diff --git a/META b/META
index 7061ad1f76192f37cdfffd26364b72c61b2e928f..60cb6677127dda29c05435513c9d5d17fce65178 100644
--- a/META
+++ b/META
@@ -3,9 +3,9 @@
Api_revision: 0
Major: 1
Meta: 1
- Micro: 22
+ Micro: 23
Minor: 2
Name: slurm
Release: 1
Release_tags:
- Version: 1.2.22
+ Version: 1.2.23
diff --git a/NEWS b/NEWS
index 212ceda0a22478fbf8768e691b9c8be50abaf865..2cb6a555b8246a25bfa11a631f98d7614020286a 100644
--- a/NEWS
+++ b/NEWS
@@ -3,6 +3,31 @@ documents those changes that are of interest to users and admins.
* Changes in SLURM 1.2.23
=========================
+ -- Fix for libpmi to not export unneeded variables like xstr*
+ -- BLUEGENE - added per partition dynamic block creation
+ -- fix infinite loop bug in sview when there were multiple partitions
+ -- Send message to srun command when a job is requeued due to node failure.
+ Note this will be overwritten in the output file unless JobFileAppend
+ is set in slurm.conf. In slurm version 1.3, srun's --open-mode=append
+ option will offer this control for each job.
+ -- Change a node's default TmpDisk from 1MB to 0MB and change job's default
+ disk space requirement from 1MB to 0MB.
+ -- In sched/wiki (Maui scheduler) specify a QOS (quality of service) by
+ specifying an account of the form "qos-name".
+ -- In select/linear, fix bug in scheduling required nodes that already have
+ a job running on them (req.load.patch from Chris Holmes, HP).
+ -- For use with Moab only: change timeout for srun/sbatch --get-user-env
+ option to 2 secs, don't get DISPLAY environment variables, but explicitly
+ set ENVIRONMENT=BATCH and HOSTNAME to the execution host of the batch script.
+ -- Add configuration parameter GetEnvTimeout for use with Moab. See
+ "man slurm.conf" for details.
+ -- Modify salloc and sbatch to accept both "--tasks" and "--ntasks" as
+ equivalent options for compatibility with srun.
+ -- If a partition's node list contains space separators, replace them with
+ commas for easier parsing.
+ -- BLUEGENE - fixed bug in geometry specs when creating a block.
+ -- Add support for Moab and Maui to start jobs with select/cons_res plugin
+ and jobs requiring more than one CPU per task.
* Changes in SLURM 1.2.22
=========================
@@ -2790,4 +2815,4 @@ 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 13118 2008-01-29 19:09:00Z da $
+$Id: NEWS 13293 2008-02-15 21:51:16Z jette $
diff --git a/doc/html/configurator.html.in b/doc/html/configurator.html.in
index 8b0faed7b7031b37a72a44c8b6741dc6b7f602e2..b7d09d968884ab511373b02607b77191ca33902b 100644
--- a/doc/html/configurator.html.in
+++ b/doc/html/configurator.html.in
@@ -159,6 +159,7 @@ function displayfile()
"InactiveLimit=" + document.config.inactive_limit.value + "<br>" +
"MinJobAge=" + document.config.min_job_age.value + "<br>" +
"KillWait=" + document.config.kill_wait.value + "<br>" +
+ "#GetEnvTimeout=2 <br>" +
"#UnkillableStepProgram= <br>" +
"#UnkillableStepTimeout=60 <br>" +
"Waittime=" + document.config.wait_time.value + "<br>" +
diff --git a/doc/man/man1/srun.1 b/doc/man/man1/srun.1
index 15bf0201312f86b53a49cc6e9d8dab45908e10f9..e58121e2c43061c1761865692133789bb830f0ab 100644
--- a/doc/man/man1/srun.1
+++ b/doc/man/man1/srun.1
@@ -1,4 +1,4 @@
-\." $Id: srun.1 13066 2008-01-23 16:34:55Z jette $
+\." $Id: srun.1 13256 2008-02-12 22:02:36Z jette $
.\"
.TH SRUN "1" "July 2007" "srun 1.2" "slurm components"
@@ -335,6 +335,7 @@ use only one core in each physical CPU
.B [no]multithread
[don't] use extra threads with in-core multi-threading
which can benefit communication intensive applications
+.TP
.B help
show this help message
.RE
diff --git a/doc/man/man5/slurm.conf.5 b/doc/man/man5/slurm.conf.5
index b330630f1e8b05ef246a9f57e3141eac361d2f9b..43a9508727d550b35efeabebf834f5e1a6bda995 100644
--- a/doc/man/man5/slurm.conf.5
+++ b/doc/man/man5/slurm.conf.5
@@ -145,6 +145,14 @@ specific requested value. Job id values generated will incremented by 1
for each subsequent job. This may be used to provide a meta\-scheduler
with a job id space which is disjoint from the interactive jobs.
The default value is 1.
+
+.TP
+\fBGetEnvTimeout\fR
+Used for Moab scheduled jobs only. Controls how long job should wait
+in seconds for loading the user's environment before attempting to
+load it from a cache file. Applies when the srun or sbatch
+\fI--get-user-env\fR option is used. Default value is 2 seconds.
+
.TP
\fBHeartbeatInterval\fR
Defunct paramter.
@@ -950,7 +958,7 @@ system is purged as needed so that user jobs have access to
most of this space.
The Prolog and/or Epilog programs (specified in the configuration file)
might be used to insure the file system is kept clean.
-The default value is 1.
+The default value is 0.
.TP
\fBWeight\fR
The priority of the node for scheduling purposes.
diff --git a/slurm.spec b/slurm.spec
index de1d4ded7900b27aa0c0872f0c930fdfcc8aad41..5f7e05c94087d9b53dfa38ce63391e11052fcdae 100644
--- a/slurm.spec
+++ b/slurm.spec
@@ -1,4 +1,4 @@
-# $Id: slurm.spec 13075 2008-01-23 20:39:30Z da $
+# $Id: slurm.spec 13266 2008-02-13 21:54:50Z da $
#
# Note that this package is not relocatable
@@ -60,14 +60,14 @@
%endif
Name: slurm
-Version: 1.2.22
+Version: 1.2.23
Release: 1
Summary: Simple Linux Utility for Resource Management
License: GPL
Group: System Environment/Base
-Source: slurm-1.2.22.tar.bz2
+Source: slurm-1.2.23.tar.bz2
BuildRoot: %{_tmppath}/%{name}-%{version}-%{release}
URL: https://computing.llnl.gov/linux/slurm/
BuildRequires: openssl-devel >= 0.9.6 openssl >= 0.9.6
@@ -211,7 +211,7 @@ SLURM process tracking plugin for SGI job containers.
#############################################################################
%prep
-%setup -n slurm-1.2.22
+%setup -n slurm-1.2.23
%build
%configure --program-prefix=%{?_program_prefix:%{_program_prefix}} \
@@ -219,7 +219,7 @@ SLURM process tracking plugin for SGI job containers.
%{?with_proctrack} \
%{?with_ssl} \
%{?with_munge} \
- %{!?with_readline:--without-readline} \
+ %{!?slurm_with_readline:--without-readline} \
%{?with_cflags}
make %{?_smp_mflags}
diff --git a/slurm/slurm.h.in b/slurm/slurm.h.in
index fa7596d071c40e00d24f600e7628c1a01ed767cd..7f76820d69042b95ded602167ded5015018c424f 100644
--- a/slurm/slurm.h.in
+++ b/slurm/slurm.h.in
@@ -868,6 +868,7 @@ typedef struct slurm_ctl_conf {
char *job_comp_type; /* job completion logger type */
char *job_comp_loc; /* job completion logging location */
uint16_t job_file_append; /* if set, append to stdout/err file */
+ uint16_t get_env_timeout; /* secs allowed for srun --get-user-env */
uint16_t kill_wait; /* seconds between SIGXCPU to SIGKILL
* on job termination */
char *mail_prog; /* pathname of mail program */
diff --git a/src/api/Makefile.am b/src/api/Makefile.am
index 6aeddd8699f614fef3f3dcb098dc456125b7b501..819f5bd2c47d09355e9cfd077668e35f0755f651 100644
--- a/src/api/Makefile.am
+++ b/src/api/Makefile.am
@@ -10,6 +10,10 @@ VERSION_SCRIPT = \
version.map
OTHER_FLAGS = \
-Wl,--version-script=$(VERSION_SCRIPT)
+PMI_VERSION_SCRIPT = \
+ pmi_version.map
+PMI_OTHER_FLAGS = \
+ -Wl,--version-script=$(PMI_VERSION_SCRIPT)
endif
# libslurm version information :
@@ -46,7 +50,7 @@ lib_LTLIBRARIES = libslurm.la libpmi.la
# need to be built before anything else. For instance, this takes care
# of libpmi.la's dependency on libslurm.la, as long as you use "make"
# and NOT "make libpmi.la".
-BUILT_SOURCES = $(VERSION_SCRIPT) libslurm.la
+BUILT_SOURCES = $(VERSION_SCRIPT) $(PMI_VERSION_SCRIPT) libslurm.la
# Note that libslurmhelper is mostly the same as libslurm, except that
# it exports ALL symbols, including those from libcommon, libeio, etc.
@@ -101,8 +105,9 @@ libslurm_la_LDFLAGS = \
$(OTHER_FLAGS)
libpmi_la_SOURCES = pmi.c
-libpmi_la_LIBADD = $(convenience_libs)
-libpmi_la_LDFLAGS = $(LIB_LDFLAGS)
+#libpmi_la_LIBADD = $(convenience_libs)
+libpmi_la_LDFLAGS = $(LIB_LDFLAGS) \
+ $(PMI_OTHER_FLAGS)
force:
$(convenience_libs) : force
@@ -119,8 +124,14 @@ $(VERSION_SCRIPT) :
echo " local: *;"; \
echo "};") > $(VERSION_SCRIPT)
+$(PMI_VERSION_SCRIPT) :
+ (echo "{ global:"; \
+ echo " PMI_*;"; \
+ echo " local: *;"; \
+ echo "};") > $(PMI_VERSION_SCRIPT)
+
CLEANFILES = \
- $(VERSION_SCRIPT)
+ $(VERSION_SCRIPT) $(PMI_VERSION_SCRIPT)
DISTCLEANFILES = \
- $(VERSION_SCRIPT)
+ $(VERSION_SCRIPT) $(PMI_VERSION_SCRIPT)
diff --git a/src/api/Makefile.in b/src/api/Makefile.in
index 4e9f9cfe6b75c957a9f2e949d16feeb8addaf491..2d94ff509015fd52153f80e14bce785c5f5305e6 100644
--- a/src/api/Makefile.in
+++ b/src/api/Makefile.in
@@ -74,7 +74,7 @@ am__strip_dir = `echo $$p | sed -e 's|^.*/||'`;
am__installdirs = "$(DESTDIR)$(libdir)"
libLTLIBRARIES_INSTALL = $(INSTALL)
LTLIBRARIES = $(lib_LTLIBRARIES) $(noinst_LTLIBRARIES)
-libpmi_la_DEPENDENCIES = $(convenience_libs)
+libpmi_la_LIBADD =
am_libpmi_la_OBJECTS = pmi.lo
libpmi_la_OBJECTS = $(am_libpmi_la_OBJECTS)
libpmi_la_LINK = $(LIBTOOL) --tag=CC $(AM_LIBTOOLFLAGS) \
@@ -280,6 +280,12 @@ INCLUDES = -I$(top_srcdir)
@WITH_GNU_LD_TRUE@OTHER_FLAGS = \
@WITH_GNU_LD_TRUE@ -Wl,--version-script=$(VERSION_SCRIPT)
+@WITH_GNU_LD_TRUE@PMI_VERSION_SCRIPT = \
+@WITH_GNU_LD_TRUE@ pmi_version.map
+
+@WITH_GNU_LD_TRUE@PMI_OTHER_FLAGS = \
+@WITH_GNU_LD_TRUE@ -Wl,--version-script=$(PMI_VERSION_SCRIPT)
