partition_t *block_ptr, int *nc_id)
+extern int find_nodecard_num(rm_partition_t *block_ptr, rm_nodecard_t *ncard,
+ int *nc_id)
{
char *my_card_name = NULL;
char *card_name = NULL;
@@ -276,19 +277,12 @@ static int _find_nodecard(rm_partition_t *block_ptr, int *nc_id)
int i=0;
int rc;
rm_nodecard_list_t *ncard_list = NULL;
- rm_nodecard_t *ncard = NULL;
rm_BP_t *curr_bp = NULL;
+ rm_nodecard_t *ncard2;
xassert(block_ptr);
xassert(nc_id);
- if((rc = bridge_get_data(block_ptr,
- RM_PartitionFirstNodeCard,
- &ncard))
- != STATUS_OK) {
- error("bridge_get_data(RM_FirstCard): %s",
- bg_err_str(rc));
- }
if((rc = bridge_get_data(ncard,
RM_NodeCardID,
&my_card_name))
@@ -330,7 +324,7 @@ static int _find_nodecard(rm_partition_t *block_ptr, int *nc_id)
if ((rc =
bridge_get_data(ncard_list,
RM_NodeCardListNext,
- &ncard)) != STATUS_OK) {
+ &ncard2)) != STATUS_OK) {
error("bridge_get_data"
"(RM_NodeCardListNext): %s",
rc);
@@ -340,7 +334,7 @@ static int _find_nodecard(rm_partition_t *block_ptr, int *nc_id)
} else {
if ((rc = bridge_get_data(ncard_list,
RM_NodeCardListFirst,
- &ncard)) != STATUS_OK) {
+ &ncard2)) != STATUS_OK) {
error("bridge_get_data"
"(RM_NodeCardListFirst: %s",
rc);
@@ -348,7 +342,7 @@ static int _find_nodecard(rm_partition_t *block_ptr, int *nc_id)
goto cleanup;
}
}
- if ((rc = bridge_get_data(ncard,
+ if ((rc = bridge_get_data(ncard2,
RM_NodeCardID,
&card_name)) != STATUS_OK) {
error("bridge_get_data(RM_NodeCardID: %s",
@@ -593,7 +587,7 @@ int read_bg_blocks(List curr_block_list)
nc_id = 0;
if(nc_cnt == 1)
- _find_nodecard(block_ptr, &nc_id);
+ find_nodecard_num(block_ptr, ncard, &nc_id);
bg_record->node_cnt =
nc_cnt * bg_conf->nodecard_node_cnt;
diff --git a/src/plugins/select/bluegene/plugin/bluegene.h b/src/plugins/select/bluegene/plugin/bluegene.h
index 55f21f788abffa89f75d1aadcef5d45bb6735cf5..bff5f46f14ce6b26b326e53bd191ebe7aed1d598 100644
--- a/src/plugins/select/bluegene/plugin/bluegene.h
+++ b/src/plugins/select/bluegene/plugin/bluegene.h
@@ -1,7 +1,7 @@
/*****************************************************************************\
* bluegene.h - header for blue gene configuration processing module.
*
- * $Id: bluegene.h 19095 2009-12-01 22:59:18Z da $
+ * $Id: bluegene.h 19995 2010-04-09 20:55:46Z da $
*****************************************************************************
* Copyright (C) 2004 The Regents of the University of California.
* Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER).
@@ -168,6 +168,12 @@ extern int validate_current_blocks(char *dir);
/* block_sys.c */
/*****************************************************/
+#ifdef HAVE_BG_FILES
+#ifdef HAVE_BGL
+extern int find_nodecard_num(rm_partition_t *block_ptr, rm_nodecard_t *ncard,
+ int *nc_id);
+#endif
+#endif
extern int configure_block(bg_record_t * bg_conf_record);
extern int read_bg_blocks();
extern int load_state_file(List curr_block_list, char *dir_name);
diff --git a/src/plugins/select/bluegene/plugin/dynamic_block.c b/src/plugins/select/bluegene/plugin/dynamic_block.c
index 627a6c045a8ab5626b77b5e52ed1423d8a42821e..bca33d8e548299edfd6e0b77080a49a3f6bb65bb 100644
--- a/src/plugins/select/bluegene/plugin/dynamic_block.c
+++ b/src/plugins/select/bluegene/plugin/dynamic_block.c
@@ -271,7 +271,8 @@ extern List create_dynamic_block(List block_list,
debug3("removing %s for request %d",
bg_record->nodes, request->size);
- remove_block(bg_record->bg_block_list, (int)NO_VAL);
+ remove_block(bg_record->bg_block_list, (int)NO_VAL,
+ (int)bg_record->conn_type);
/* need to set any unusable nodes that this last block
used */
removable_set_bps(unusable_nodes);
@@ -404,7 +405,8 @@ extern bg_record_t *create_small_record(bg_record_t *bg_record,
found_record->ionode_bitmap = bit_copy(ionodes);
bit_fmt(bitstring, BITSIZE, found_record->ionode_bitmap);
found_record->ionodes = xstrdup(bitstring);
-
+ debug4("made small block of %s[%s]",
+ found_record->nodes, found_record->ionodes);
return found_record;
}
@@ -608,6 +610,7 @@ static int _breakup_blocks(List block_list, List new_blocks,
char tmp_char[256];
bitstr_t *ionodes = bit_alloc(bg_conf->numpsets);
int cnodes = request->procs / bg_conf->cpu_ratio;
+ int curr_bp_bit = -1;
debug2("proc count = %d cnodes = %d size = %d",
request->procs, cnodes, request->size);
@@ -676,6 +679,36 @@ static int _breakup_blocks(List block_list, List new_blocks,
if(bg_record->node_cnt < cnodes) {
char bitstring[BITSIZE];
bitstr_t *bitstr = NULL;
+ int num_over = 0;
+ int num_cnodes = bg_record->node_cnt;
+ int rec_bp_bit = bit_ffs(bg_record->bitmap);
+
+ if(curr_bp_bit != rec_bp_bit) {
+ /* Got a different node than
+ * previously, since the list should
+ * be in order of nodes for small blocks
+ * just clear here since the last node
+ * doesn't have any more. */
+ curr_bp_bit = rec_bp_bit;
+ bit_nclear(ionodes, 0, (bg_conf->numpsets-1));
+ total_cnode_cnt = 0;
+ }
+
+ /* On really busy systems we can get
+ overlapping blocks here. If that is the
+ case only add that which doesn't overlap.
+ */
+ if((num_over = bit_overlap(
+ ionodes, bg_record->ionode_bitmap))) {
+ /* Since the smallest block size is
+ the number of cnodes in an io node,
+ just multiply the num_over by that to
+ get the number of cnodes to remove.
+ */
+ if((num_cnodes -=
+ num_over * bg_conf->smallest_block) <= 0)
+ continue;
+ }
bit_or(ionodes, bg_record->ionode_bitmap);
/* check and see if the bits set are a valid
@@ -690,15 +723,16 @@ static int _breakup_blocks(List block_list, List new_blocks,
if(!bitstr) {
bit_nclear(ionodes, 0, (bg_conf->numpsets-1));
bit_or(ionodes, bg_record->ionode_bitmap);
- total_cnode_cnt = bg_record->node_cnt;
+ total_cnode_cnt = num_cnodes =
+ bg_record->node_cnt;
} else
- total_cnode_cnt += bg_record->node_cnt;
+ total_cnode_cnt += num_cnodes;
bit_fmt(bitstring, BITSIZE, ionodes);
- debug2("1 adding %s %d got %d set "
+ debug2("1 adding %s %s %d got %d set "
"ionodes %s total is %s",
- bg_record->bg_block_id,
- bg_record->node_cnt, total_cnode_cnt,
+ bg_record->bg_block_id, bg_record->nodes,
+ num_cnodes, total_cnode_cnt,
bg_record->ionodes, bitstring);
if(total_cnode_cnt == cnodes) {
request->save_name = xstrdup_printf(
diff --git a/src/plugins/select/bluegene/plugin/state_test.c b/src/plugins/select/bluegene/plugin/state_test.c
index 600de3398656829abdffcd25bab907cebb58a772..03c3595bcb77899c99fc8611dea5afe4d049231c 100644
--- a/src/plugins/select/bluegene/plugin/state_test.c
+++ b/src/plugins/select/bluegene/plugin/state_test.c
@@ -2,7 +2,7 @@
* state_test.c - Test state of Bluegene base partitions and switches.
