From c0fbb5b43d31253bb5cde8ea6aac3b2548610278 Mon Sep 17 00:00:00 2001
From: Moe Jette <jette1@llnl.gov>
Date: Thu, 19 Oct 2006 22:40:00 +0000
Subject: [PATCH] svn merge -r9843:9854
https://eris.llnl.gov/svn/slurm/branches/slurm-1.1
---
NEWS | 6 ++++++
doc/html/slurm.shtml | 15 ++++++++-------
src/common/plugin.c | 16 +++++++++-------
src/plugins/sched/wiki2/msg.c | 3 ++-
.../select/bluegene/plugin/libsched_if64.c | 13 +++++++++++--
.../select/bluegene/plugin/slurm_epilog.c | 2 +-
.../select/bluegene/plugin/slurm_prolog.c | 10 ++++++----
7 files changed, 43 insertions(+), 22 deletions(-)
diff --git a/NEWS b/NEWS
index 259ce99919b..6103472fa80 100644
--- a/NEWS
+++ b/NEWS
@@ -82,6 +82,12 @@ documents those changes that are of interest to users and admins.
* Changes in SLURM 1.1.16
=========================
+ - BLUEGENE - fix to make prolog run 5 minutes longer to make sure we have
+ enough time to free the overlapping blocks when starting a new job on a
+ block.
+ - BLUEGENE - edit to the libsched_if.so to read env and look at
+ MPIRUN_PARTITION to see if we are in slurm or running mpirun natively.
+ - Plugins are now dlopened RTLD_LAZY instead of RTLD_NOW.
* Changes in SLURM 1.1.15
=========================
diff --git a/doc/html/slurm.shtml b/doc/html/slurm.shtml
index 3426f75f894..61a422bc219 100644
--- a/doc/html/slurm.shtml
+++ b/doc/html/slurm.shtml
@@ -33,12 +33,13 @@ add functionality.</li>
<p>SLURM provides resource management on about 1000 computers world-wide including
many of the most powerful computers in the world including:
<ul>
-<li><a href="http://www.llnl.gov/asci/platforms/bluegenel/">BlueGene/L</a> with 65,536
-dual-processor compute nodes</li>
-<li><a href="http://www.llnl.gov/linux/thunder/">Thunder</a> a Linux cluster with 1024
-nodes, each having four Itanium2 processors</li>
-<li><a href="http://www.llnl.gov/asci/platforms/purple/">ASC Purple</a> an IBM SP/AIX
-cluster with 1500 nodes, each having eight Power5 processors</li>
+<li><a href="http://www.llnl.gov/asc/computing_resources/bluegenel/bluegene_home.html">BlueGene/L</a>
+with 65,536 dual-processor compute nodes</li>
+<li><a href="http://www.llnl.gov/asc/computing_resources/purple/purple_index.html">ASC Purple</a>
+an IBM SP/AIX cluster with 1532 nodes each having eight Power5 processors</li>
+<li>Peloton
+with 1152 nodes each having four sockets with dual core Opteron processors and an
+InfiniBand switch</li>
</ul>
There are about 150 downloads of SLURM per month from LLNL's FTP server alone.
As of September 2006, SLURM has been downloaded over 3500 times to over 500
@@ -46,6 +47,6 @@ distinct sites in 38 countries.
SLURM is also distributed and supported by <a href="http://www.hp.com">
Hewlett-Packard</a> as the resource manager in their XC System Software.</p>
-<p style="text-align:center;">Last modified 29 September 2006</p>
+<p style="text-align:center;">Last modified 19 October 2006</p>
<!--#include virtual="footer.txt"-->
diff --git a/src/common/plugin.c b/src/common/plugin.c
index 19409865377..dc7c6b02e0a 100644
--- a/src/common/plugin.c
+++ b/src/common/plugin.c
@@ -109,14 +109,16 @@ plugin_load_from_file( const char *fq_path )
int (*init)( void );
/*
- * Try to open the shared object. We have a choice of trying to
- * resolve all the symbols (in both directions) now or when the
- * symbols are first dereferenced and used. While it's slower to
- * do it this way, it's a lot easier to debug. If you get an
- * error somewhere down the line, you're likely to think it's
- * some condition that happened then instead of way back here.
+ * Try to open the shared object.
+ *
+ * Use RTLD_LAZY to allow plugins to use symbols that may be
+ * defined in only one slurm entity (e.g. srun and not slurmd),
+ * when the use of that symbol is restricted to within the
+ * entity from which it is available. (i.e. srun symbols are only
+ * used in the context of srun, not slurmd.)