+
# libslurm version information :
#
@@ -314,7 +320,7 @@ lib_LTLIBRARIES = libslurm.la libpmi.la
# need to be built before anything else. For instance, this takes care
# of libpmi.la's dependency on libslurm.la, as long as you use "make"
# and NOT "make libpmi.la".
-BUILT_SOURCES = $(VERSION_SCRIPT) libslurm.la
+BUILT_SOURCES = $(VERSION_SCRIPT) $(PMI_VERSION_SCRIPT) libslurm.la
# Note that libslurmhelper is mostly the same as libslurm, except that
# it exports ALL symbols, including those from libcommon, libeio, etc.
@@ -366,13 +372,15 @@ libslurm_la_LDFLAGS = \
$(OTHER_FLAGS)
libpmi_la_SOURCES = pmi.c
-libpmi_la_LIBADD = $(convenience_libs)
-libpmi_la_LDFLAGS = $(LIB_LDFLAGS)
+#libpmi_la_LIBADD = $(convenience_libs)
+libpmi_la_LDFLAGS = $(LIB_LDFLAGS) \
+ $(PMI_OTHER_FLAGS)
+
CLEANFILES = \
- $(VERSION_SCRIPT)
+ $(VERSION_SCRIPT) $(PMI_VERSION_SCRIPT)
DISTCLEANFILES = \
- $(VERSION_SCRIPT)
+ $(VERSION_SCRIPT) $(PMI_VERSION_SCRIPT)
all: $(BUILT_SOURCES)
$(MAKE) $(AM_MAKEFLAGS) all-am
@@ -707,6 +715,12 @@ $(VERSION_SCRIPT) :
echo " client_io_handler_*;"; \
echo " local: *;"; \
echo "};") > $(VERSION_SCRIPT)
+
+$(PMI_VERSION_SCRIPT) :
+ (echo "{ global:"; \
+ echo " PMI_*;"; \
+ echo " local: *;"; \
+ echo "};") > $(PMI_VERSION_SCRIPT)
# Tell versions [3.59,3.63) of GNU make to not export all variables.
# Otherwise a system limit (for SysV at least) may be exceeded.
.NOEXPORT:
diff --git a/src/common/env.c b/src/common/env.c
index 741152a0c14d7d976536e588c0247a7144c241a9..424faa2773bab5d791412e1752ce9e6b8b92f4ea 100644
--- a/src/common/env.c
+++ b/src/common/env.c
@@ -1,6 +1,6 @@
/*****************************************************************************\
* src/common/env.c - add an environment variable to environment vector
- * $Id: env.c 12970 2008-01-07 20:16:53Z jette $
+ * $Id: env.c 13237 2008-02-08 23:16:16Z jette $
*****************************************************************************
* Copyright (C) 2002-2007 The Regents of the University of California.
* Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER).
@@ -53,6 +53,7 @@
#include "slurm/slurm.h"
#include "src/common/log.h"
#include "src/common/env.h"
+#include "src/common/read_config.h"
#include "src/common/xassert.h"
#include "src/common/xmalloc.h"
#include "src/common/xstring.h"
@@ -75,8 +76,6 @@ strong_alias(env_array_append_fmt, slurm_env_array_append_fmt);
strong_alias(env_array_overwrite, slurm_env_array_overwrite);
strong_alias(env_array_overwrite_fmt, slurm_env_array_overwrite_fmt);
-#define SU_WAIT_MSEC 8000 /* 8000 msec for /bin/su to return user
- * env vars for --get-user-env option */
#define ENV_BUFSIZE (64 * 1024)
/*
@@ -132,6 +131,18 @@ _extend_env(char ***envp)
return (++ep);
}
+/* return true if the environment variables should not be set for
+ * srun's --get-user-env option */
+static bool _discard_env(char *name, char *value)
+{
+ if ((strcmp(name, "DISPLAY") == 0) ||
+ (strcmp(name, "ENVIRONMENT") == 0) ||
+ (strcmp(name, "HOSTNAME") == 0))
+ return true;
+
+ return false;
+}
+
/*
* Return the number of elements in the environment `env'
*/
@@ -787,6 +798,7 @@ env_array_for_job(char ***dest, const resource_allocation_response_msg_t *alloc)
* SLURM_JOB_NODELIST
* SLURM_JOB_CPUS_PER_NODE
* ENVIRONMENT=BATCH
+ * HOSTNAME
* LOADLBATCH (AIX only)
*
* Sets OBSOLETE variables:
@@ -796,8 +808,9 @@ env_array_for_job(char ***dest, const resource_allocation_response_msg_t *alloc)
* SLURM_TASKS_PER_NODE <- poorly named, really CPUs per node
* ? probably only needed for users...
*/
-void
-env_array_for_batch_job(char ***dest, const batch_job_launch_msg_t *batch)
+extern void
+env_array_for_batch_job(char ***dest, const batch_job_launch_msg_t *batch,
+ const char *node_name)
{
char *tmp;
uint32_t num_nodes = 0;
@@ -817,6 +830,8 @@ env_array_for_batch_job(char ***dest, const batch_job_launch_msg_t *batch)
batch->cpu_count_reps);
env_array_overwrite_fmt(dest, "SLURM_JOB_CPUS_PER_NODE", "%s", tmp);
env_array_overwrite_fmt(dest, "ENVIRONMENT", "BATCH");
+ if (node_name)
+ env_array_overwrite_fmt(dest, "HOSTNAME", "%s", node_name);
#ifdef HAVE_AIX
/* this puts the "poe" command into batch mode */
env_array_overwrite(dest, "LOADLBATCH", "yes");
@@ -1254,9 +1269,10 @@ char **_load_env_cache(const char *username)
if (!fgets(line, ENV_BUFSIZE, fp))
break;
_strip_cr_nl(line);
- _env_array_entry_splitter(line, name, ENV_BUFSIZE, value,
- ENV_BUFSIZE);
- env_array_overwrite(&env, name, value);
+ if (_env_array_entry_splitter(line, name, ENV_BUFSIZE, value,
+ ENV_BUFSIZE) &&
+ (!_discard_env(name, value)))
+ env_array_overwrite(&env, name, value);
}
fclose(fp);
return env;
@@ -1271,7 +1287,7 @@ char **_load_env_cache(const char *username)
* 2. Load the user environment from a cache file. This is used
* in the event that option 1 times out.
*
- * timeout value is in seconds or zero for default (8 secs)
+ * timeout value is in seconds or zero for default (2 secs)
* mode is 1 for short ("su <user>"), 2 for long ("su - <user>")
* On error, returns NULL.
*
@@ -1293,7 +1309,7 @@ char **env_array_user_default(const char *username, int timeout, int mode)
struct pollfd ufds;
if (geteuid() != (uid_t)0) {
- info("WARNING: you must be root to use --get-user-env");
+ fatal("WARNING: you must be root to use --get-user-env");
return NULL;
}
@@ -1340,15 +1356,14 @@ char **env_array_user_default(const char *username, int timeout, int mode)
ufds.events = POLLIN;
/* Read all of the output from /bin/su into buffer */
+ if ((timeout == 0) && ((timeout = slurm_get_env_timeout()) == 0))
+ timeleft = DEFAULT_GET_ENV_TIMEOUT;
found = 0;
buf_read = 0;
bzero(buffer, sizeof(buffer));
while (1) {
gettimeofday(&now, NULL);
- if (timeout)
- timeleft = timeout * 1000;
- else
- timeleft = SU_WAIT_MSEC;
+ timeleft = timeout * 1000;
timeleft -= (now.tv_sec - begin.tv_sec) * 1000;
timeleft -= (now.tv_usec - begin.tv_usec) / 1000;
if (timeleft <= 0) {
@@ -1394,7 +1409,7 @@ char **env_array_user_default(const char *username, int timeout, int mode)
close(fildes[0]);
if (!found) {
error("Failed to load current user environment variables");
- _load_env_cache(username);
+ return _load_env_cache(username);
}
/* First look for the start token in the output */
@@ -1424,15 +1439,16 @@ char **env_array_user_default(const char *username, int timeout, int mode)
break;
}
if (_env_array_entry_splitter(line, name, sizeof(name),
- value, sizeof(value)))
+ value, sizeof(value)) &&
+ (!_discard_env(name, value)))
env_array_overwrite(&env, name, value);
line = strtok_r(NULL, "\n", &last);
}
if (!found) {
error("Failed to get all user environment variables");
+ env_array_free(env);
return _load_env_cache(username);
}
return env;
}
-
diff --git a/src/common/env.h b/src/common/env.h
index 94000ed24f202354845e431ba3c42b5d84fa4925..3f8901d66565847d0ecda3680c5f93be959fe611 100644
--- a/src/common/env.h
+++ b/src/common/env.h
@@ -116,6 +116,7 @@ void env_array_for_job(char ***dest,
* SLURM_JOB_NODELIST
* SLURM_JOB_CPUS_PER_NODE
* ENVIRONMENT=BATCH
+ * HOSTNAME
* LOADLBATCH (AIX only)
*
* Sets OBSOLETE variables:
@@ -125,7 +126,9 @@ void env_array_for_job(char ***dest,
* SLURM_TASKS_PER_NODE <- poorly named, really CPUs per node
* ? probably only needed for users...
*/
-void env_array_for_batch_job(char ***dest, const batch_job_launch_msg_t *batch);
+extern void env_array_for_batch_job(char ***dest,
+ const batch_job_launch_msg_t *batch,
+ const char* node_name);
/*
* Set in "dest the environment variables relevant to a SLURM job step,
diff --git a/src/common/hostlist.c b/src/common/hostlist.c
index 36b489af1844a10cf32b1cd9987a308d8a555c11..19f340942d715b26125b95caa7cb57b5c5df01f0 100644
--- a/src/common/hostlist.c
+++ b/src/common/hostlist.c
@@ -1,5 +1,5 @@
/*****************************************************************************\
- * $Id: hostlist.c 12632 2007-11-06 23:27:07Z da $
+ * $Id: hostlist.c 13270 2008-02-14 19:40:44Z da $
*****************************************************************************
* $LSDId: hostlist.c,v 1.14 2003/10/14 20:11:54 grondo Exp $
*****************************************************************************
@@ -1558,7 +1558,8 @@ error:
* RET 1 if str contained a valid number or range,
* 0 if conversion of str to a range failed.
*/
-static int _parse_box_range(char *str, struct _range *ranges, int len, int *count)
+static int _parse_box_range(char *str, struct _range *ranges,
+ int len, int *count)
{
int a[3], b[3], i1, i2, i;
char new_str[8];
@@ -2411,14 +2412,28 @@ _get_boxes(char *buf, int max_len)
start_box = i;
end_box = i;
}
+
+
if (((len+8) < max_len) && (start_box != -1)
&& ((is_box == 0) || (i == axis_max_x))) {
- sprintf(buf+len,"%c%c%cx%c%c%c,",
- alpha_num[start_box], alpha_num[axis_min_y],
- alpha_num[axis_min_z],
- alpha_num[end_box], alpha_num[axis_max_y],
- alpha_num[axis_max_z]);
- len += 8;
+ if(start_box == end_box
+ && axis_min_y == axis_max_y
+ && axis_min_z == axis_max_z) {
+ sprintf(buf+len,"%c%c%c,",
+ alpha_num[start_box],
+ alpha_num[axis_min_y],
+ alpha_num[axis_min_z]);
+ len += 4;
+ } else {
+ sprintf(buf+len,"%c%c%cx%c%c%c,",
+ alpha_num[start_box],
+ alpha_num[axis_min_y],
+ alpha_num[axis_min_z],
+ alpha_num[end_box],
+ alpha_num[axis_max_y],
+ alpha_num[axis_max_z]);
+ len += 8;
+ }
start_box = -1;
end_box = -1;
}
diff --git a/src/common/node_select.c b/src/common/node_select.c
index 2f9a57d0cb56856cd0d57f96ce6d211bb85fb2d5..ede1c38fcfb1b80684b1afd52ed25c7b35e88c4f 100644
--- a/src/common/node_select.c
+++ b/src/common/node_select.c
@@ -9,7 +9,7 @@
* the plugin. This is because functions required by the plugin can not be
* resolved on the front-end nodes, so we can't load the plugins there.