* DRAIN nodes in SLURM that are not usable.
*
- * $Id: state_test.c 19755 2010-03-16 19:15:43Z da $
+ * $Id: state_test.c 19998 2010-04-12 19:18:46Z da $
*****************************************************************************
* Copyright (C) 2004-2007 The Regents of the University of California.
* Copyright (C) 2008-2009 Lawrence Livermore National Security.
@@ -122,6 +122,17 @@ static void _configure_node_down(rm_bp_id_t bp_id, my_bluegene_t *my_bg)
continue;
}
+ /* make sure we have this midplane in the system */
+ if(bp_loc.X >= DIM_SIZE[X]
+ || bp_loc.Y >= DIM_SIZE[Y]
+ || bp_loc.Z >= DIM_SIZE[Z]) {
+ debug4("node %s%c%c%c isn't configured",
+ bg_conf->slurm_node_prefix,
+ alpha_num[bp_loc.X], alpha_num[bp_loc.Y],
+ alpha_num[bp_loc.Z]);
+ continue;
+ }
+
snprintf(bg_down_node, sizeof(bg_down_node), "%s%c%c%c",
bg_conf->slurm_node_prefix,
alpha_num[bp_loc.X], alpha_num[bp_loc.Y],
@@ -141,6 +152,136 @@ static void _configure_node_down(rm_bp_id_t bp_id, my_bluegene_t *my_bg)
}
}
+static char *_get_bp_node_name(rm_BP_t *bp_ptr)
+{
+ rm_location_t bp_loc;
+ int rc;
+
+ errno = SLURM_SUCCESS;
+
+ if ((rc = bridge_get_data(bp_ptr, RM_BPLoc, &bp_loc))
+ != STATUS_OK) {
+ error("bridge_get_data(RM_BPLoc): %s", bg_err_str(rc));
+ errno = SLURM_ERROR;
+ return NULL;
+ }
+
+ /* make sure we have this midplane in the system */
+ if(bp_loc.X >= DIM_SIZE[X]
+ || bp_loc.Y >= DIM_SIZE[Y]
+ || bp_loc.Z >= DIM_SIZE[Z]) {
+ debug4("node %s%c%c%c isn't configured",
+ bg_conf->slurm_node_prefix,
+ alpha_num[bp_loc.X], alpha_num[bp_loc.Y],
+ alpha_num[bp_loc.Z]);
+ return NULL;
+ }
+
+ return xstrdup_printf("%s%c%c%c",
+ bg_conf->slurm_node_prefix,
+ alpha_num[bp_loc.X], alpha_num[bp_loc.Y],
+ alpha_num[bp_loc.Z]);
+}
+
+/* To fake a nodecard down do this on the service node.
+db2 "update bg{l|p}nodecard set status = 'E' where location =
+'Rxx-Mx-Nx' and status='A'"
+Reverse the A, and E to bring it back up.
+*/
+static int _test_nodecard_state(rm_nodecard_t *ncard, int nc_id,
+ char *node_name, bool slurmctld_locked)
+{
+ int rc = SLURM_SUCCESS;
+ rm_nodecard_id_t nc_name = NULL;
+ rm_nodecard_state_t state;
+ int io_start = 0;
+
+ if ((rc = bridge_get_data(ncard,
+ RM_NodeCardState,
+ &state)) != STATUS_OK) {
+ error("bridge_get_data(RM_NodeCardState: %s",
+ rc);
+ return SLURM_ERROR;
+ }
+
+ if(state == RM_NODECARD_UP)
+ return SLURM_SUCCESS;
+
+ if ((rc = bridge_get_data(ncard,
+ RM_NodeCardID,
+ &nc_name)) != STATUS_OK) {
+ error("bridge_get_data(RM_NodeCardID): %d", rc);
+ return SLURM_ERROR;
+ }
+
+ if(!nc_name) {
+ error("We didn't get an RM_NodeCardID but rc was STATUS_OK?");
+ return SLURM_ERROR;
+ }
+
+#ifdef HAVE_BGL
+ if ((rc = bridge_get_data(ncard,
+ RM_NodeCardQuarter,
+ &io_start)) != STATUS_OK) {
+ error("bridge_get_data(CardQuarter): %d", rc);
+ rc = SLURM_ERROR;
+ goto clean_up;
+ }
+ io_start *= bg_conf->quarter_ionode_cnt;
+ io_start += bg_conf->nodecard_ionode_cnt * (nc_id%4);
+#else
+ /* From the first nodecard id we can figure
+ out where to start from with the alloc of ionodes.
+ */
+ io_start = atoi((char*)nc_name+1);
+ io_start *= bg_conf->io_ratio;
+#endif
+ /* On small systems with less than a midplane the
+ database may see the nodecards there but in missing
+ state. To avoid getting a bunch of warnings here just
+ skip over the ones missing.
+ */
+ if(io_start >= bg_conf->numpsets) {
+ rc = SLURM_SUCCESS;
+ if(state == RM_NODECARD_MISSING) {
+ debug3("Nodecard %s is missing",
+ nc_name);
+ } else {
+ error("We don't have the system configured "
+ "for this nodecard %s, we only have "
+ "%d ionodes and this starts at %d",
+ nc_name, io_start, bg_conf->numpsets);
+ }
+ goto clean_up;
+ }
+
+ /* if(!ionode_bitmap) */
+ /* ionode_bitmap = bit_alloc(bg_conf->numpsets); */
+ /* info("setting %s start %d of %d", */
+ /* nc_name, io_start, bg_conf->numpsets); */
+ /* bit_nset(ionode_bitmap, io_start, io_start+io_cnt); */
+
+ /* we have to handle each nodecard separately to make
+ sure we don't create holes in the system */
+ if(down_nodecard(node_name, io_start, slurmctld_locked)
+ == SLURM_SUCCESS) {
+ debug("nodecard %s on %s is in an error state",
+ nc_name, node_name);
+ } else
+ debug2("nodecard %s on %s is in an error state, "
+ "but error was returned when trying to make it so",
+ nc_name, node_name);
+
+ /* Here we want to keep track of any nodecard that
+ isn't up and return error if it is in the system. */
+ rc = SLURM_ERROR;
+
+clean_up:
+
+ free(nc_name);
+ return rc;
+}
+
/*
* This could potentially lock the node lock in the slurmctld with
* slurm_drain_node, so if nodes_locked is called we will call the
@@ -149,17 +290,14 @@ static void _configure_node_down(rm_bp_id_t bp_id, my_bluegene_t *my_bg)
static int _test_down_nodecards(rm_BP_t *bp_ptr, bool slurmctld_locked)
{
rm_bp_id_t bp_id = NULL;
- rm_nodecard_id_t nc_name = NULL;
int num = 0;
int marked_down = 0;
int i=0;
int rc = SLURM_SUCCESS;
rm_nodecard_list_t *ncard_list = NULL;
rm_nodecard_t *ncard = NULL;
- rm_nodecard_state_t state;
//bitstr_t *ionode_bitmap = NULL;
//bg_record_t *bg_record = NULL;
- int *coord = NULL;
char *node_name = NULL;
//int bp_bit = 0;
//int io_cnt = 1;
@@ -183,28 +321,7 @@ static int _test_down_nodecards(rm_BP_t *bp_ptr, bool slurmctld_locked)
goto clean_up;
}
- coord = find_bp_loc(bp_id);
- if(!coord) {
- error("Could not find coordinates for "
- "BP ID %s", (char *) bp_id);
- rc = SLURM_ERROR;
- goto clean_up;
- }
-
- /* make sure we have this midplane in the system */
- if(coord[X] >= DIM_SIZE[X]
- || coord[Y] >= DIM_SIZE[Y]
- || coord[Z] >= DIM_SIZE[Z]) {
- debug4("node %s isn't configured", bp_id);
- rc = SLURM_SUCCESS;
- goto clean_up;
- }
-
- node_name = xstrdup_printf("%s%c%c%c",
- bg_conf->slurm_node_prefix,
- alpha_num[coord[X]],
- alpha_num[coord[Y]],
- alpha_num[coord[Z]]);
+ node_name = _get_bp_node_name(bp_ptr);
if((rc = bridge_get_data(ncard_list, RM_NodeCardListSize, &num))
!= STATUS_OK) {
@@ -215,8 +332,6 @@ static int _test_down_nodecards(rm_BP_t *bp_ptr, bool slurmctld_locked)
}
for(i=0; i<num; i++) {
- int io_start = 0;
-
if (i) {
if ((rc = bridge_get_data(ncard_list,
RM_NodeCardListNext,
@@ -238,83 +353,10 @@ static int _test_down_nodecards(rm_BP_t *bp_ptr, bool slurmctld_locked)
goto clean_up;
}
}
- if ((rc = bridge_get_data(ncard,
- RM_NodeCardState,
- &state)) != STATUS_OK) {
- error("bridge_get_data(RM_NodeCardState: %s",
- rc);
- rc = SLURM_ERROR;
- goto clean_up;
- }
-
- if(state == RM_NODECARD_UP)
- continue;
-
- /* Here we want to keep track of any nodecard that
- isn't up and return error if this is not 0 since
- we could be checking to see if we could run here. */
- marked_down++;
-
- if ((rc = bridge_get_data(ncard,
- RM_NodeCardID,
- &nc_name)) != STATUS_OK) {
- error("bridge_get_data(RM_NodeCardID): %d",rc);
- rc = SLURM_ERROR;
- goto clean_up;
- }
-
- if(!nc_name) {
- rc = SLURM_ERROR;
- goto clean_up;
- }
-#ifdef HAVE_BGL
- if ((rc = bridge_get_data(ncard,
- RM_NodeCardQuarter,
- &io_start)) != STATUS_OK) {
- error("bridge_get_data(CardQuarter): %d",rc);
- goto clean_up;
- }
- io_start *= bg_conf->quarter_ionode_cnt;
- io_start += bg_conf->nodecard_ionode_cnt * (i%4);
-#else
- /* From the first nodecard id we can figure
- out where to start from with the alloc of ionodes.