+ *
*/
- plug = dlopen( fq_path, RTLD_NOW );
+ plug = dlopen( fq_path, RTLD_LAZY );
if ( plug == NULL ) {
error( "plugin_load_from_file: dlopen(%s): %s",
fq_path,
diff --git a/src/plugins/sched/wiki2/msg.c b/src/plugins/sched/wiki2/msg.c
index 1ea7d91aeec..40916cffa53 100644
--- a/src/plugins/sched/wiki2/msg.c
+++ b/src/plugins/sched/wiki2/msg.c
@@ -389,7 +389,8 @@ static int _parse_msg(char *msg, char **req)
if (delta_t > 300) {
err_code = -350;
err_msg = "TS value too far from NOW";
- error("wiki: TS delta_t=%u", delta_t);
+ error("wiki: TimeStamp too far from NOW (%u secs)",
+ delta_t);
return -1;
}
diff --git a/src/plugins/select/bluegene/plugin/libsched_if64.c b/src/plugins/select/bluegene/plugin/libsched_if64.c
index b64740eb288..37c24e5399d 100644
--- a/src/plugins/select/bluegene/plugin/libsched_if64.c
+++ b/src/plugins/select/bluegene/plugin/libsched_if64.c
@@ -37,11 +37,20 @@
* 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
\*****************************************************************************/
#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
int get_parameters(void *params)
{
- printf("YOU ARE OUTSIDE OF SLURM!!!! NOT RUNNING MPIRUN!\n");
- return -1;
+ char *partition = getenv("MPIRUN_PARTITION"); /* get MPIRUN env
+ * var to see if we
+ * are inside slurm
+ * or not */
+ if (!partition || (strlen(partition) < 3)) {
+ printf("YOU ARE OUTSIDE OF SLURM!!!! NOT RUNNING MPIRUN!\n");
+ return 1;
+ }
+ return 2;
}
void mpirun_done(int res)
diff --git a/src/plugins/select/bluegene/plugin/slurm_epilog.c b/src/plugins/select/bluegene/plugin/slurm_epilog.c
index 3e9b168ddb9..b5f5bef4b3b 100644
--- a/src/plugins/select/bluegene/plugin/slurm_epilog.c
+++ b/src/plugins/select/bluegene/plugin/slurm_epilog.c
@@ -19,7 +19,7 @@
* 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
+ * 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
diff --git a/src/plugins/select/bluegene/plugin/slurm_prolog.c b/src/plugins/select/bluegene/plugin/slurm_prolog.c
index 95c932f236c..25ef0fa4e6f 100644
--- a/src/plugins/select/bluegene/plugin/slurm_prolog.c
+++ b/src/plugins/select/bluegene/plugin/slurm_prolog.c
@@ -19,7 +19,7 @@
* 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
+ * 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
@@ -61,17 +61,18 @@
/*
* Check the bgblock's status every POLL_SLEEP seconds.
- * Retry for a period of MIN_DELAY +
+ * Retry for a period of MIN_FREE_PERVIOUS_BLOCK_DELAY + MIN_DELAY +
* (INCR_DELAY * POLL_SLEEP * base partition count).
* For example if MIN_DELAY=300 and INCR_DELAY=20 and POLL_SLEEP=3,
* wait up to 1260 seconds.
* For a 16 base partition bgblock to be ready (300 + (20 * 3 * 16).
*/
#define POLL_SLEEP 3 /* retry interval in seconds */
+#define MIN_FREE_PERVIOUS_BLOCK_DELAY 300 /* time in seconds */
#define MIN_DELAY 300 /* time in seconds */
#define INCR_DELAY 20 /* time in seconds per BP */
-int max_delay = MIN_DELAY;
+int max_delay = MIN_DELAY + MIN_FREE_PERVIOUS_BLOCK_DELAY;
int cur_delay = 0;
enum rm_partition_state {RM_PARTITION_FREE,
@@ -114,7 +115,8 @@ static int _wait_part_ready(uint32_t job_id)
{
int is_ready = 0, i, rc;
- max_delay = MIN_DELAY + (INCR_DELAY * _get_job_size(job_id));
+ max_delay = MIN_DELAY + MIN_FREE_PERVIOUS_BLOCK_DELAY +
+ (INCR_DELAY * _get_job_size(job_id));
#if _DEBUG
printf("Waiting for job %u to become ready.", job_id);
--
GitLab