*
- * $Id: node_select.c 12627 2007-11-06 19:48:55Z jette $
+ * $Id: node_select.c 13270 2008-02-14 19:40:44Z da $
*****************************************************************************
* Copyright (C) 2002-2006 The Regents of the University of California.
* Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER).
@@ -1114,16 +1114,20 @@ extern char *select_g_sprint_jobinfo(select_jobinfo_t jobinfo,
sprintf(start_char, "None");
else {
snprintf(start_char, sizeof(start_char),
- "%1ux%1ux%1u", jobinfo->start[0],
- jobinfo->start[1], jobinfo->start[2]);
+ "%cx%cx%c",
+ alpha_num[jobinfo->start[0]],
+ alpha_num[jobinfo->start[1]],
+ alpha_num[jobinfo->start[2]]);
}
snprintf(buf, size,
- "%7.7s %6.6s %6.6s %9s %1ux%1ux%1u %5s %-16s",
+ "%7.7s %6.6s %6.6s %9s %cx%cx%c %5s %-16s",
_job_conn_type_string(jobinfo->conn_type),
_yes_no_string(jobinfo->reboot),
_yes_no_string(jobinfo->rotate),
max_procs_char,
- geometry[0], geometry[1], geometry[2],
+ alpha_num[geometry[0]],
+ alpha_num[geometry[1]],
+ alpha_num[geometry[2]],
start_char, jobinfo->bg_block_id);
break;
case SELECT_PRINT_MIXED:
@@ -1137,18 +1141,22 @@ extern char *select_g_sprint_jobinfo(select_jobinfo_t jobinfo,
sprintf(start_char, "None");
else {
snprintf(start_char, sizeof(start_char),
- "%1ux%1ux%1u", jobinfo->start[0],
- jobinfo->start[1], jobinfo->start[2]);
+ "%cx%cx%c",
+ alpha_num[jobinfo->start[0]],
+ alpha_num[jobinfo->start[1]],
+ alpha_num[jobinfo->start[2]]);
}
snprintf(buf, size,
"Connection=%s Reboot=%s Rotate=%s MaxProcs=%s "
- "Geometry=%1ux%1ux%1u Start=%s Block_ID=%s",
+ "Geometry=%cx%cx%c Start=%s Block_ID=%s",
_job_conn_type_string(jobinfo->conn_type),
_yes_no_string(jobinfo->reboot),
_yes_no_string(jobinfo->rotate),
max_procs_char,
- geometry[0], geometry[1], geometry[2],
+ alpha_num[geometry[0]],
+ alpha_num[geometry[1]],
+ alpha_num[geometry[2]],
start_char, jobinfo->bg_block_id);
break;
case SELECT_PRINT_BG_ID:
@@ -1167,16 +1175,20 @@ extern char *select_g_sprint_jobinfo(select_jobinfo_t jobinfo,
_yes_no_string(jobinfo->rotate));
break;
case SELECT_PRINT_GEOMETRY:
- snprintf(buf, size, "%1ux%1ux%1u",
- geometry[0], geometry[1], geometry[2]);
+ snprintf(buf, size, "%cx%cx%c",
+ alpha_num[geometry[0]],
+ alpha_num[geometry[1]],
+ alpha_num[geometry[2]]);
break;
case SELECT_PRINT_START:
if (jobinfo->start[0] == (uint16_t) NO_VAL)
sprintf(buf, "None");
else {
snprintf(buf, size,
- "%1ux%1ux%1u", jobinfo->start[0],
- jobinfo->start[1], jobinfo->start[2]);
+ "%cx%cx%c",
+ alpha_num[jobinfo->start[0]],
+ alpha_num[jobinfo->start[1]],
+ alpha_num[jobinfo->start[2]]);
}
case SELECT_PRINT_MAX_PROCS:
if (jobinfo->max_procs == NO_VAL)
diff --git a/src/common/read_config.c b/src/common/read_config.c
index 01976c812a94bd3fe7bb6aead329a88118917b1f..79d44156de5ab70d7d04b4efd6bbe7257785bbcd 100644
--- a/src/common/read_config.c
+++ b/src/common/read_config.c
@@ -140,6 +140,7 @@ s_p_options_t slurm_conf_options[] = {
{"JobCredentialPrivateKey", S_P_STRING},
{"JobCredentialPublicCertificate", S_P_STRING},
{"JobFileAppend", S_P_UINT16},
+ {"GetEnvTimeout", S_P_UINT16},
{"KillTree", S_P_UINT16, defunct_option},
{"KillWait", S_P_UINT16},
{"MailProg", S_P_STRING},
@@ -315,7 +316,7 @@ static int parse_nodename(void **dest, slurm_parser_enum_t type,
if (!s_p_get_uint32(&n->tmp_disk, "TmpDisk", tbl)
&& !s_p_get_uint32(&n->tmp_disk, "TmpDisk", dflt))
- n->tmp_disk = 1;
+ n->tmp_disk = 0;
if (!s_p_get_uint32(&n->weight, "Weight", tbl)
&& !s_p_get_uint32(&n->weight, "Weight", dflt))
@@ -456,6 +457,13 @@ static int parse_partitionname(void **dest, slurm_parser_enum_t type,
if (!s_p_get_string(&p->nodes, "Nodes", tbl)
&& !s_p_get_string(&p->nodes, "Nodes", dflt))
p->nodes = NULL;
+ else {
+ int i;
+ for (i=0; p->nodes[i]; i++) {
+ if (isspace(p->nodes[i]))
+ p->nodes[i] = ',';
+ }
+ }
if (!s_p_get_boolean(&p->root_only_flag, "RootOnly", tbl)
&& !s_p_get_boolean(&p->root_only_flag, "RootOnly", dflt))
@@ -1432,6 +1440,9 @@ validate_and_set_defaults(slurm_ctl_conf_t *conf, s_p_hashtbl_t *hashtbl)
if (!s_p_get_uint16(&conf->job_file_append, "JobFileAppend", hashtbl))
conf->job_file_append = 0;
+ if (!s_p_get_uint16(&conf->get_env_timeout, "GetEnvTimeout", hashtbl))
+ conf->get_env_timeout = DEFAULT_GET_ENV_TIMEOUT;
+
if (!s_p_get_uint16(&conf->kill_wait, "KillWait", hashtbl))
conf->kill_wait = DEFAULT_KILL_WAIT;
diff --git a/src/common/read_config.h b/src/common/read_config.h
index 6817fe34ab109bf7f2c060ad043991eb9147ce15..70215d19926e7507becde5773613d36c9ddd85a1 100644
--- a/src/common/read_config.h
+++ b/src/common/read_config.h
@@ -52,6 +52,7 @@ extern char *default_plugstack;
#define DEFAULT_CACHE_GROUPS 0
#define DEFAULT_FAST_SCHEDULE 1
#define DEFAULT_FIRST_JOB_ID 1
+#define DEFAULT_GET_ENV_TIMEOUT 2
/* NOTE: DEFAULT_INACTIVE_LIMIT must be 0 for Blue Gene/L systems */
#define DEFAULT_INACTIVE_LIMIT 0
#define DEFAULT_JOB_ACCT_LOGFILE "/var/log/slurm_accounting.log"
diff --git a/src/common/slurm_protocol_api.c b/src/common/slurm_protocol_api.c
index 550e855148bc9dca7bffd87449c0d74c29a5cd68..d8d87d88cd3167e575df0c5987dcfc1ca8ee8ef9 100644
--- a/src/common/slurm_protocol_api.c
+++ b/src/common/slurm_protocol_api.c
@@ -182,6 +182,20 @@ void slurm_api_clear_config(void)
/* slurm_mutex_lock(&config_lock); */
/* } */
+/* slurm_get_env_timeout
+ * return default timeout for srun/sbatch --get-user-env option
+ */
+int inline slurm_get_env_timeout(void)
+{
+ int timeout;
+ slurm_ctl_conf_t *conf;
+
+ conf = slurm_conf_lock();
+ timeout = conf->get_env_timeout;
+ slurm_conf_unlock();
+ return timeout;
+}
+
/* slurm_get_mpi_default
* get default mpi value from slurmctld_conf object
* RET char * - mpi default value from slurm.conf, MUST be xfreed by caller
diff --git a/src/common/slurm_protocol_api.h b/src/common/slurm_protocol_api.h
index bd5614c56b4f7efd8d8edcc009b3c1710be746fd..10c19e124a8906530dddc99c7861f8a2d560673c 100644
--- a/src/common/slurm_protocol_api.h
+++ b/src/common/slurm_protocol_api.h
@@ -96,6 +96,11 @@ int inline slurm_set_api_config(slurm_protocol_config_t * protocol_conf);
*/
inline slurm_protocol_config_t *slurm_get_api_config();
+/* slurm_get_env_timeout
+ * return default timeout for srun/sbatch --get-user-env option
+ */
+int inline slurm_get_env_timeout(void);
+
/* slurm_get_mpi_default
* get default mpi value from slurmctld_conf object
* RET char * - mpi default value from slurm.conf, MUST be xfreed by caller
diff --git a/src/common/slurm_step_layout.c b/src/common/slurm_step_layout.c
index 555f7d65fea78b9d956efa60452a18bcb44142aa..e49e01dc443aa7cbc81fb2c34d86fb8f5401043f 100644
--- a/src/common/slurm_step_layout.c
+++ b/src/common/slurm_step_layout.c
@@ -679,12 +679,12 @@ static int _task_layout_plane(slurm_step_layout_t *step_layout,
return SLURM_ERROR;
}
+#if(0)
+ /* debugging only */
for (i=0; i < step_layout->node_cnt; i++) {
info("tasks[%d]: %u", i, step_layout->tasks[i]);
}
-#if(0)
- /* debugging only */
for (i=0; i < step_layout->node_cnt; i++) {
info ("Host %d _plane_ # of tasks %u", i, step_layout->tasks[i]);
for (j=0; j<step_layout->tasks[i]; j++) {
diff --git a/src/plugins/mpi/mvapich/mvapich.c b/src/plugins/mpi/mvapich/mvapich.c
index ac07bedab2c97337be5d166b0b1a3cf2c8124b8a..38f1f5d8ecab0bd3c67861158089aa08f123ca6e 100644
--- a/src/plugins/mpi/mvapich/mvapich.c
+++ b/src/plugins/mpi/mvapich/mvapich.c
@@ -320,8 +320,12 @@ static int mvapich_read_n (mvapich_state_t *st, struct mvapich_info *mvi,
return (-1);
}
- if (n == 0) /* unexpected EOF */
+ if (n == 0) { /* unexpected EOF */
+ error ("mvapich: rank %d: "
+ "Unexpected EOF (%dB left to read)",
+ mvi->rank, nleft);
return (-1);
+ }
nleft -= n;
p += n;
@@ -554,52 +558,55 @@ static void mvapich_bcast_hostids (mvapich_state_t *st)
}
/* Write size bytes from buf into socket for rank */
-static void mvapich_send (mvapich_state_t *st, void* buf, int size, int rank)
+static int mvapich_send (mvapich_state_t *st, void* buf, int size, int rank)
{
struct mvapich_info *mvi = st->mvarray [rank];
- if (mvapich_write_n (st, mvi, buf, size) < 0)
- error ("mvapich: write hostid rank %d: %m", mvi->rank);
+ return (mvapich_write_n (st, mvi, buf, size));
}
/* Read size bytes from socket for rank into buf */
-static void mvapich_recv (mvapich_state_t *st, void* buf, int size, int rank)
+static int mvapich_recv (mvapich_state_t *st, void* buf, int size, int rank)
{
struct mvapich_info *mvi = st->mvarray [rank];
- if (mvapich_read_n (st, mvi, buf, size) <= 0)
- error("mvapich reading from %d: %m", mvi->rank);
-}
-
-/* Read an integer from socket for rank */
-static int mvapich_recv_int (mvapich_state_t *st, int rank)
-{
- int buf;
- mvapich_recv(st, &buf, sizeof(buf), rank);
- return buf;
+ return (mvapich_read_n (st, mvi, buf, size));
}
/* Scatter data in buf to ranks using chunks of size bytes */
-static void mvapich_scatterbcast (mvapich_state_t *st, void* buf, int size)
+static int mvapich_scatterbcast (mvapich_state_t *st, void* buf, int size)
{
- int i;
- for (i = 0; i < st->nprocs; i++)
- mvapich_send(st, buf + i*size, size, i);
+ int i, rc;
+ int n = 0;
+
+ for (i = 0; i < st->nprocs; i++) {
+ if ((rc = mvapich_send (st, buf + i*size, size, i)) <= 0)
+ return (-1);
+ n += rc;
+ }
+ return (n);
}
/* Broadcast buf to each rank, which is size bytes big */