- */
- io_start = atoi((char*)nc_name+1);
- io_start *= bg_conf->io_ratio;
-#endif
- /* On small systems with less than a midplane the
- database may see the nodecards there but in missing
- state. To avoid getting a bunch of warnings here just
- skip over the ones missing.
- */
- if(io_start >= bg_conf->numpsets) {
- if(state == RM_NODECARD_MISSING) {
- debug3("Nodecard %s is missing continue",
- nc_name);
- } else {
- error("We don't have the system configured "
- "for this nodecard %s, we only have "
- "%d ionodes and this starts at %d",
- nc_name, io_start, bg_conf->numpsets);
- }
- free(nc_name);
- continue;
- }
-/* if(!ionode_bitmap) */
-/* ionode_bitmap = bit_alloc(bg_conf->numpsets); */
-/* info("setting %s start %d of %d", */
-/* nc_name, io_start, bg_conf->numpsets); */
-/* bit_nset(ionode_bitmap, io_start, io_start+io_cnt); */
- /* we have to handle each nodecard separately to make
- sure we don't create holes in the system */
- if(down_nodecard(node_name, io_start, slurmctld_locked)
- == SLURM_SUCCESS) {
- debug("nodecard %s on %s is in an error state",
- nc_name, node_name);
- }
- free(nc_name);
+ if(_test_nodecard_state(ncard, i, node_name, slurmctld_locked)
+ != SLURM_SUCCESS)
+ marked_down++;
}
/* this code is here to bring up a block after it is in an
@@ -374,6 +416,7 @@ clean_up:
/* If we marked any nodecard down we need to state it here */
if((rc == SLURM_SUCCESS) && marked_down)
rc = SLURM_ERROR;
+
return rc;
}
@@ -523,8 +566,9 @@ extern int check_block_bp_states(char *bg_block_id, bool slurmctld_locked)
#ifdef HAVE_BG_FILES
rm_partition_t *block_ptr = NULL;
rm_BP_t *bp_ptr = NULL;
- int bp_cnt = 0;
+ int cnt = 0;
int i = 0;
+ bool small = false;
if ((rc = bridge_get_block(bg_block_id, &block_ptr)) != STATUS_OK) {
error("Block %s doesn't exist.", bg_block_id);
@@ -534,14 +578,103 @@ extern int check_block_bp_states(char *bg_block_id, bool slurmctld_locked)
}
- if ((rc = bridge_get_data(block_ptr, RM_PartitionBPNum, &bp_cnt))
+ if ((rc = bridge_get_data(block_ptr, RM_PartitionSmall, &small))
+ != STATUS_OK) {
+ error("bridge_get_data(RM_PartitionSmall): %s",
+ bg_err_str(rc));
+ rc = SLURM_ERROR;
+
+ goto cleanup;
+ }
+
+ if(small) {
+ rm_nodecard_t *ncard = NULL;
+ char *node_name = NULL;
+
+ /* If this is a small block we can just check the
+ nodecard list of the block.
+ */
+ if((rc = bridge_get_data(block_ptr,
+ RM_PartitionNodeCardNum,
+ &cnt))
+ != STATUS_OK) {
+ error("bridge_get_data(RM_PartitionNodeCardNum): %s",
+ bg_err_str(rc));
+ rc = SLURM_ERROR;
+ goto cleanup;
+ }
+
+ if ((rc = bridge_get_data(block_ptr,
+ RM_PartitionFirstBP,
+ &bp_ptr))
+ != STATUS_OK) {
+ error("bridge_get_data(RM_FirstBP): %s",
+ bg_err_str(rc));
+ rc = SLURM_ERROR;
+ goto cleanup;
+ }
+
+ if(!(node_name = _get_bp_node_name(bp_ptr))) {
+ rc = errno;
+ goto cleanup;
+ }
+
+ for(i=0; i<cnt; i++) {
+ int nc_id = 0;
+ if(i) {
+ if ((rc = bridge_get_data(
+ block_ptr,
+ RM_PartitionNextNodeCard,
+ &ncard))
+ != STATUS_OK) {
+ error("bridge_get_data("
+ "RM_PartitionNextNodeCard): %s",
+ bg_err_str(rc));
+ rc = SLURM_ERROR;
+ break;
+ }
+ } else {
+ if ((rc = bridge_get_data(
+ block_ptr,
+ RM_PartitionFirstNodeCard,
+ &ncard))
+ != STATUS_OK) {
+ error("bridge_get_data("
+ "RM_PartitionFirstNodeCard): %s",
+ bg_err_str(rc));
+ rc = SLURM_ERROR;
+ break;
+ }
+ }
+#ifdef HAVE_BGL
+ find_nodecard_num(block_ptr, ncard, &nc_id);
+#endif
+ /* If we find any nodecards in an error state just
+ break here since we are seeing if we can run. If
+ any nodecard is down this can't happen.
+ */
+ if(_test_nodecard_state(
+ ncard, nc_id, node_name, slurmctld_locked)
+ != SLURM_SUCCESS) {
+ rc = SLURM_ERROR;
+ break;
+ }
+ }
+ xfree(node_name);
+ goto cleanup;
+ }
+
+ /* If this isn't a small block we have to check the list of
+ nodecards on each midplane.
+ */
+ if ((rc = bridge_get_data(block_ptr, RM_PartitionBPNum, &cnt))
!= STATUS_OK) {
error("bridge_get_data(RM_BPNum): %s", bg_err_str(rc));
rc = SLURM_ERROR;
goto cleanup;
}
- for(i=0; i<bp_cnt; i++) {
+ for(i=0; i<cnt; i++) {
if(i) {
if ((rc = bridge_get_data(block_ptr,
RM_PartitionNextBP,
@@ -568,9 +701,11 @@ extern int check_block_bp_states(char *bg_block_id, bool slurmctld_locked)
break here since we are seeing if we can run. If
any nodecard is down this can't happen.