-static void mvapich_allgatherbcast (mvapich_state_t *st, void* buf, int size)
+static int mvapich_allgatherbcast (mvapich_state_t *st, void* buf, int size)
{
- int i;
- for (i = 0; i < st->nprocs; i++)
- mvapich_send(st, buf, size, i);
+ int i, rc;
+ int n = 0;
+
+ for (i = 0; i < st->nprocs; i++) {
+ if ((rc = mvapich_send (st, buf, size, i)) <= 0)
+ return (-1);
+ n += rc;
+ }
+ return (n);
}
/* Perform alltoall using data in buf with elements of size bytes */
-static void mvapich_alltoallbcast (mvapich_state_t *st, void* buf, int size)
+static int mvapich_alltoallbcast (mvapich_state_t *st, void* buf, int size)
{
int pbufsize = size * st->nprocs;
void* pbuf = xmalloc(pbufsize);
+ int i, src, rc;
+ int n = 0;
- int i, src;
for (i = 0; i < st->nprocs; i++) {
for (src = 0; src < st->nprocs; src++) {
memcpy( pbuf + size*src,
@@ -607,22 +614,141 @@ static void mvapich_alltoallbcast (mvapich_state_t *st, void* buf, int size)
size
);
}
- mvapich_send(st, pbuf, pbufsize, i);
+ if ((rc = mvapich_send (st, pbuf, pbufsize, i)) <= 0)
+ goto out;
+ n += rc;
}
+ out:
xfree(pbuf);
+ return (rc < 0 ? rc : n);
+}
+
+static int recv_common_value (mvapich_state_t *st, int *valp, int rank)
+{
+ int val;
+ if (mvapich_recv (st, &val, sizeof (int), rank) <= 0) {
+ error ("mvapich: recv: rank %d: %m\n", rank);
+ return (-1);
+ }
+
+ /*
+ * If value is uninitialized, set it to current value,
+ * otherwise ensure that current value matches previous
+ */
+ if (*valp == -1)
+ *valp = val;
+ else if (val != *valp) {
+ error ("mvapich: PMGR: unexpected value from rank %d: "
+ "expected %d, recvd %d", rank, *valp, val);
+ return (-1);
+ }
+ return (0);
+}
+
+/*
+ * PMGR_BCAST (root, size of message, then message data (from root only))
+ */
+static int process_pmgr_bcast (mvapich_state_t *st, int *rootp, int *sizep,
+ void ** bufp, int rank)
+{
+ if (recv_common_value (st, rootp, rank) < 0)
+ return (-1);
+ if (recv_common_value (st, sizep, rank) < 0)
+ return (-1);
+ if (rank != *rootp)
+ return (0);
+
+ /*
+ * Recv data from root
+ */
+ *bufp = xmalloc (*sizep);
+ if (mvapich_recv (st, *bufp, *sizep, rank) < 0) {
+ error ("mvapich: PMGR_BCAST: Failed to recv from root: %m");
+ return (-1);
+ }
+ return (0);
+}
+
+/*
+ * PMGR_GATHER (root, size of message, then message data)
+ */
+static int process_pmgr_gather (mvapich_state_t *st, int *rootp,
+ int *sizep, void **bufp, int rank)
+{
+ if (recv_common_value (st, rootp, rank) < 0)
+ return (-1);
+ if (recv_common_value (st, sizep, rank) < 0)
+ return (-1);
+ if (*bufp == NULL)
+ *bufp = xmalloc (*sizep * st->nprocs);
+
+ if (mvapich_recv(st, (*bufp) + (*sizep)*rank, *sizep, rank) < 0) {
+ error ("mvapich: PMGR_/GATHER: rank %d: recv: %m", rank);
+ return (-1);
+ }
+ return (0);
}
-/* Check that new == curr value if curr has been initialized */
-static int set_current (int curr, int new)
+/*
+ * PMGR_SCATTER (root, size of message, then message data)
+ */
+static int process_pmgr_scatter (mvapich_state_t *st, int *rootp,
+ int *sizep, void **bufp, int rank)
{
- if (curr == -1)
- curr = new;
- if (new != curr) {
- error("PMGR unexpected value: received %d, expecting %d",
- new, curr);
+ if (recv_common_value (st, rootp, rank) < 0)
+ return (-1);
+ if (recv_common_value (st, sizep, rank) < 0)
+ return (-1);
+ if (rank != *rootp)
+ return (0);
+
+ if (*bufp == NULL)
+ *bufp = xmalloc (*sizep * st->nprocs);
+
+ if (mvapich_recv(st, *bufp, (*sizep) * st->nprocs, rank) < 0) {
+ error ("mvapich: PMGR_SCATTER: rank %d: recv: %m", rank);
+ return (-1);
}
- return curr;
+ return (0);
+}
+
+/*
+ * PMGR_ALLGATHER (size of message, then message data)
+ */
+static int process_pmgr_allgather (mvapich_state_t *st, int *sizep,
+ void **bufp, int rank)
+{
+ if (recv_common_value (st, sizep, rank) < 0)
+ return (-1);
+ if (*bufp == NULL)
+ *bufp = xmalloc (*sizep * st->nprocs);
+ if (mvapich_recv (st, (*bufp) + *sizep*rank, *sizep, rank) < 0) {
+ error ("mvapich: PMGR_ALLGATHER: rank %d: %m", rank);
+ return (-1);
+ }
+ return (0);
+}
+
+/*
+ * PMGR_ALLTOALL (size of message, then message data)
+ */
+static int process_pmgr_alltoall (mvapich_state_t *st, int *sizep,
+ void **bufp, int rank)
+{
+ if (recv_common_value (st, sizep, rank) < 0)
+ return (-1);
+
+ if (*bufp == NULL)
+ *bufp = xmalloc (*sizep * st->nprocs * st->nprocs);
+ if (mvapich_recv ( st,
+ *bufp + (*sizep * st->nprocs)*rank,
+ *sizep * st->nprocs, rank ) < 0) {
+ error ("mvapich: PMGR_ALLTOALL: recv: rank %d: %m", rank);
+ return (-1);
+ }
+
+ return (0);
}
/*
@@ -643,7 +769,7 @@ static int set_current (int curr, int new)
* Note: Although there are op codes available for PMGR_OPEN and
* PMGR_ABORT, neither is fully implemented and should not be used.
*/
-static void mvapich_processops (mvapich_state_t *st)
+static int mvapich_processops (mvapich_state_t *st)
{
/* Until a 'CLOSE' or 'ABORT' message is seen, we continuously
* loop processing ops
@@ -663,57 +789,57 @@ static void mvapich_processops (mvapich_state_t *st)
struct mvapich_info *mvi = st->mvarray [i];
// read in opcode
- opcode = set_current(opcode, mvapich_recv_int(st, i));
+ if (recv_common_value (st, &opcode, i) < 0) {
+ error ("mvapich: rank %d: Failed to read opcode: %m",
+ mvi->rank);
+ return (-1);
+ }
// read in additional data depending on current opcode
int rank, code;
switch(opcode) {
case 0: // PMGR_OPEN (followed by rank)
- rank = mvapich_recv_int(st, i);
+ if (mvapich_recv (st, &rank, sizeof (int), i) <= 0) {
+ error ("mvapich: PMGR_OPEN: recv: %m");
+ exit = 1;
+ }
break;
case 1: // PMGR_CLOSE (no data, close the socket)
close(mvi->fd);
break;
case 2: // PMGR_ABORT (followed by exit code)
- code = mvapich_recv_int(st, i);
+ if (mvapich_recv (st, &code, sizeof (int), i) <= 0) {
+ error ("mvapich: PMGR_ABORT: recv: %m");
+ }
error("mvapich abort with code %d from rank %d",
code, i);
break;
case 3: // PMGR_BARRIER (no data)
break;
- case 4: // PMGR_BCAST (root, size of message,
- // then message data (from root only))
- root = set_current(root, mvapich_recv_int(st, i));
- size = set_current(size, mvapich_recv_int(st, i));
- if (!buf) buf = (void*) xmalloc(size);
- if (i == root) mvapich_recv(st, buf, size, i);
+ case 4: // PMGR_BCAST
+ if (process_pmgr_bcast (st, &root, &size, &buf, i) < 0)
+ return (-1);
break;
- case 5: // PMGR_GATHER (root, size of message,
- // then message data)
- root = set_current(root, mvapich_recv_int(st, i));
- size = set_current(size, mvapich_recv_int(st, i));
- if (!buf) buf = (void*) xmalloc(size * st->nprocs);
- mvapich_recv(st, buf + size*i, size, i);
+ case 5: // PMGR_GATHER
+ if (process_pmgr_gather (st, &root, &size, &buf, i) < 0)
+ return (-1);
break;
- case 6: // PMGR_SCATTER (root, size of message,
- // then message data)
- root = set_current(root, mvapich_recv_int(st, i));
- size = set_current(size, mvapich_recv_int(st, i));
- if (!buf) buf = (void*) xmalloc(size * st->nprocs);
- if (i == root) mvapich_recv(st, buf, size * st->nprocs, i);
+ case 6: // PMGR_SCATTER
+ if (process_pmgr_scatter (st, &root,
+ &size, &buf, i) < 0)
+ return (-1);
break;
- case 7: // PMGR_ALLGATHER (size of message, then message data)
- size = set_current(size, mvapich_recv_int(st, i));
- if (!buf) buf = (void*) xmalloc(size * st->nprocs);
- mvapich_recv(st, buf + size*i, size, i);
+ case 7: // PMGR_ALLGATHER
+ if (process_pmgr_allgather (st, &size, &buf, i) < 0)
+ return (-1);
break;
- case 8: // PMGR_ALLTOALL (size of message, then message data)
- size = set_current(size, mvapich_recv_int(st, i));
- if (!buf) buf = (void*) xmalloc(size * st->nprocs * st->nprocs);
- mvapich_recv(st, buf + (size*st->nprocs)*i, size * st->nprocs, i);
+ case 8: // PMGR_ALLTOALL
+ if (process_pmgr_alltoall (st, &size, &buf, i) < 0)
+ return (-1);
break;
default:
error("Unrecognized PMGR opcode: %d", opcode);
+ return (-1);
}
}
@@ -767,6 +893,7 @@ static void mvapich_processops (mvapich_state_t *st)
xfree(buf);
} // while(!exit)
mvapich_debug ("Completed processing PMGR opcodes");
+ return (0);
}
static void mvapich_bcast (mvapich_state_t *st)
@@ -1158,7 +1285,8 @@ again:
}
if (st->protocol_version == 8) {
- mvapich_processops(st);
+ if (mvapich_processops(st) < 0)
+ goto fail;
} else {
mvapich_debug ("bcasting mvapich info to %d tasks", st->nprocs);
mvapich_bcast (st);
diff --git a/src/plugins/sched/wiki/get_jobs.c b/src/plugins/sched/wiki/get_jobs.c
index 70c9d91aa8a9107333ca3259bc82238bbe939edf..6c6fd1f626b198e50007562da7f232e07c224e6b 100644
--- a/src/plugins/sched/wiki/get_jobs.c
+++ b/src/plugins/sched/wiki/get_jobs.c
@@ -301,8 +301,14 @@ static char * _dump_job(struct job_record *job_ptr, int state_info)
}
if (job_ptr->account) {
- snprintf(tmp, sizeof(tmp),
- "ACCOUNT=%s;", job_ptr->account);
+ /* allow QOS spec in form "qos-name" */
+ if (!strncmp(job_ptr->account,"qos-",4)) {
+ snprintf(tmp, sizeof(tmp),
+ "QOS=%s;", job_ptr->account + 4);
+ } else {
+ snprintf(tmp, sizeof(tmp),
+ "ACCOUNT=%s;", job_ptr->account);
+ }
xstrcat(buf, tmp);
}
diff --git a/src/plugins/sched/wiki/start_job.c b/src/plugins/sched/wiki/start_job.c
index f7cdd46d189ad8b080c780beb3ac794e7a9ca0ee..180e79742dd7a46e2e30c8830cfc217023da8274 100644
--- a/src/plugins/sched/wiki/start_job.c
+++ b/src/plugins/sched/wiki/start_job.c