*/
- if((rc = _test_down_nodecards(bp_ptr, slurmctld_locked))
- != SLURM_SUCCESS)
+ if(_test_down_nodecards(bp_ptr, slurmctld_locked)
+ != SLURM_SUCCESS) {
+ rc = SLURM_ERROR;
break;
+ }
}
cleanup:
diff --git a/src/plugins/select/cons_res/job_test.c b/src/plugins/select/cons_res/job_test.c
index 0fa4b6cd5638188ec721a0cff8a1e47589e60da6..596577adf6c5f04c4e250c52483fe590ddd4df3b 100644
--- a/src/plugins/select/cons_res/job_test.c
+++ b/src/plugins/select/cons_res/job_test.c
@@ -1794,8 +1794,6 @@ extern int cr_job_test(struct job_record *job_ptr, bitstr_t *bitmap,
"idle resources found");
goto alloc_job;
}
- debug3("cons_res: cr_job_test: test 1 fail - "
- "not enough idle resources");
if (job_node_req == NODE_CR_ONE_ROW) {
/* This job CANNOT share CPUs regardless of priority,
@@ -1803,8 +1801,12 @@ extern int cr_job_test(struct job_record *job_ptr, bitstr_t *bitmap,
* addressed in _verify_node_state() and job preemption
* removes jobs from simulated resource allocation map
* before this point. */
+ debug3("cons_res: cr_job_test: test 1 fail - "
+ "no idle resources available");
goto alloc_job;
}
+ debug3("cons_res: cr_job_test: test 1 fail - "
+ "not enough idle resources");
/*** Step 2 ***/
bit_copybits(bitmap, orig_map);
diff --git a/src/plugins/select/cons_res/select_cons_res.c b/src/plugins/select/cons_res/select_cons_res.c
index ef2e047a568a220c264516e86faffa210e6899d9..62500a7b125fe9728b1a3fd92785f37d64fc553b 100644
--- a/src/plugins/select/cons_res/select_cons_res.c
+++ b/src/plugins/select/cons_res/select_cons_res.c
@@ -1087,9 +1087,8 @@ static uint16_t _get_job_node_req(struct job_record *job_ptr)
if (max_share == 0) /* Partition Shared=EXCLUSIVE */
return NODE_CR_RESERVED;
- /* Partition Shared=FORCE with count > 1 */
- if ((max_share & SHARED_FORCE) &&
- ((max_share & (~SHARED_FORCE)) > 1))
+ /* Partition is Shared=FORCE */
+ if (max_share & SHARED_FORCE)
return NODE_CR_AVAILABLE;
/* Partition is Shared=NO or Shared=YES */
diff --git a/src/salloc/opt.c b/src/salloc/opt.c
index 5b6d23a90eaf00fe3a200be110eec97db94dedef..aab3a2c759e0f17a4c8d68debc55b54f13ae3dc7 100644
--- a/src/salloc/opt.c
+++ b/src/salloc/opt.c
@@ -188,7 +188,6 @@ static bool _opt_verify(void);
static void _proc_get_user_env(char *optarg);
static void _process_env_var(env_vars_t *e, const char *val);
-static int _parse_signal(const char *signal_name);
static void _usage(void);
/*---[ end forward declarations of static functions ]---------------------*/
@@ -737,9 +736,11 @@ void set_options(const int argc, char **argv)
break;
case 'K': /* argument is optional */
if (optarg) {
- opt.kill_command_signal =_parse_signal(optarg);
- if (opt.kill_command_signal == 0)
+ opt.kill_command_signal = sig_name2num(optarg);
+ if (opt.kill_command_signal == 0) {
+ error("Invalid signal name %s", optarg);
exit(error_exit);
+ }
}
opt.kill_command_signal_set = true;
break;
@@ -1564,61 +1565,6 @@ static char *print_constraints()
return buf;
}
-/*
- * Takes a string containing the number or name of a signal and returns
- * the signal number. The signal name is case insensitive, and may be of
- * the form "SIGHUP" or just "HUP".
- *
- * Allowed signal names are HUP, INT, QUIT, KILL, TERM, USR1, USR2, and CONT.
- */
-static int _parse_signal(const char *signal_name)
-{
- char *sig_name[] = {"HUP", "INT", "QUIT", "KILL", "TERM",
- "USR1", "USR2", "CONT", NULL};
- int sig_num[] = {SIGHUP, SIGINT, SIGQUIT, SIGKILL, SIGTERM,
- SIGUSR1, SIGUSR2, SIGCONT};
- char *ptr;
- long tmp;
- int sig;
- int i;
-
- tmp = strtol(signal_name, &ptr, 10);
- if (ptr != signal_name) { /* found a number */
- if (xstring_is_whitespace(ptr)) {
- sig = (int)tmp;
- } else {
- goto fail;
- }
- } else {
- ptr = (char *)signal_name;
- while (isspace(*ptr))
- ptr++;
- if (strncasecmp(ptr, "SIG", 3) == 0)
- ptr += 3;
- for (i = 0; ; i++) {
- if (sig_name[i] == NULL) {
- goto fail;
- }
- if (strncasecmp(ptr, sig_name[i],
- strlen(sig_name[i])) == 0) {
- /* found the signal name */
- if (!xstring_is_whitespace(
- ptr + strlen(sig_name[i]))) {
- goto fail;
- }
- sig = sig_num[i];
- break;
- }
- }
- }
-
- return sig;
-
-fail:
- error("\"%s\" is not a valid signal", signal_name);
- return 0;
-}
-
#define tf_(b) (b == true) ? "true" : "false"
static void _opt_list()
diff --git a/src/salloc/salloc.c b/src/salloc/salloc.c
index 2b05f2078b4a6eaaa052fbae265b3fddb58c25e0..567f6189323aa9ce0d9eecdb5866d5371ec50dc0 100644
--- a/src/salloc/salloc.c
+++ b/src/salloc/salloc.c
@@ -491,6 +491,8 @@ static int _fill_job_desc_from_opts(job_desc_msg_t *desc)
desc->begin_time = opt.begin;
if (opt.account)
desc->account = xstrdup(opt.account);
+ if (opt.acctg_freq >= 0)
+ desc->acctg_freq = opt.acctg_freq;
if (opt.comment)
desc->comment = xstrdup(opt.comment);
if (opt.qos)
diff --git a/src/sinfo/opts.c b/src/sinfo/opts.c
index 5b618ffa097097efbd33b8cf6a2f07df15233c76..7de33968f0975c62caef09da85e63e9dc8c0dcc0 100644
--- a/src/sinfo/opts.c
+++ b/src/sinfo/opts.c
@@ -232,7 +232,7 @@ extern void parse_command_line(int argc, char *argv[])
#ifdef HAVE_BG
params.format = "%9P %.5a %.10l %.32F %N";
#else
- params.format = "%9P %.5a %.10l %.15F %N";
+ params.format = "%9P %.5a %.10l %.16F %N";
#endif
} else if ( params.node_flag ) {
params.node_field_flag = true; /* compute size later */
diff --git a/src/sinfo/sinfo.c b/src/sinfo/sinfo.c
index 49c6c40499011b5154f35cdd9028f7c7d2199f9b..108a069855c32d740bb3a9fced1be66dbbbad6e4 100644
--- a/src/sinfo/sinfo.c
+++ b/src/sinfo/sinfo.c
@@ -59,10 +59,6 @@ struct sinfo_parameters params;
static int g_node_scaling = 1;
-#ifdef HAVE_BG
-static int cpus_per_node = 1;
-#endif
-
/************
* Funtions *
************/
@@ -76,7 +72,7 @@ static sinfo_data_t *_create_sinfo(partition_info_t* part_ptr,
static bool _filter_out(node_info_t *node_ptr);
static void _sinfo_list_delete(void *data);
static bool _match_node_data(sinfo_data_t *sinfo_ptr,
- node_info_t *node_ptr);
+ node_info_t *node_ptr, uint32_t node_scaling);
static bool _match_part_data(sinfo_data_t *sinfo_ptr,
partition_info_t* part_ptr);
static int _query_server(partition_info_msg_t ** part_pptr,
@@ -334,10 +330,6 @@ static int _build_sinfo_data(List sinfo_list,
g_node_scaling = node_msg->node_scaling;
-#ifdef HAVE_BG
- cpus_per_node = node_msg->node_array[0].cpus / g_node_scaling;
-#endif
-
/* by default every partition is shown, even if no nodes */
if ((!params.node_flag) && params.match_flags.partition_flag) {
part_ptr = partition_msg->partition_array;
@@ -517,7 +509,7 @@ static void _sort_hostlist(List sinfo_list)
}