@@ -195,6 +195,7 @@ static int _start_job(uint32_t jobid, int task_cnt, char *hostlist,
* performs many string compares. */
xfree(job_ptr->details->req_node_layout);
if (task_cnt && cr_enabled) {
+ uint16_t cpus_per_task = MAX(1, job_ptr->details->cpus_per_task);
job_ptr->details->req_node_layout = (uint16_t *)
xmalloc(bit_set_count(new_bitmap) * sizeof(uint16_t));
bsize = bit_size(new_bitmap);
@@ -212,7 +213,8 @@ static int _start_job(uint32_t jobid, int task_cnt, char *hostlist,
if ((node_idx[node_name_len] == ',') ||
(node_idx[node_name_len] == '\0')) {
job_ptr->details->
- req_node_layout[ll]++;
+ req_node_layout[ll] +=
+ cpus_per_task;
}
node_cur = strchr(node_idx, ',');
if (node_cur)
diff --git a/src/plugins/sched/wiki2/start_job.c b/src/plugins/sched/wiki2/start_job.c
index d6fd7dd8201d8e1f633723ccceb90d5c1e63f800..564919f93ad682e69be342566ead6493d7cfc754 100644
--- a/src/plugins/sched/wiki2/start_job.c
+++ b/src/plugins/sched/wiki2/start_job.c
@@ -251,6 +251,7 @@ static int _start_job(uint32_t jobid, int task_cnt, char *hostlist,
* performs many string compares. */
xfree(job_ptr->details->req_node_layout);
if (task_cnt && cr_enabled) {
+ uint16_t cpus_per_task = MAX(1, job_ptr->details->cpus_per_task);
job_ptr->details->req_node_layout = (uint16_t *)
xmalloc(bit_set_count(new_bitmap) * sizeof(uint16_t));
bsize = bit_size(new_bitmap);
@@ -268,7 +269,8 @@ static int _start_job(uint32_t jobid, int task_cnt, char *hostlist,
if ((node_idx[node_name_len] == ',') ||
(node_idx[node_name_len] == '\0')) {
job_ptr->details->
- req_node_layout[ll]++;
+ req_node_layout[ll] +=
+ cpus_per_task;
}
node_cur = strchr(node_idx, ',');
if (node_cur)
diff --git a/src/plugins/select/bluegene/block_allocator/block_allocator.c b/src/plugins/select/bluegene/block_allocator/block_allocator.c
index f8cc7dc0967409d637ea40ee885847846b71556e..9c0467374b800c4b20621eacd58353a18f3bd019 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 bglblocks,
* wiring, mapping for smap, etc.
- * $Id: block_allocator.c 12543 2007-10-23 22:19:49Z jette $
+ * $Id: block_allocator.c 13150 2008-01-31 22:59:13Z da $
*****************************************************************************
* Copyright (C) 2004 The Regents of the University of California.
* Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER).
@@ -1264,6 +1264,7 @@ extern int copy_node_path(List nodes, List dest_nodes)
#endif
return rc;
}
+
extern int check_and_set_node_list(List nodes)
{
int rc = SLURM_ERROR;
diff --git a/src/plugins/select/bluegene/block_allocator/block_allocator.h b/src/plugins/select/bluegene/block_allocator/block_allocator.h
index cc69f7d3595b8a25310b0d120a1b88c0c26763dc..91fe587478f55e963a80a597abcc149801ce8259 100644
--- a/src/plugins/select/bluegene/block_allocator/block_allocator.h
+++ b/src/plugins/select/bluegene/block_allocator/block_allocator.h
@@ -109,6 +109,7 @@ typedef struct {
bool rotate;
bool elongate;
List elongate_geos;
+ bitstr_t *avail_node_bitmap; /* pointer to available nodes */
} ba_request_t;
typedef struct {
@@ -172,7 +173,7 @@ typedef struct
* ba_node_t: node within the allocation system.
*/
typedef struct {
- /* set if using this node in a block*/
+ /* set if using this node in a block */
bool used;
/* coordinates */
diff --git a/src/plugins/select/bluegene/plugin/bg_job_place.c b/src/plugins/select/bluegene/plugin/bg_job_place.c
index f03eae075ad2d68af028d2bc7f778946e34fdfe1..30d60327b72102027de95209dd15428506f0c44f 100644
--- a/src/plugins/select/bluegene/plugin/bg_job_place.c
+++ b/src/plugins/select/bluegene/plugin/bg_job_place.c
@@ -2,7 +2,7 @@
* bg_job_place.c - blue gene job placement (e.g. base block selection)
* functions.
*
- * $Id: bg_job_place.c 12627 2007-11-06 19:48:55Z jette $
+ * $Id: bg_job_place.c 13271 2008-02-14 20:02:00Z da $
*****************************************************************************
* Copyright (C) 2004-2007 The Regents of the University of California.
* Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER).
@@ -209,7 +209,6 @@ static int _find_best_block_match(struct job_record* job_ptr,
uint32_t req_procs = job_ptr->num_procs;
uint32_t proc_cnt;
ba_request_t request;
- ba_request_t *try_request = NULL;
int i;
int rot_cnt = 0;
int created = 0;
@@ -256,50 +255,7 @@ static int _find_best_block_match(struct job_record* job_ptr,
if(start[X] != (uint16_t)NO_VAL)
start_req = 1;
- if(num_unused_cpus != total_cpus) {
- /*
- see if we have already tried to create this
- size but couldn't make it right now no reason
- to try again
- */
- slurm_mutex_lock(&request_list_mutex);
- itr = list_iterator_create(bg_request_list);
- while ((try_request = list_next(itr))) {
- if(start_req) {
- if ((try_request->start[X] != start[X])
- || (try_request->start[Y] != start[Y])
- || (try_request->start[Z] != start[Z])) {
- debug4("got %c%c%c looking for %c%c%c",
- alpha_num[try_request->start[X]],
- alpha_num[try_request->start[Y]],
- alpha_num[try_request->start[Z]],
- alpha_num[start[X]],
- alpha_num[start[Y]],
- alpha_num[start[Z]]);
- continue;
- }
- debug3("found %c%c%c looking for %c%c%c",
- alpha_num[try_request->start[X]],
- alpha_num[try_request->start[Y]],
- alpha_num[try_request->start[Z]],
- alpha_num[start[X]],
- alpha_num[start[Y]],
- alpha_num[start[Z]]);
- }
- if(try_request->procs == req_procs) {
- debug("already tried to create but "
- "can't right now.");
- list_iterator_destroy(itr);
- slurm_mutex_unlock(&request_list_mutex);
- if(test_only)
- return SLURM_SUCCESS;
- else
- return SLURM_ERROR;
- }
- }
- list_iterator_destroy(itr);
- slurm_mutex_unlock(&request_list_mutex);
- }
+
select_g_get_jobinfo(job_ptr->select_jobinfo,
SELECT_DATA_CONN_TYPE, &conn_type);
select_g_get_jobinfo(job_ptr->select_jobinfo,
@@ -471,8 +427,8 @@ try_again:
req_procs, max_procs, proc_cnt);
if ((proc_cnt < req_procs)
|| ((max_procs != NO_VAL) && (proc_cnt > max_procs))) {
- /* We use the proccessor count per partition here
- mostly to see if we can run on a smaller partition.
+ /* We use the proccessor count per block here
+ mostly to see if we can run on a smaller block.
*/
convert_num_unit((float)proc_cnt, tmp_char,
sizeof(tmp_char), UNIT_NONE);
@@ -522,7 +478,7 @@ try_again:
continue;
}
- /* Make sure no other partitions are under this partition
+ /* Make sure no other blocks are under this block
are booted and running jobs
*/
itr2 = list_iterator_create(bg_list);
@@ -719,7 +675,12 @@ try_again:
request.linuximage = linuximage;
request.mloaderimage = mloaderimage;
request.ramdiskimage = ramdiskimage;
-
+ if(job_ptr->details->req_node_bitmap)
+ request.avail_node_bitmap =
+ job_ptr->details->req_node_bitmap;
+ else
+ request.avail_node_bitmap = slurm_block_bitmap;
+
debug("trying with all free blocks");
if(create_dynamic_block(&request, NULL) == SLURM_ERROR) {
error("this job will never run on "
@@ -735,24 +696,6 @@ try_again:
goto end_it;
}
- /*
- add request to list so we don't try again until
- something happens like a job finishing or
- something so we can try again
- */
- debug3("adding %d %d",
- request.procs, request.conn_type);
- try_request = xmalloc(sizeof(ba_request_t));
- try_request->procs = req_procs;
- try_request->save_name = NULL;
- try_request->elongate_geos = NULL;
- try_request->start_req = request.start_req;
- for(i=0; i<BA_SYSTEM_DIMENSIONS; i++)
- try_request->start[i] = start[i];
- slurm_mutex_lock(&request_list_mutex);
- list_push(bg_request_list, try_request);
- slurm_mutex_unlock(&request_list_mutex);
-
slurm_conf_lock();
snprintf(tmp_char, sizeof(tmp_char), "%s%s",
slurmctld_conf.node_prefix,
@@ -812,6 +755,12 @@ try_again:
request.linuximage = linuximage;
request.mloaderimage = mloaderimage;
request.ramdiskimage = ramdiskimage;
+ if(job_ptr->details->req_node_bitmap)
+ request.avail_node_bitmap =
+ job_ptr->details->req_node_bitmap;
+ else
+ request.avail_node_bitmap = slurm_block_bitmap;
+
/* 1- try empty space
2- we see if we can create one in the
unused bps
diff --git a/src/plugins/select/bluegene/plugin/bg_job_run.c b/src/plugins/select/bluegene/plugin/bg_job_run.c
index f656c0525af4e0d881931a9c0d5a7dd6a2b2b787..5153de5689352b5fa20f5fac44a27691416fe642 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 11274 2007-03-30 19:39:49Z da $
+ * $Id: bg_job_run.c 13271 2008-02-14 20:02:00Z da $
*****************************************************************************
* Copyright (C) 2004-2006 The Regents of the University of California.
* Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER).
@@ -589,9 +589,6 @@ static void _term_agent(bg_update_t *bg_update_ptr)
bg_record->bg_block_id,
bg_record->user_name);
- if(bluegene_layout_mode == LAYOUT_DYNAMIC)
- remove_from_request_list();
-
if(job_remove_failed) {
char time_str[32];
slurm_make_time_str(&now, time_str, sizeof(time_str));
diff --git a/src/plugins/select/bluegene/plugin/bluegene.c b/src/plugins/select/bluegene/plugin/bluegene.c
index c684d32f53578c3149d8ad49244e927e0f86b4fb..30684bf23bbe289a75e13cf728871205fd3dbff2 100644
--- a/src/plugins/select/bluegene/plugin/bluegene.c
+++ b/src/plugins/select/bluegene/plugin/bluegene.c
@@ -1,7 +1,7 @@
/*****************************************************************************\
* bluegene.c - blue gene node configuration processing module.