static bool _match_node_data(sinfo_data_t *sinfo_ptr,
- node_info_t *node_ptr)
+ node_info_t *node_ptr, uint32_t node_scaling)
{
if (sinfo_ptr->nodes &&
params.match_flags.features_flag &&
@@ -543,7 +535,7 @@ static bool _match_node_data(sinfo_data_t *sinfo_ptr,
return true;
if (params.match_flags.cpus_flag &&
- (node_ptr->cpus != sinfo_ptr->min_cpus))
+ ((node_ptr->cpus / node_scaling) != sinfo_ptr->min_cpus))
return false;
if (params.match_flags.sockets_flag &&
(node_ptr->sockets != sinfo_ptr->min_sockets))
@@ -630,6 +622,9 @@ static void _update_sinfo(sinfo_data_t *sinfo_ptr, node_info_t *node_ptr,
uint16_t base_state;
uint16_t used_cpus = 0, error_cpus = 0;
int total_cpus = 0, total_nodes = 0;
+ /* since node_scaling could be less here we need to use the
+ global node scaling which should never change. */
+ int single_node_cpus = (node_ptr->cpus / g_node_scaling);
base_state = node_ptr->node_state & NODE_STATE_BASE;
@@ -637,8 +632,8 @@ static void _update_sinfo(sinfo_data_t *sinfo_ptr, node_info_t *node_ptr,
sinfo_ptr->node_state = node_ptr->node_state;
sinfo_ptr->features = node_ptr->features;
sinfo_ptr->reason = node_ptr->reason;
- sinfo_ptr->min_cpus = node_ptr->cpus;
- sinfo_ptr->max_cpus = node_ptr->cpus;
+ sinfo_ptr->min_cpus = single_node_cpus;
+ sinfo_ptr->max_cpus = single_node_cpus;
sinfo_ptr->min_sockets = node_ptr->sockets;
sinfo_ptr->max_sockets = node_ptr->sockets;
sinfo_ptr->min_cores = node_ptr->cores;
@@ -656,10 +651,10 @@ static void _update_sinfo(sinfo_data_t *sinfo_ptr, node_info_t *node_ptr,
* just return, don't duplicate */
return;
} else {
- if (sinfo_ptr->min_cpus > node_ptr->cpus)
- sinfo_ptr->min_cpus = node_ptr->cpus;
- if (sinfo_ptr->max_cpus < node_ptr->cpus)
- sinfo_ptr->max_cpus = node_ptr->cpus;
+ if (sinfo_ptr->min_cpus > single_node_cpus)
+ sinfo_ptr->min_cpus = single_node_cpus;
+ if (sinfo_ptr->max_cpus < single_node_cpus)
+ sinfo_ptr->max_cpus = single_node_cpus;
if (sinfo_ptr->min_sockets > node_ptr->sockets)
sinfo_ptr->min_sockets = node_ptr->sockets;
@@ -711,31 +706,33 @@ static void _update_sinfo(sinfo_data_t *sinfo_ptr, node_info_t *node_ptr,
node_ptr->reason = xstrdup("Block(s) in error state");
sinfo_ptr->reason = node_ptr->reason;
}
- if(params.match_flags.cpus_flag && (used_cpus || error_cpus)) {
- /* we only get one shot at this (because the node name
- is the same), so we need to make
- sure we get all the subgrps accounted for here */
+ if(!params.match_flags.state_flag && (used_cpus || error_cpus)) {
+ /* We only get one shot at this (because all states
+ are combined together), so we need to make
+ sure we get all the subgrps accounted. (So use
+ g_node_scaling for safe measure) */
total_nodes = g_node_scaling;
sinfo_ptr->nodes_alloc += used_cpus;
sinfo_ptr->nodes_other += error_cpus;
sinfo_ptr->nodes_idle +=
(total_nodes - (used_cpus + error_cpus));
- used_cpus *= cpus_per_node;
- error_cpus *= cpus_per_node;
+ used_cpus *= single_node_cpus;
+ error_cpus *= single_node_cpus;
} else {
/* process only for this subgrp and then return */
- total_cpus = total_nodes * cpus_per_node;
+ total_cpus = total_nodes * single_node_cpus;
if ((base_state == NODE_STATE_ALLOCATED)
|| (node_ptr->node_state & NODE_STATE_COMPLETING)) {
sinfo_ptr->nodes_alloc += total_nodes;
sinfo_ptr->cpus_alloc += total_cpus;
- } else if (base_state == NODE_STATE_IDLE) {
- sinfo_ptr->nodes_idle += total_nodes;
- sinfo_ptr->cpus_idle += total_cpus;
- } else {
+ } else if (IS_NODE_DRAIN(node_ptr) ||
+ (base_state == NODE_STATE_DOWN)) {
sinfo_ptr->nodes_other += total_nodes;
+ sinfo_ptr->cpus_other += total_cpus;
+ } else {
+ sinfo_ptr->nodes_idle += total_nodes;
sinfo_ptr->cpus_idle += total_cpus;
}
@@ -748,10 +745,11 @@ static void _update_sinfo(sinfo_data_t *sinfo_ptr, node_info_t *node_ptr,
if ((base_state == NODE_STATE_ALLOCATED) ||
IS_NODE_COMPLETING(node_ptr))
sinfo_ptr->nodes_alloc += total_nodes;
- else if (base_state == NODE_STATE_IDLE)
+ else if (IS_NODE_DRAIN(node_ptr) ||
+ (base_state == NODE_STATE_DOWN))
+ sinfo_ptr->nodes_other += total_nodes;
+ else
sinfo_ptr->nodes_idle += total_nodes;
- else
- sinfo_ptr->nodes_other += total_nodes;
#endif
sinfo_ptr->nodes_total += total_nodes;
@@ -769,9 +767,10 @@ static void _update_sinfo(sinfo_data_t *sinfo_ptr, node_info_t *node_ptr,
} else
sinfo_ptr->cpus_idle += total_cpus;
-/* info("count is now %d %d %d %d", */
-/* sinfo_ptr->cpus_alloc, sinfo_ptr->cpus_idle, */
-/* sinfo_ptr->cpus_other, sinfo_ptr->cpus_total); */
+ /* info("count is now %d %d %d %d %d", */
+ /* sinfo_ptr->cpus_alloc, sinfo_ptr->cpus_idle, */
+ /* sinfo_ptr->cpus_other, sinfo_ptr->cpus_total, */
+ /* sinfo_ptr->nodes_total); */
}
static int _insert_node_ptr(List sinfo_list, uint16_t part_num,
@@ -787,7 +786,7 @@ static int _insert_node_ptr(List sinfo_list, uint16_t part_num,
if (!_match_part_data(sinfo_ptr, part_ptr))
continue;
if (sinfo_ptr->nodes_total &&
- (!_match_node_data(sinfo_ptr, node_ptr)))
+ (!_match_node_data(sinfo_ptr, node_ptr, node_scaling)))
continue;
_update_sinfo(sinfo_ptr, node_ptr, node_scaling);
break;
diff --git a/src/slurmctld/agent.c b/src/slurmctld/agent.c
index 678f71b78e6f466a6be16bf0efd21432b8b1a860..a59048363a4a2eecd14d297feeb597745a04d880 100644
--- a/src/slurmctld/agent.c
+++ b/src/slurmctld/agent.c
@@ -1395,6 +1395,8 @@ static void _purge_agent_args(agent_arg_t *agent_arg_ptr)
slurm_free_srun_exec_msg(agent_arg_ptr->msg_args);
else if (agent_arg_ptr->msg_type == SRUN_NODE_FAIL)
slurm_free_srun_node_fail_msg(agent_arg_ptr->msg_args);
+ else if (agent_arg_ptr->msg_type == SRUN_STEP_MISSING)
+ slurm_free_srun_step_missing_msg(agent_arg_ptr->msg_args);
else
xfree(agent_arg_ptr->msg_args);
}
diff --git a/src/slurmctld/job_mgr.c b/src/slurmctld/job_mgr.c
index a91b69c7880338c6fab31773a6d86ba3fcebcfaf..e3b15059c9d99b2ccc8eec6f10b9e7194ebefbc0 100644
--- a/src/slurmctld/job_mgr.c
+++ b/src/slurmctld/job_mgr.c
@@ -2894,7 +2894,7 @@ static int _validate_job_create_req(job_desc_msg_t * job_desc)
if (job_desc->mloaderimage &&
(strlen(job_desc->mloaderimage) > MAX_STR_LEN)) {
info("_validate_job_create_req: strlen(mloaderimage) too big (%d)",
- strlen(job_desc->features));
+ strlen(job_desc->mloaderimage));
return ESLURM_PATHNAME_TOO_LONG;
}
if (job_desc->name && (strlen(job_desc->name) > MAX_STR_LEN)) {
@@ -5029,13 +5029,33 @@ int update_job(job_desc_msg_t * job_specs, uid_t uid)
else {
#ifdef HAVE_BG
uint32_t node_cnt = job_specs->num_procs;
+ /* Set the block type to NAV here since if the
+ job size changes from a small block to a
+ regular size block the block types won't
+ jive.