*
- * $Id: bluegene.c 12450 2007-10-05 18:22:36Z da $
+ * $Id: bluegene.c 13271 2008-02-14 20:02:00Z da $
*****************************************************************************
* Copyright (C) 2004 The Regents of the University of California.
* Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER).
@@ -59,8 +59,6 @@ List bg_found_block_list = NULL; /* found bg blocks already on system */
List bg_job_block_list = NULL; /* jobs running in these blocks */
List bg_booted_block_list = NULL; /* blocks that are booted */
List bg_freeing_list = NULL; /* blocks that being freed */
-List bg_request_list = NULL; /* list of request that can't
- be made just yet */
List bg_blrtsimage_list = NULL;
List bg_linuximage_list = NULL;
@@ -80,7 +78,6 @@ uint16_t bridge_api_verb = 0;
bool agent_fini = false;
time_t last_bg_update;
pthread_mutex_t block_state_mutex = PTHREAD_MUTEX_INITIALIZER;
-pthread_mutex_t request_list_mutex = PTHREAD_MUTEX_INITIALIZER;
int num_block_to_free = 0;
int num_block_freed = 0;
int blocks_are_created = 0;
@@ -185,13 +182,6 @@ extern void fini_bg(void)
while(destroy_cnt > 0)
usleep(1000);
- slurm_mutex_lock(&request_list_mutex);
- if (bg_request_list) {
- list_destroy(bg_request_list);
- bg_request_list = NULL;
- }
- slurm_mutex_unlock(&request_list_mutex);
-
if(bg_blrtsimage_list) {
list_destroy(bg_blrtsimage_list);
bg_blrtsimage_list = NULL;
@@ -407,6 +397,7 @@ extern void process_nodes(bg_record_t *bg_record)
start);
if(bg_record->nodes[j] != ',')
break;
+ j--;
}
j++;
}
@@ -423,6 +414,8 @@ extern void process_nodes(bg_record_t *bg_record)
itr = list_iterator_create(bg_record->bg_block_list);
while ((ba_node = list_next(itr)) != NULL) {
+ if(!ba_node->used)
+ continue;
debug4("%c%c%c is included in this block",
alpha_num[ba_node->coord[X]],
alpha_num[ba_node->coord[Y]],
@@ -1205,6 +1198,89 @@ extern int create_dynamic_block(ba_request_t *request, List my_block_list)
debug("No list was given");
}
+ if(request->avail_node_bitmap) {
+ int j=0, number;
+ int x,y,z;
+ char *nodes = NULL;
+ bitstr_t *bitmap = bit_alloc(node_record_count);
+ int start[BA_SYSTEM_DIMENSIONS];
+ int end[BA_SYSTEM_DIMENSIONS];
+
+ /* we want the bps that aren't in this partition to
+ * mark them as used
+ */
+ bit_or(bitmap, request->avail_node_bitmap);
+ bit_not(bitmap);
+ nodes = bitmap2node_name(bitmap);
+ //info("not using %s", nodes);
+ while(nodes[j] != '\0') {
+ if ((nodes[j] == '[' || nodes[j] == ',')
+ && (nodes[j+8] == ']' || nodes[j+8] == ',')
+ && (nodes[j+4] == 'x' || nodes[j+4] == '-')) {
+ j++;
+ number = xstrntol(nodes + j,
+ NULL, BA_SYSTEM_DIMENSIONS,
+ HOSTLIST_BASE);
+ start[X] = number /
+ (HOSTLIST_BASE * HOSTLIST_BASE);
+ start[Y] = (number %
+ (HOSTLIST_BASE * HOSTLIST_BASE))
+ / HOSTLIST_BASE;
+ start[Z] = (number % HOSTLIST_BASE);
+ j += 4;
+ number = xstrntol(nodes + j,
+ NULL, 3, HOSTLIST_BASE);
+ end[X] = number /
+ (HOSTLIST_BASE * HOSTLIST_BASE);
+ end[Y] = (number
+ % (HOSTLIST_BASE * HOSTLIST_BASE))
+ / HOSTLIST_BASE;
+ end[Z] = (number % HOSTLIST_BASE);
+ j += 3;
+ for (x = start[X]; x <= end[X]; x++) {
+ for (y = start[Y]; y <= end[Y]; y++) {
+ for (z = start[Z];
+ z <= end[Z]; z++) {
+ ba_system_ptr->
+ grid[x]
+#ifdef HAVE_BG
+ [y][z]
+#endif
+ .used = 1;
+ }
+ }
+ }
+
+ if(nodes[j] != ',')
+ break;
+ j--;
+ } else if((nodes[j] >= '0' && nodes[j] <= '9')
+ || (nodes[j] >= 'A' && nodes[j] <= 'Z')) {
+
+ number = xstrntol(nodes + j,
+ NULL, BA_SYSTEM_DIMENSIONS,
+ HOSTLIST_BASE);
+ x = number / (HOSTLIST_BASE * HOSTLIST_BASE);
+ y = (number % (HOSTLIST_BASE * HOSTLIST_BASE))
+ / HOSTLIST_BASE;
+ z = (number % HOSTLIST_BASE);
+ j+=3;
+ ba_system_ptr->grid[x]
+#ifdef HAVE_BG
+ [y][z]
+#endif
+ .used = 1;
+
+ if(nodes[j] != ',')
+ break;
+ j--;
+ }
+ j++;
+ }
+ xfree(nodes);
+ FREE_NULL_BITMAP(bitmap);
+ }
+
if(request->size==1 && request->procs < bluegene_bp_node_cnt) {
request->conn_type = SELECT_SMALL;
if(request->procs == (procs_per_node/16)) {
@@ -1228,14 +1304,14 @@ extern int create_dynamic_block(ba_request_t *request, List my_block_list)
}
num_quarter=4;
}
-
+
if(_breakup_blocks(request, my_block_list) != SLURM_SUCCESS) {
debug2("small block not able to be placed");
//rc = SLURM_ERROR;
} else
goto finished;
}
-
+
if(request->conn_type == SELECT_NAV)
request->conn_type = SELECT_TORUS;
@@ -1531,31 +1607,6 @@ extern int remove_from_bg_list(List my_bg_list, bg_record_t *bg_record)
return rc;
}
-extern int remove_from_request_list()
-{
- ba_request_t *try_request = NULL;
- ListIterator itr;
- int rc = SLURM_ERROR;
-
- /*
- remove all requests out of the list.
- */
-
- slurm_mutex_lock(&request_list_mutex);
- itr = list_iterator_create(bg_request_list);
- while ((try_request = list_next(itr)) != NULL) {
- debug3("removing size %d",
- try_request->procs);
- list_remove(itr);
- delete_ba_request(try_request);
- //list_iterator_reset(itr);
- rc = SLURM_SUCCESS;
- }
- list_iterator_destroy(itr);
- slurm_mutex_unlock(&request_list_mutex);
- return rc;
-}
-
extern int bg_free_block(bg_record_t *bg_record)
{
#ifdef HAVE_BG_FILES
@@ -1705,8 +1756,6 @@ extern void *mult_destroy_block(void *args)
*/
sort_bg_record_inc_size(bg_freeing_list);
- remove_from_request_list();
-
slurm_mutex_lock(&block_state_mutex);
if(remove_from_bg_list(bg_job_block_list, bg_record)
== SLURM_SUCCESS) {
@@ -2361,7 +2410,7 @@ static int _addto_node_list(bg_record_t *bg_record, int *start, int *end)
alpha_num[DIM_SIZE[X]], alpha_num[DIM_SIZE[Y]],
alpha_num[DIM_SIZE[Z]]);
}
- debug3("bluegene.conf: %c%c%cx%c%c%c",
+ debug3("adding bps: %c%c%cx%c%c%c",
alpha_num[start[X]], alpha_num[start[Y]], alpha_num[start[Z]],
alpha_num[end[X]], alpha_num[end[Y]], alpha_num[end[Z]]);
debug3("slurm.conf: %c%c%c",
@@ -2380,6 +2429,7 @@ static int _addto_node_list(bg_record_t *bg_record, int *start, int *end)
slurm_conf_unlock();
ba_node = ba_copy_node(
&ba_system_ptr->grid[x][y][z]);
+ ba_node->used = 1;
list_append(bg_record->bg_block_list, ba_node);
node_count++;
}
@@ -2436,12 +2486,6 @@ static void _set_bg_lists()
list_destroy(bg_list);
bg_list = list_create(destroy_bg_record);
- slurm_mutex_lock(&request_list_mutex);
- if(bg_request_list)
- list_destroy(bg_request_list);
- bg_request_list = list_create(delete_ba_request);
- slurm_mutex_unlock(&request_list_mutex);
-
slurm_mutex_unlock(&block_state_mutex);
if(bg_blrtsimage_list)
@@ -2865,6 +2909,14 @@ static int _breakup_blocks(ba_request_t *request, List my_block_list)
continue;
if(bg_record->state != RM_PARTITION_FREE)
continue;
+ if (request->avail_node_bitmap &&
+ !bit_super_set(bg_record->bitmap,
+ request->avail_node_bitmap)) {
+ debug2("bg block %s has nodes not usable by this job",
+ bg_record->bg_block_id);
+ continue;
+ }
+
if(request->start_req) {
if ((request->start[X] != bg_record->start[X])
|| (request->start[Y] != bg_record->start[Y])
@@ -2889,7 +2941,7 @@ static int _breakup_blocks(ba_request_t *request, List my_block_list)
proc_cnt = bg_record->bp_count *
bg_record->cpus_per_bp;
if(proc_cnt == request->procs) {
- debug2("found it here %s, %s",
+ debug("found it here %s, %s",
bg_record->bg_block_id,
bg_record->nodes);
request->save_name = xmalloc(4);
@@ -2961,6 +3013,14 @@ static int _breakup_blocks(ba_request_t *request, List my_block_list)
!= NULL) {
if(bg_record->job_running != NO_JOB_RUNNING)
continue;
+ if (request->avail_node_bitmap &&
+ !bit_super_set(bg_record->bitmap,
+ request->avail_node_bitmap)) {
+ debug2("bg block %s has nodes not usable by this job",
+ bg_record->bg_block_id);
+ continue;
+ }
+
if(request->start_req) {
if ((request->start[X] != bg_record->start[X])
|| (request->start[Y] != bg_record->start[Y])
diff --git a/src/plugins/select/bluegene/plugin/select_bluegene.c b/src/plugins/select/bluegene/plugin/select_bluegene.c
index 48752fc999b5a46c19328d2774e648061c9c459d..085543cd20292be2674b9db38fdee1d087829d0d 100644
--- a/src/plugins/select/bluegene/plugin/select_bluegene.c
+++ b/src/plugins/select/bluegene/plugin/select_bluegene.c
@@ -1,7 +1,7 @@
/*****************************************************************************\
* select_bluegene.c - node selection plugin for Blue Gene system.
*
- * $Id: select_bluegene.c 12409 2007-09-26 16:32:16Z jette $
+ * $Id: select_bluegene.c 13271 2008-02-14 20:02:00Z da $
*****************************************************************************
* Copyright (C) 2004-2006 The Regents of the University of California.
* Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER).