+ */
+ uint16_t conn_type = (uint16_t)SELECT_NAV;
+ /* Also, if geometry is set we need to 0 out
+ the first part of it so the bluegene plugin
+ doesn't look at it any more.
+ */
+ uint16_t req_geometry[SYSTEM_DIMENSIONS] = { 0 };
+
if(cpus_per_node)
node_cnt /= cpus_per_node;
select_g_select_jobinfo_set(job_ptr->select_jobinfo,
SELECT_JOBDATA_NODE_CNT,
&node_cnt);
+ select_g_select_jobinfo_set(job_ptr->select_jobinfo,
+ SELECT_JOBDATA_CONN_TYPE,
+ &conn_type);
+ select_g_select_jobinfo_set(job_ptr->select_jobinfo,
+ SELECT_JOBDATA_GEOMETRY,
+ &req_geometry);
#endif
job_ptr->num_procs = job_specs->num_procs;
+ job_ptr->details->job_min_cpus =
+ job_specs->job_min_cpus;
info("update_job: setting num_procs to %u for "
"job_id %u", job_specs->num_procs,
job_specs->job_id);
diff --git a/src/slurmctld/reservation.c b/src/slurmctld/reservation.c
index 6f6816254f9b23a9599ea7b1432e3d048838b028..46f05a2112e22dfd89d410b632f7a5abdfeb9828 100644
--- a/src/slurmctld/reservation.c
+++ b/src/slurmctld/reservation.c
@@ -2825,7 +2825,7 @@ extern int job_resv_check(struct job_record *job_ptr)
else
job_ptr->resv_ptr->job_pend_cnt++;
- if (job_ptr->resv_ptr->end_time < (time(NULL) + resv_over_run))
+ if ((job_ptr->resv_ptr->end_time + resv_over_run) < time(NULL))
return ESLURM_INVALID_TIME_VALUE;
return SLURM_SUCCESS;
}
diff --git a/src/slurmd/slurmd/req.c b/src/slurmd/slurmd/req.c
index 879b93ddbfa8931fe03827da5a11764d7537d0da..1cc41c4a4c5e83dc572630d188da8d6496a19711 100644
--- a/src/slurmd/slurmd/req.c
+++ b/src/slurmd/slurmd/req.c
@@ -1234,7 +1234,7 @@ _abort_step(uint32_t job_id, uint32_t step_id)
step_complete_msg_t resp;
slurm_msg_t resp_msg;
slurm_msg_t_init(&resp_msg);
- int rc; /* Note: we are ignoring return code */
+ int rc, rc2;
resp.job_id = job_id;
resp.job_step_id = step_id;
@@ -1244,7 +1244,10 @@ _abort_step(uint32_t job_id, uint32_t step_id)
resp.jobacct = jobacct_gather_g_create(NULL);
resp_msg.msg_type = REQUEST_STEP_COMPLETE;
resp_msg.data = &resp;
- return slurm_send_recv_controller_rc_msg(&resp_msg, &rc);
+ rc2 = slurm_send_recv_controller_rc_msg(&resp_msg, &rc);
+ /* Note: we are ignoring the RPC return code */
+ jobacct_gather_g_destroy(resp.jobacct);
+ return rc2;
}
static void
diff --git a/src/slurmd/slurmd/slurmd.c b/src/slurmd/slurmd/slurmd.c
index 472f31aed42d01804d67f802ba9c79e08abdf8cd..51d7b98f79a663f09e14b36ec45392cdbeb8d4c1 100644
--- a/src/slurmd/slurmd/slurmd.c
+++ b/src/slurmd/slurmd/slurmd.c
@@ -1,6 +1,6 @@
/*****************************************************************************\
* src/slurmd/slurmd/slurmd.c - main slurm node server daemon
- * $Id: slurmd.c 19759 2010-03-16 20:32:26Z jette $
+ * $Id: slurmd.c 20026 2010-04-14 20:31:58Z jette $
*****************************************************************************
* Copyright (C) 2002-2007 The Regents of the University of California.
* Copyright (C) 2008-2009 Lawrence Livermore National Security.
@@ -724,9 +724,11 @@ _read_config(void)
_massage_pathname(&conf->logfile);
/* set node_addr if relevant */
- if((conf->node_addr = slurm_conf_get_nodeaddr(conf->hostname)))
- if (strcmp(conf->node_addr, conf->hostname) == 0)
- xfree(conf->node_addr); /* Sets to NULL */
+ if ((conf->node_addr == NULL) &&
+ (conf->node_addr = slurm_conf_get_nodeaddr(conf->hostname)) &&
+ (strcmp(conf->node_addr, conf->hostname) == 0)) {
+ xfree(conf->node_addr); /* Sets to NULL */
+ }
conf->port = slurm_conf_get_port(conf->node_name);
slurm_conf_get_cpus_sct(conf->node_name,
diff --git a/src/slurmd/slurmstepd/mgr.c b/src/slurmd/slurmstepd/mgr.c
index fc9f0e7d2aa91756ac2f82582d792d089d420fed..a32336089d32546ca402b235d546215c9e1100d1 100644
--- a/src/slurmd/slurmstepd/mgr.c
+++ b/src/slurmd/slurmstepd/mgr.c
@@ -1,6 +1,6 @@
/*****************************************************************************\
* src/slurmd/slurmstepd/mgr.c - job manager functions for slurmstepd
- * $Id: mgr.c 19858 2010-03-23 21:29:05Z jette $
+ * $Id: mgr.c 20036 2010-04-15 18:52:13Z da $
*****************************************************************************
* Copyright (C) 2002-2007 The Regents of the University of California.
* Copyright (C) 2008-2009 Lawrence Livermore National Security.
@@ -1649,7 +1649,7 @@ _send_launch_resp(slurmd_job_t *job, int rc)
}
if (slurm_send_only_node_msg(&resp_msg) != SLURM_SUCCESS)
- error("failed to send RESPONSE_LAUNCH_TASKS: %m");
+ error("failed to send RESPONSE_LAUNCH_TASKS: %m");
xfree(resp.local_pids);
xfree(resp.task_ids);
diff --git a/src/slurmd/slurmstepd/slurmstepd_job.h b/src/slurmd/slurmstepd/slurmstepd_job.h
index 3a9735f0ee50d71ba195c006960ef3ba3e424c05..f7c66c3e88d95e68fb520435da49c32c96ab966e 100644
--- a/src/slurmd/slurmstepd/slurmstepd_job.h
+++ b/src/slurmd/slurmstepd/slurmstepd_job.h
@@ -1,38 +1,38 @@
/*****************************************************************************\
* src/slurmd/slurmstepd/slurmstepd_job.h slurmd_job_t definition
- * $Id: slurmstepd_job.h 19152 2009-12-10 22:29:52Z da $
+ * $Id: slurmstepd_job.h 19917 2010-03-30 16:41:15Z da $
*****************************************************************************
* Copyright (C) 2002-2007 The Regents of the University of California.
* Copyright (C) 2008-2009 Lawrence Livermore National Security.
* Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER).
* Written by Mark Grondona <mgrondona@llnl.gov>.
* CODE-OCEC-09-009. All rights reserved.
- *
+ *
* This file is part of SLURM, a resource management program.
* For details, see <https://computing.llnl.gov/linux/slurm/>.
* Please also read the included file: DISCLAIMER.
- *
+ *
* SLURM is free software; you can redistribute it and/or modify it under
* the terms of the GNU General Public License as published by the Free
* Software Foundation; either version 2 of the License, or (at your option)
* any later version.
*
- * In addition, as a special exception, the copyright holders give permission
- * to link the code of portions of this program with the OpenSSL library under
- * certain conditions as described in each individual source file, and
- * distribute linked combinations including the two. You must obey the GNU
- * General Public License in all respects for all of the code used other than
- * OpenSSL. If you modify file(s) with this exception, you may extend this
- * exception to your version of the file(s), but you are not obligated to do
+ * In addition, as a special exception, the copyright holders give permission
+ * to link the code of portions of this program with the OpenSSL library under
+ * certain conditions as described in each individual source file, and
+ * distribute linked combinations including the two. You must obey the GNU
+ * General Public License in all respects for all of the code used other than
+ * OpenSSL. If you modify file(s) with this exception, you may extend this
+ * exception to your version of the file(s), but you are not obligated to do
* so. If you do not wish to do so, delete this exception statement from your
- * version. If you delete this exception statement from all source files in
+ * version. If you delete this exception statement from all source files in
* the program, then also delete it here.
- *
+ *
* SLURM is distributed in the hope that it will be useful, but WITHOUT ANY
* WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
* FOR A PARTICULAR PURPOSE. See the GNU General Public License for more
* details.
- *
+ *
* You should have received a copy of the GNU General Public License along
* with SLURM; if not, write to the Free Software Foundation, Inc.,
* 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
@@ -84,7 +84,7 @@ typedef enum task_state {
typedef struct task_info {
pthread_mutex_t mutex; /* mutex to protect task state */
slurmd_task_state_t state; /* task state */
-
+
int id; /* local task id */
uint32_t gtid; /* global task id */
pid_t pid; /* task pid */
@@ -185,7 +185,7 @@ typedef struct slurmd_job {
pid_t jmgr_pid; /* job manager pid */
pid_t pgid; /* process group id for tasks */
- uint16_t task_flags;
+ uint16_t task_flags;
uint16_t multi_prog;
uint16_t overcommit;
env_t *envtp;
@@ -212,8 +212,8 @@ void job_kill(slurmd_job_t *job, int signal);
void job_destroy(slurmd_job_t *job);
-struct srun_info * srun_info_create(slurm_cred_t *cred, slurm_addr *respaddr,
- slurm_addr *ioaddr);
+struct srun_info * srun_info_create(slurm_cred_t *cred, slurm_addr *respaddr,
+ slurm_addr *ioaddr);
void srun_info_destroy(struct srun_info *srun);
diff --git a/src/slurmd/slurmstepd/task.c b/src/slurmd/slurmstepd/task.c
index 1436f72e15a0a4ec93c2bab55c1f75655dd6a984..daa5e5c4257db566e304c24b63b81084a2ff2d60 100644
--- a/src/slurmd/slurmstepd/task.c
+++ b/src/slurmd/slurmstepd/task.c
@@ -100,68 +100,81 @@
* Static prototype definitions.
*/
static void _make_tmpdir(slurmd_job_t *job);
-static void _print_stdout(char *buf);
static int _run_script_and_set_env(const char *name, const char *path,
slurmd_job_t *job);
-static void _update_env(char *buf, char ***env);
+static void _proc_stdout(char *buf, char ***env);
static char *_uint32_array_to_str(int array_len, const uint32_t *array);
-/* Search for "export NAME=value" records in buf and
- * use them to add environment variables to env */
-static void _update_env(char *buf, char ***env)
+/*
+ * Process TaskProlog output
+ * "export NAME=value" adds environment variables
+ * "unset NAME" clears an environment variable
+ * "print <whatever>" writes that to the job's stdout
+ */
+static void _proc_stdout(char *buf, char ***env)
{
- char *tmp_ptr, *name_ptr, *val_ptr, *buf_ptr = buf;
-
- while ((tmp_ptr = strstr(buf_ptr, "export"))) {
- buf_ptr += 6;
- while (isspace(buf_ptr[0]))
- buf_ptr++;
- if (buf_ptr[0] == '=') /* mal-formed */
- continue;
- name_ptr = buf_ptr; /* start of env var name */
- while ((buf_ptr[0] != '=') && (buf_ptr[0] != '\0'))
- buf_ptr++;
- if (buf_ptr[0] == '\0') /* mal-formed */
- continue;
- buf_ptr[0] = '\0'; /* end of env var name */
- buf_ptr++;
- val_ptr = buf_ptr; /* start of env var value */
- while ((!isspace(buf_ptr[0])) && (buf_ptr[0] != '\0'))
- buf_ptr++;
- if (isspace(buf_ptr[0])) {
- buf_ptr[0] = '\0';/* end of env var value */
- buf_ptr++;
+ bool end_buf = false;
+ int len;
+ char *buf_ptr, *name_ptr, *val_ptr;
+ char *end_line, *equal_ptr;
+
+ buf_ptr = buf;
+ while (buf_ptr[0]) {
+ end_line = strchr(buf_ptr, '\n');
+ if (!end_line) {
+ end_line = buf_ptr + strlen(buf_ptr);
+ end_buf = true;
}
- debug("name:%s:val:%s:",name_ptr,val_ptr);
- if (setenvf(env, name_ptr, "%s", val_ptr))
- error("Unable to set %s environment variable",
- name_ptr);
- }
-}
-
-/* Search for "print <whatever>" records in buf and
- * write that to the job's stdout */
-static void _print_stdout(char *buf)
-{
- char *tmp_ptr, *buf_ptr = buf;
-
- while ((tmp_ptr = strstr(buf_ptr, "print "))) {
- if ((tmp_ptr != buf_ptr) && (tmp_ptr[-1] != '\n')) {
- /* Skip "print " if not at start of a line */
- buf_ptr +=6;
- continue;
+ if (!strncmp(buf_ptr, "print ", 6)) {
+ buf_ptr += 6;
+ while (isspace(buf_ptr[0]))
+ buf_ptr++;
+ len = end_line - buf_ptr + 1;
+ safe_write(1, buf_ptr, len);
+ } else if (!strncmp(buf_ptr, "export ",7)) {
+ name_ptr = buf_ptr + 7;
+ while (isspace(name_ptr[0]))
+ name_ptr++;
+ equal_ptr = strchr(name_ptr, '=');
+ if (!equal_ptr || (equal_ptr > end_line))
+ goto rwfail;
+ val_ptr = equal_ptr + 1;
+ while (isspace(equal_ptr[-1]))
+ equal_ptr--;
+ equal_ptr[0] = '\0';
+ end_line[0] = '\0';
+ debug("export name:%s:val:%s:", name_ptr, val_ptr);
+ if (setenvf(env, name_ptr, "%s", val_ptr)) {
+ error("Unable to set %s environment variable",
+ buf_ptr);
+ }
+ equal_ptr[0] = '=';
+ if (end_buf)
+ end_line[0] = '\0';
+ else
+ end_line[0] = '\n';
+ } else if (!strncmp(buf_ptr, "unset ", 6)) {
+ name_ptr = buf_ptr + 6;