@@ -1085,8 +1085,6 @@ extern int select_p_update_node_state (int index, uint16_t state)
}
}
#endif
- if(bluegene_layout_mode == LAYOUT_DYNAMIC)
- remove_from_request_list();
return SLURM_ERROR;
}
diff --git a/src/salloc/opt.c b/src/salloc/opt.c
index 80b5435bcbf52f36d907a8906961f208a9c70672..6833d240862f86003e8ed20b9e6f90318785ce39 100644
--- a/src/salloc/opt.c
+++ b/src/salloc/opt.c
@@ -669,6 +669,7 @@ void set_options(const int argc, char **argv)
{"no-kill", no_argument, 0, 'k'},
{"kill-command", optional_argument, 0, 'K'},
{"tasks", required_argument, 0, 'n'},
+ {"ntasks", required_argument, 0, 'n'},
{"nodes", required_argument, 0, 'N'},
{"overcommit", no_argument, 0, 'O'},
{"partition", required_argument, 0, 'p'},
@@ -827,6 +828,9 @@ void set_options(const int argc, char **argv)
#ifdef HAVE_BG
info("\tThe nodelist option should only be used if\n"
"\tthe block you are asking for can be created.\n"
+ "\tIt should also include all the midplanes you\n"
+ "\twant to use, partial lists may not\n"
+ "\twork correctly.\n"
"\tPlease consult smap before using this option\n"
"\tor your job may be stuck with no way to run.");
#endif
diff --git a/src/sbatch/opt.c b/src/sbatch/opt.c
index 82633fa422d69808094505e26d9f5a7e36fc6c18..d94a2466156301a0cb6b0f1fe98e530908299240 100644
--- a/src/sbatch/opt.c
+++ b/src/sbatch/opt.c
@@ -622,6 +622,7 @@ static struct option long_options[] = {
{"immediate", no_argument, 0, 'I'},
{"job-name", required_argument, 0, 'J'},
{"no-kill", no_argument, 0, 'k'},
+ {"tasks", required_argument, 0, 'n'},
{"ntasks", required_argument, 0, 'n'},
{"nodes", required_argument, 0, 'N'},
{"output", required_argument, 0, 'o'},
@@ -1154,6 +1155,9 @@ static void _set_options(int argc, char **argv)
#ifdef HAVE_BG
info("\tThe nodelist option should only be used if\n"
"\tthe block you are asking for can be created.\n"
+ "\tIt should also include all the midplanes you\n"
+ "\twant to use, partial lists may not\n"
+ "\twork correctly.\n"
"\tPlease consult smap before using this option\n"
"\tor your job may be stuck with no way to run.");
#endif
diff --git a/src/sbatch/sbatch.c b/src/sbatch/sbatch.c
index 2fd84f78dd8eaaae998df5162709fc14ef6be174..9ad0b2a77f610ebeb0a1dd77390b1db6c7fe15a9 100644
--- a/src/sbatch/sbatch.c
+++ b/src/sbatch/sbatch.c
@@ -1,7 +1,7 @@
/*****************************************************************************\
* sbatch.c - Submit a SLURM batch script.
*
- * $Id: sbatch.c 12856 2007-12-19 00:18:44Z jette $
+ * $Id: sbatch.c 13231 2008-02-08 17:16:47Z jette $
*****************************************************************************
* Copyright (C) 2006 The Regents of the University of California.
* Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER).
@@ -216,11 +216,12 @@ static int fill_job_desc_from_opts(job_desc_msg_t *desc)
struct passwd *pw = NULL;
pw = getpwuid(opt.uid);
if (pw != NULL) {
- desc->environment = env_array_user_default(pw->pw_name,
+ desc->environment = env_array_user_default(
+ pw->pw_name,
opt.get_user_env_time,
opt.get_user_env_mode);
- /* FIXME - should we abort if j->environment
- * is NULL? */
+ if (desc->environment == NULL)
+ exit(1); /* error already logged */
}
}
env_array_merge(&desc->environment, (const char **)environ);
diff --git a/src/slaunch/opt.c b/src/slaunch/opt.c
index 83658f41f5660ed313b3448f930820d696da8376..de1b25d64ee66f54fa7f4838ce1edee2584b6a35 100644
--- a/src/slaunch/opt.c
+++ b/src/slaunch/opt.c
@@ -1145,6 +1145,9 @@ void set_options(const int argc, char **argv)
#ifdef HAVE_BG
info("\tThe nodelist option should only be used if\n"
"\tthe block you are asking for can be created.\n"
+ "\tIt should also include all the midplanes you\n"
+ "\twant to use, partial lists may not\n"
+ "\twork correctly.\n"
"\tPlease consult smap before using this option\n"
"\tor your job may be stuck with no way to run.");
#endif
diff --git a/src/slurmctld/controller.c b/src/slurmctld/controller.c
index 99769cf93c59635273668751667d40109ea30989..58d26e45f88dc9e69668d8d91d546c460679548c 100644
--- a/src/slurmctld/controller.c
+++ b/src/slurmctld/controller.c
@@ -1,6 +1,6 @@
/*****************************************************************************\
* controller.c - main control machine daemon for slurm
- * $Id: controller.c 13077 2008-01-23 22:31:44Z da $
+ * $Id: controller.c 13156 2008-02-01 17:43:01Z da $
*****************************************************************************
* Copyright (C) 2002-2006 The Regents of the University of California.
* Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER).
@@ -842,8 +842,8 @@ static int _gold_mark_all_nodes_down(char *reason, time_t event_time)
state_file = xstrdup (slurmctld_conf.state_save_location);
xstrcat (state_file, "/node_state");
if (stat(state_file, &stat_buf)) {
- error("_gold_mark_all_nodes_down: could not stat(%s) to record "
- "node down time", state_file);
+ debug("_gold_mark_all_nodes_down: could not stat(%s) "
+ "to record node down time", state_file);
xfree(state_file);
return rc;
}
diff --git a/src/slurmctld/job_mgr.c b/src/slurmctld/job_mgr.c
index 5109f4cf6df2bba5ee8859cf7a70bc956f5d30c7..923670c928a352f0fc9227081fbe00bb4aeceef8 100644
--- a/src/slurmctld/job_mgr.c
+++ b/src/slurmctld/job_mgr.c
@@ -3,7 +3,7 @@
* Note: there is a global job list (job_list), time stamp
* (last_job_update), and hash table (job_hash)
*
- * $Id: job_mgr.c 13083 2008-01-24 16:28:23Z jette $
+ * $Id: job_mgr.c 13176 2008-02-04 16:56:57Z jette $
*****************************************************************************
* Copyright (C) 2002-2007 The Regents of the University of California.
* Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER).
@@ -1064,6 +1064,7 @@ extern int kill_running_job_by_node_name(char *node_name, bool step_test)
} else if (job_ptr->batch_flag && job_ptr->details &&
(job_ptr->details->no_requeue == 0)) {
uint16_t save_state;
+ srun_node_fail(job_ptr->job_id, node_name);
info("requeue job %u due to failure of node %s",
job_ptr->job_id, node_name);
_set_job_prio(job_ptr);
@@ -2741,7 +2742,7 @@ static int _validate_job_desc(job_desc_msg_t * job_desc_msg, int allocate,
if (job_desc_msg->job_max_memory == NO_VAL)
job_desc_msg->job_max_memory = 1; /* default 1MB mem per node */
if (job_desc_msg->job_min_tmp_disk == NO_VAL)
- job_desc_msg->job_min_tmp_disk = 1;/* default 1MB disk per node */
+ job_desc_msg->job_min_tmp_disk = 0;/* default 0MB disk per node */
return SLURM_SUCCESS;
}
diff --git a/src/slurmctld/node_mgr.c b/src/slurmctld/node_mgr.c
index 7478bc6062aeed2917abe713367a82316bbad159..5f07cff6f16cfa4e0dd227d7faf6e527da0f0abb 100644
--- a/src/slurmctld/node_mgr.c
+++ b/src/slurmctld/node_mgr.c
@@ -4,7 +4,7 @@
* hash table (node_hash_table), time stamp (last_node_update) and
* configuration list (config_list)
*
- * $Id: node_mgr.c 13068 2008-01-23 18:41:54Z jette $
+ * $Id: node_mgr.c 13274 2008-02-14 21:32:07Z jette $
*****************************************************************************
* Copyright (C) 2002-2006 The Regents of the University of California.
* Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER).
@@ -1919,7 +1919,8 @@ void node_not_resp (char *name, time_t msg_time)
error ("node_not_resp unable to find node %s", name);
return;
}
- error("Node %s not responding", node_ptr->name);
+ if ((node_ptr->node_state & NODE_STATE_BASE) != NODE_STATE_DOWN)
+ error("Node %s not responding", node_ptr->name);
_node_not_resp(node_ptr, msg_time);
#endif
}
diff --git a/src/slurmctld/node_scheduler.c b/src/slurmctld/node_scheduler.c
index 16f7661360d4da19c34a06d64368893e596f2d99..b8903bc6280316183b046cfdf839d9fa78d8859a 100644
--- a/src/slurmctld/node_scheduler.c
+++ b/src/slurmctld/node_scheduler.c
@@ -2,7 +2,7 @@
* node_scheduler.c - select and allocated nodes to jobs
* Note: there is a global node table (node_record_table_ptr)
*
- * $Id: node_scheduler.c 13061 2008-01-22 21:23:56Z da $
+ * $Id: node_scheduler.c 13234 2008-02-08 22:13:39Z jette $
*****************************************************************************
* Copyright (C) 2002-2006 The Regents of the University of California.
* Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER).
@@ -300,18 +300,26 @@ _pick_best_load(struct job_record *job_ptr, bitstr_t * bitmap,
uint32_t req_nodes, bool test_only)
{
bitstr_t *basemap;
- int i, error_code = EINVAL, node_cnt = 0, prev_cnt = 0, set_cnt;
-
- basemap = bit_copy(bitmap);
- if (basemap == NULL)
- fatal("bit_copy malloc failure");
+ int i, max_bit, error_code = EINVAL;
+ int node_cnt = 0, prev_cnt = 0, set_cnt;
set_cnt = bit_set_count(bitmap);
if ((set_cnt < min_nodes) ||
((req_nodes > min_nodes) && (set_cnt < req_nodes)))
return error_code; /* not usable */
+ basemap = bit_copy(bitmap);
+ if (basemap == NULL)
+ fatal("bit_copy malloc failure");
+
+ max_bit = bit_size(bitmap) - 1;
for (i=0; node_cnt<set_cnt; i++) {
+ /* if req_nodes, then start with those as a baseline */
+ if (job_ptr->details && job_ptr->details->req_node_bitmap) {
+ bit_copybits(bitmap, job_ptr->details->req_node_bitmap);
+ } else {
+ bit_nclear(bitmap, 0, max_bit);
+ }
node_cnt = _job_count_bitmap(basemap, bitmap, i);
if ((node_cnt == 0) || (node_cnt == prev_cnt))
continue; /* nothing new to test */
@@ -332,7 +340,7 @@ _pick_best_load(struct job_record *job_ptr, bitstr_t * bitmap,
/*
* Set the bits in 'jobmap' that correspond to bits in the 'bitmap'
- * that are running 'job_cnt' jobs or less, and clear the rest.
+ * that are running 'job_cnt' jobs or less.
*/
static int
_job_count_bitmap(bitstr_t * bitmap, bitstr_t * jobmap, int job_cnt)
@@ -345,8 +353,6 @@ _job_count_bitmap(bitstr_t * bitmap, bitstr_t * jobmap, int job_cnt)
(node_record_table_ptr[i].run_job_cnt <= job_cnt)) {
bit_set(jobmap, i);
count++;
- } else {
- bit_clear(jobmap, i);
}
}
return count;
diff --git a/src/slurmctld/partition_mgr.c b/src/slurmctld/partition_mgr.c
index 62ab07a80b6ff2323ad1bf63397e78ffe391f5be..4d31c895af885f1945209136ace689ca2f3c10f2 100644
--- a/src/slurmctld/partition_mgr.c
+++ b/src/slurmctld/partition_mgr.c
@@ -2,7 +2,7 @@
* partition_mgr.c - manage the partition information of slurm
* Note: there is a global partition list (part_list) and
* time stamp (last_part_update)
- * $Id: partition_mgr.c 12452 2007-10-05 19:07:07Z da $
+ * $Id: partition_mgr.c 13279 2008-02-15 00:14:16Z jette $
*****************************************************************************
* Copyright (C) 2002-2007 The Regents of the University of California.
* Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER).
@@ -835,12 +835,14 @@ int update_part(update_part_msg_t * part_desc)
xfree(part_ptr->allow_uids);
if ((strcasecmp(part_desc->allow_groups, "ALL") == 0) ||
(part_desc->allow_groups[0] == '\0')) {
- info("update_part: setting allow_groups to ALL for partition %s",
+ info("update_part: setting allow_groups to ALL for "
+ "partition %s",
part_desc->name);
} else {
part_ptr->allow_groups = part_desc->allow_groups;
part_desc->allow_groups = NULL;
- info("update_part: setting allow_groups to %s for partition %s",
+ info("update_part: setting allow_groups to %s for "
+ "partition %s",
part_ptr->allow_groups, part_desc->name);
part_ptr->allow_uids =
_get_groups_members(part_ptr->allow_groups);
@@ -852,8 +854,14 @@ int update_part(update_part_msg_t * part_desc)
if (part_desc->nodes[0] == '\0')
part_ptr->nodes = NULL; /* avoid empty string */
- else
+ else {
+ int i;
part_ptr->nodes = xstrdup(part_desc->nodes);
+ for (i=0; part_ptr->nodes[i]; i++) {
+ if (isspace(part_ptr->nodes[i]))
+ part_ptr->nodes[i] = ',';
+ }
+ }
error_code = _build_part_bitmap(part_ptr);
if (error_code) {
@@ -861,7 +869,7 @@ int update_part(update_part_msg_t * part_desc)
part_ptr->nodes = backup_node_list;
} else {
info("update_part: setting nodes to %s for partition %s",
- part_desc->nodes, part_desc->name);
+ part_ptr->nodes, part_desc->name);
xfree(backup_node_list);
}
}
diff --git a/src/slurmctld/proc_req.c b/src/slurmctld/proc_req.c
index f5768a3fb9d687662d8013cd58777262aea2727f..e222d1e86bc1702bdd6d0ce0e700b08da3ec9389 100644
--- a/src/slurmctld/proc_req.c
+++ b/src/slurmctld/proc_req.c
@@ -1,7 +1,7 @@
/*****************************************************************************\
* proc_req.c - process incomming messages to slurmctld
*
- * $Id: proc_req.c 12413 2007-09-26 17:32:47Z jette $
+ * $Id: proc_req.c 13237 2008-02-08 23:16:16Z jette $
*****************************************************************************
* Copyright (C) 2002-2007 The Regents of the University of California.
* Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER).
@@ -324,6 +324,7 @@ void _fill_ctld_conf(slurm_ctl_conf_t * conf_ptr)
conf_ptr->job_credential_public_certificate = xstrdup(conf->
job_credential_public_certificate);
conf_ptr->job_file_append = conf->job_file_append;
+ conf_ptr->get_env_timeout = conf->get_env_timeout;
conf_ptr->kill_wait = conf->kill_wait;
conf_ptr->mail_prog = xstrdup(conf->mail_prog);
conf_ptr->max_job_cnt = conf->max_job_cnt;
diff --git a/src/slurmctld/step_mgr.c b/src/slurmctld/step_mgr.c
index 43aa77897f19ea43a3e85d9bb5c09bafee175639..2da3cfdc5c5efa1e9d5e9392ffe2d2d573382f0c 100644
--- a/src/slurmctld/step_mgr.c
+++ b/src/slurmctld/step_mgr.c
@@ -1,6 +1,6 @@
/*****************************************************************************\
* step_mgr.c - manage the job step information of slurm
- * $Id: step_mgr.c 12681 2007-11-26 18:56:25Z jette $
+ * $Id: step_mgr.c 13155 2008-02-01 17:30:43Z jette $
*****************************************************************************
* Copyright (C) 2002-2006 The Regents of the University of California.
* Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER).
@@ -701,8 +701,16 @@ step_create(job_step_create_request_msg_t *step_specs,
if (job_ptr == NULL)
return ESLURM_INVALID_JOB_ID ;
- if ((job_ptr->job_state == JOB_SUSPENDED) || IS_JOB_PENDING(job_ptr))
+ if (job_ptr->job_state == JOB_SUSPENDED)
return ESLURM_DISABLED;
+ if (IS_JOB_PENDING(job_ptr)) {
+ /* NOTE: LSF creates a job allocation for batch jobs.
+ * After the allocation has been made, LSF submits a
+ * job to run in that allocation (sbatch --jobid= ...).
+ * If that job is pending either LSF messed up or LSF is
+ * not being used. We have seen this problem with Moab. */
+ return ESLURM_DUPLICATE_JOB_ID;
+ }
if (batch_step) {
info("user %u attempting to run batch script within "
diff --git a/src/slurmd/slurmstepd/mgr.c b/src/slurmd/slurmstepd/mgr.c
index 0091fadc28f0b3c48fb7fa12437a9a1170f4aadf..6ac03c22d6365fe23e0c382eb9cf766db941dd7e 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 12647 2007-11-12 17:09:47Z da $
+ * $Id: mgr.c 13229 2008-02-08 01:02:06Z jette $
*****************************************************************************
* Copyright (C) 2002-2007 The Regents of the University of California.
* Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER).
@@ -280,7 +280,7 @@ mgr_launch_batch_job_setup(batch_job_launch_msg_t *msg, slurm_addr *cli)
}
/* this is the new way of setting environment variables */
- env_array_for_batch_job(&job->env, msg);
+ env_array_for_batch_job(&job->env, msg, conf->node_name);
/* this is the old way of setting environment variables */
job->envtp->nprocs = msg->nprocs;
diff --git a/src/smap/configure_functions.c b/src/smap/configure_functions.c
index fdbb327e2b0362d00be71766366c0fe423605beb..aa5294f7138600b687b946b93ff77f240bd1e128 100644
--- a/src/smap/configure_functions.c
+++ b/src/smap/configure_functions.c
@@ -1,6 +1,6 @@
/*****************************************************************************\
* configure_functions.c - Functions related to configure mode of smap.
- * $Id: configure_functions.c 11985 2007-08-09 23:07:08Z da $
+ * $Id: configure_functions.c 13270 2008-02-14 19:40:44Z da $
*****************************************************************************
* Copyright (C) 2002 The Regents of the University of California.
* Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER).
@@ -213,7 +213,8 @@ static int _create_allocation(char *com, List allocated_blocks)
request->nodecards = 0;
request->quarters = 0;
request->passthrough = false;
-
+ request->avail_node_bitmap = NULL;
+
while(i<len) {
if(!strncasecmp(com+i, "mesh", 4)) {
request->conn_type=SELECT_MESH;
@@ -855,7 +856,8 @@ static int _copy_allocation(char *com, List allocated_blocks)
request->rotate_count= 0;
request->elongate_count = 0;
request->elongate_geos = list_create(NULL);
-
+ request->avail_node_bitmap = NULL;
+
results_i = list_iterator_create(request->elongate_geos);
while ((geo_ptr = list_next(results_i)) != NULL) {
geo = xmalloc(sizeof(int)*3);
@@ -1098,6 +1100,7 @@ static int _add_bg_record(blockreq_t *blockreq, List allocated_blocks)
bp_count++;
if(nodes[j] != ',')
break;
+ j--;
}
j++;
}
diff --git a/src/smap/partition_functions.c b/src/smap/partition_functions.c
index 2dd270b00740fe0bfd3aece334267731ad25059b..9e8c6743b9a86961221d96d6070903ae83ac0614 100644
--- a/src/smap/partition_functions.c
+++ b/src/smap/partition_functions.c
@@ -425,6 +425,7 @@ static int _marknodes(db2_block_info_t *block_ptr, int count)
0);
if(block_ptr->nodes[j] != ',')
break;
+ j--;
}
j++;
}
@@ -914,6 +915,7 @@ static int _make_nodelist(char *nodes, List nodelist)
_addto_nodelist(nodelist, start, start);
if(nodes[j] != ',')
break;
+ j--;
}
j++;
}
diff --git a/src/srun/allocate.c b/src/srun/allocate.c
index a06ea95966d6af57affd4ec8dfbbc3c5fa0b218f..b1abd672acf29f57b0179aaf9610c21a59464eab 100644
--- a/src/srun/allocate.c
+++ b/src/srun/allocate.c
@@ -1,6 +1,6 @@
/*****************************************************************************\
* src/srun/allocate.c - srun functions for managing node allocations
- * $Id: allocate.c 12700 2007-11-27 23:39:24Z jette $
+ * $Id: allocate.c 13231 2008-02-08 17:16:47Z jette $
*****************************************************************************
* Copyright (C) 2002-2006 The Regents of the University of California.
* Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER).
@@ -561,12 +561,12 @@ job_desc_msg_create_from_opts (char *script)
struct passwd *pw = NULL;
pw = getpwuid(opt.uid);
if (pw != NULL) {
- j->environment =
- env_array_user_default(pw->pw_name,
+ j->environment = env_array_user_default(
+ pw->pw_name,
opt.get_user_env_time,
opt.get_user_env_mode);
- /* FIXME - should we abort if j->environment
- is NULL? */
+ if (j->environment == NULL)
+ exit(1); /* error already logged */
}
}
env_array_merge(&j->environment, (const char **)environ);
diff --git a/src/srun/opt.c b/src/srun/opt.c
index fca5cf8ead2e5b4f7fee62b664205e17ab79d4db..4802c930b46fedf199a374d5f2ec36bb64079b7f 100644
--- a/src/srun/opt.c
+++ b/src/srun/opt.c
@@ -1,6 +1,6 @@
/*****************************************************************************\
* opt.c - options processing for srun
- * $Id: opt.c 12856 2007-12-19 00:18:44Z jette $
+ * $Id: opt.c 13270 2008-02-14 19:40:44Z da $
*****************************************************************************
* Copyright (C) 2002-2006 The Regents of the University of California.
* Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER).
@@ -1721,6 +1721,9 @@ void set_options(const int argc, char **argv, int first)
#ifdef HAVE_BG
info("\tThe nodelist option should only be used if\n"
"\tthe block you are asking for can be created.\n"
+ "\tIt should also include all the midplanes you\n"
+ "\twant to use, partial lists may not\n"
+ "\twork correctly.\n"
"\tPlease consult smap before using this option\n"
"\tor your job may be stuck with no way to run.");
#endif
@@ -2168,7 +2171,7 @@ static void _opt_args(int argc, char **argv)
char *fullpath;
char *cmd = remote_argv[0];
bool search_cwd = (opt.batch || opt.allocate);
- int mode = (search_cwd) ? R_OK : R_OK | X_OK;
+ int mode = (search_cwd) ? R_OK : X_OK;
if ((fullpath = _search_path(cmd, search_cwd, mode))) {
xfree(remote_argv[0]);
diff --git a/src/sview/part_info.c b/src/sview/part_info.c
index 48c274ea1ea0330d8c7a3d84df225f31f8202a14..bf53ddb10457eaa20b2e1f98c3d130f553e8be5f 100644
--- a/src/sview/part_info.c
+++ b/src/sview/part_info.c
@@ -1390,6 +1390,7 @@ static List _create_part_info_list(partition_info_msg_t *part_info_ptr,
ListIterator itr = NULL;
hostlist_t hl;
#ifdef HAVE_BG
+ int j;
bg_info_record_t *bg_info_record = NULL;
int node_scaling = part_info_ptr->partition_array[0].node_scaling;
char *slurm_user = NULL;
@@ -1451,9 +1452,9 @@ static List _create_part_info_list(partition_info_msg_t *part_info_ptr,
node_ptr = _find_node(node_name, node_info_ptr);
free(node_name);
#ifdef HAVE_BG
- for(i=0; i<3; i++) {
+ for(j=0; j<3; j++) {
int norm = 0;
- switch(i) {
+ switch(j) {
case SVIEW_BG_IDLE_STATE:
/* get the idle node count if
* we don't have any error or