+ while (isspace(name_ptr[0]))
+ name_ptr++;
+ if ((name_ptr[0] == '\n') || (name_ptr[0] == '\0'))
+ goto rwfail;
+ while (isspace(end_line[-1]))
+ end_line--;
+ end_line[0] = '\0';
+ debug(" unset name:%s:", name_ptr);
+ unsetenvp(*env, name_ptr);
+ if (end_buf)
+ end_line[0] = '\0';
+ else
+ end_line[0] = '\n';
}
- buf_ptr = tmp_ptr + 6;
- tmp_ptr = strchr(buf_ptr, '\n');
- if (tmp_ptr) {
- safe_write(1, buf_ptr, (tmp_ptr - buf_ptr + 1));
- buf_ptr = tmp_ptr + 1;
- } else {
- safe_write(1, buf_ptr, strlen(buf_ptr));
+
+rwfail: /* process rest of script output */
+ if (end_buf)
break;
- }
- }
-rwfail:
+ buf_ptr = end_line + 1;
+ }
return;
}
@@ -180,7 +193,7 @@ _run_script_and_set_env(const char *name, const char *path, slurmd_job_t *job)
{
int status, rc, nread;
pid_t cpid;
- int pfd[2];
+ int pfd[2], offset = 0;
char buf[4096];
xassert(job->env);
@@ -211,6 +224,8 @@ _run_script_and_set_env(const char *name, const char *path, slurmd_job_t *job)
error("couldn't do the dup: %m");
close(2);
close(0);
+ close(pfd[0]);
+ close(pfd[1]);
#ifdef SETPGRP_TWO_ARGS
setpgrp(0, 0);
#else
@@ -222,12 +237,11 @@ _run_script_and_set_env(const char *name, const char *path, slurmd_job_t *job)
}
close(pfd[1]);
- while ((nread = read(pfd[0], buf, sizeof(buf))) > 0) {
- buf[nread] = 0;
- //debug("read %d:%s:", nread, buf);
- _update_env(buf, &job->env);
- _print_stdout(buf);
- }
+ buf[0] = '\0';
+ while ((nread = read(pfd[0], buf+offset, (sizeof(buf)-offset))) > 0)
+ offset += nread;
+ /* debug ("read %d:%s:", offset, buf); */
+ _proc_stdout(buf, &job->env);
close(pfd[0]);
while (1) {
diff --git a/src/smap/configure_functions.c b/src/smap/configure_functions.c
index 6b6b34a08982f2d07b34c39ce578004a4df8dfa2..0f1ffeec0726e6d439e18c05ec858fbcb6c058f6 100644
--- a/src/smap/configure_functions.c
+++ b/src/smap/configure_functions.c
@@ -77,7 +77,8 @@ static void _delete_allocated_blocks(List allocated_blocks)
allocated_block_t *allocated_block = NULL;
while ((allocated_block = list_pop(allocated_blocks)) != NULL) {
- remove_block(allocated_block->nodes,0);
+ remove_block(allocated_block->nodes, 0,
+ allocated_block->request->conn_type);
list_destroy(allocated_block->nodes);
delete_ba_request(allocated_block->request);
xfree(allocated_block);
@@ -835,7 +836,9 @@ static int _remove_allocation(char *com, List allocated_blocks)
} else if(allocated_block->letter == letter) {
found=1;
remove_block(allocated_block->nodes,
- color_count);
+ color_count,
+ allocated_block->request->
+ conn_type);
list_destroy(allocated_block->nodes);
delete_ba_request(allocated_block->request);
list_remove(results_i);
diff --git a/src/srun/allocate.c b/src/srun/allocate.c
index 5637ca969b04a5d7b0cbeb16228c232ae8504e56..31f76c3bbef853c132ee889b340c2bedcffcb5ed 100644
--- a/src/srun/allocate.c
+++ b/src/srun/allocate.c
@@ -387,7 +387,6 @@ allocate_nodes(void)
xsignal(SIGUSR2, _signal_while_allocating);
while (!resp) {
-//WHAT IS THIS?
resp = slurm_allocate_resources_blocking(j, opt.immediate,
_set_pending_job_id);
if (destroy_job) {
@@ -520,7 +519,7 @@ slurmctld_msg_init(void)
* (see opt.h)
*/
job_desc_msg_t *
-job_desc_msg_create_from_opts ()
+job_desc_msg_create_from_opts (void)
{
job_desc_msg_t *j = xmalloc(sizeof(*j));
char buf[8192];
@@ -536,6 +535,8 @@ job_desc_msg_create_from_opts ()
j->name = xstrdup(opt.job_name);
else
j->name = xstrdup(opt.cmd_name);
+ if (opt.acctg_freq >= 0)
+ j->acctg_freq = opt.acctg_freq;
j->reservation = xstrdup(opt.reservation);
j->wckey = xstrdup(opt.wckey);
diff --git a/src/srun/srun.c b/src/srun/srun.c
index c2f826dd3dc925bc2c8b151c87faf52d6c2fc8d2..abfc55af8def5f289814e985706dbdc44b09591c 100644
--- a/src/srun/srun.c
+++ b/src/srun/srun.c
@@ -156,6 +156,7 @@ static int _slurm_debug_env_val (void);
static void _task_start(launch_tasks_response_msg_t *msg);
static void _task_finish(task_exit_msg_t *msg);
static char *_uint16_array_to_str(int count, const uint16_t *array);
+static int _validate_relative(resource_allocation_response_msg_t *resp);
/*
* from libvirt-0.6.2 GPL2
@@ -296,6 +297,8 @@ int srun(int ac, char **av)
if (opt.exclusive)
_step_opt_exclusive();
_set_cpu_env_var(resp);
+ if (_validate_relative(resp))
+ exit(error_exit);
job = job_step_create_allocation(resp);
slurm_free_resource_allocation_response_msg(resp);
@@ -315,12 +318,22 @@ int srun(int ac, char **av)
exit(error_exit);
}
#endif
+ if (opt.relative_set && opt.relative) {
+ error("--relative option ignored in job allocation "
+ "request");
+ opt.relative = NO_VAL;
+ opt.relative_set = false;
+ }
if ( !(resp = allocate_nodes()) )
exit(error_exit);
got_alloc = 1;
_print_job_information(resp);
_set_cpu_env_var(resp);
+ if (_validate_relative(resp)) {
+ slurm_complete_job(resp->job_id, 1);
+ exit(error_exit);
+ }
job = job_create_allocation(resp);
opt.exclusive = false; /* not applicable for this step */
@@ -735,6 +748,25 @@ static void _set_cpu_env_var(resource_allocation_response_msg_t *resp)
return;
}
+static int _validate_relative(resource_allocation_response_msg_t *resp)
+{
+
+ if (opt.relative_set &&
+ ((opt.relative + opt.min_nodes) > resp->node_cnt)) {
+ if (opt.nodes_set_opt) { /* -N command line option used */
+ error("--relative and --nodes option incompatable "
+ "with count of allocated nodes (%d+%d>%d)",
+ opt.relative, opt.min_nodes, resp->node_cnt);
+ } else { /* SLURM_NNODES option used */
+ error("--relative and SLURM_NNODES option incompatable "
+ "with count of allocated nodes (%d+%d>%d)",
+ opt.relative, opt.min_nodes, resp->node_cnt);
+ }
+ return -1;
+ }
+ return 0;
+}
+
/* Set SLURM_RLIMIT_* environment variables with current resource
* limit values, reset RLIMIT_NOFILE to maximum possible value */
static int _set_rlimit_env(void)
@@ -1035,8 +1067,13 @@ _task_start(launch_tasks_response_msg_t *msg)
int taskid;
int i;
- verbose("Node %s (%d), %d tasks started",
- msg->node_name, msg->srun_node_id, msg->count_of_pids);
+ if(msg->count_of_pids)
+ verbose("Node %s, %d tasks started",
+ msg->node_name, msg->count_of_pids);
+ else
+ error("No tasks started on node %s: %s",
+ msg->node_name, slurm_strerror(msg->return_code));
+
for (i = 0; i < msg->count_of_pids; i++) {
taskid = msg->task_ids[i];
diff --git a/testsuite/expect/test21.27 b/testsuite/expect/test21.27
index 0adc4cf537bfd3759546bcffbbd3716e4e63a20f..81fa81a04dd242e21f6a07218fb501fc0e2ef180 100755
--- a/testsuite/expect/test21.27
+++ b/testsuite/expect/test21.27
@@ -50,11 +50,10 @@ set ac3 pa3
set usr user
set us1 usr1
set us2 usr2
-set us3 gdm
+set us3 mysql
set access_err 0
set timeout 120
-#set user_name "id -u -n"
print_header $test_id