From 1d799bb159f9cb3074a4f97d118536e71d67ba91 Mon Sep 17 00:00:00 2001
From: Moe Jette <jette1@llnl.gov>
Date: Fri, 24 Feb 2006 16:29:38 +0000
Subject: [PATCH] svn merge -r7220:7266
https://eris.llnl.gov/svn/slurm/branches/slurm-1.0
---
META | 4 +-
NEWS | 14 +++++
doc/html/bluegene.shtml | 12 ++--
doc/html/selectplugins.shtml | 7 ++-
doc/man/man1/sacct.1 | 2 +-
doc/man/man1/smap.1 | 8 +--
slurm.spec | 2 +-
slurm/slurm.h.in | 3 +-
src/common/node_select.c | 23 +++----
src/common/node_select.h | 5 +-
.../select/bluegene/plugin/bg_job_place.c | 63 ++++---------------
.../select/bluegene/plugin/bg_job_place.h | 5 +-
src/plugins/select/bluegene/plugin/bluegene.c | 2 +-
.../select/bluegene/plugin/select_bluegene.c | 5 +-
src/plugins/select/cons_res/select_cons_res.c | 4 +-
src/plugins/select/linear/select_linear.c | 4 +-
src/plugins/switch/federation/federation.c | 10 ++-
src/sacct/sacct.c | 2 +-
src/slurmctld/node_scheduler.c | 22 +++----
src/squeue/print.c | 2 +-
src/srun/srun.c | 4 +-
21 files changed, 90 insertions(+), 113 deletions(-)
diff --git a/META b/META
index 9b01fcf7f95..8eacdbb59a1 100644
--- a/META
+++ b/META
@@ -9,8 +9,8 @@
Name: slurm
Major: 1
Minor: 0
- Micro: 2
- Version: 1.0.2
+ Micro: 4
+ Version: 1.0.4
Release: 1
API_CURRENT: 9
API_AGE: 4
diff --git a/NEWS b/NEWS
index 7d1196a614c..5fd5dc33b89 100644
--- a/NEWS
+++ b/NEWS
@@ -17,6 +17,18 @@ documents those changes that are of interest to users and admins.
-- Add hash tables for select/cons_res plugin (Susanne Balle, HP,
patch_02222006).
+* Changes in SLURM 1.0.5
+========================
+ -- Restructure logic for scheduling BlueGene small block jobs. Added
+ "test_only" flag to select_p_job_test() in select plugin.
+ -- Correct squeue "NODELIST" output for BlueGene small block jobs.
+
+* Changes in SLURM 1.0.4
+========================
+ -- Release job allocation if step creation fails (especially for BlueGene).
+ -- Fix bug select/bluegene warm start with changed bglblock layout.
+ -- Fix bug for queuing full-system BlueGene jobs.
+
* Changes in SLURM 1.0.3
========================
-- Fix bug that could refuse to queue batch jobs for BlueGene system.
@@ -30,6 +42,8 @@ documents those changes that are of interest to users and admins.
SLURM_JOBID environment variable is invalid.
-- Federation driver: allow selection of a sepecific switch interface
(sni0, sni1, etc.) with -euidevice/MP_EUIDEVICE.
+ -- Return an error for "scontrol reconfig" if there is already one in
+ progress
* Changes in SLURM 1.0.2
========================
diff --git a/doc/html/bluegene.shtml b/doc/html/bluegene.shtml
index 3ee61d8f526..c7a9a68d3ab 100644
--- a/doc/html/bluegene.shtml
+++ b/doc/html/bluegene.shtml
@@ -96,12 +96,12 @@ Values in the X dimension increase to the right.
Values in the Z dimension increase down and toward the left.</p>
<pre>
- a a a a b b d d ID JOBID PARTITION BGL_BLOCK USER NAME ST TIME NODES NODELIST
- a a a a b b d d a 12345 batch RMP0 joseph tst1 R 43:12 64 bg[000x333]
- a a a a b b c c b 12346 debug RMP1 chris sim3 R 12:34 16 bg[420x533]
-a a a a b b c c c 12350 debug RMP2 danny job3 R 0:12 8 bg[622x733]
- d 12356 debug RMP3 dan colu R 18:05 16 bg[600x731]
- a a a a b b d d e 12378 debug RMP4 joseph asx4 R 0:34 4 bg[612x713]
+ a a a a b b d d ID JOBID PARTITION BG_BLOCK USER NAME ST TIME NODES NODELIST
+ a a a a b b d d a 12345 batch RMP0 joseph tst1 R 43:12 64 bg[000x333]
+ a a a a b b c c b 12346 debug RMP1 chris sim3 R 12:34 16 bg[420x533]
+a a a a b b c c c 12350 debug RMP2 danny job3 R 0:12 8 bg[622x733]
+ d 12356 debug RMP3 dan colu R 18:05 16 bg[600x731]
+ a a a a b b d d e 12378 debug RMP4 joseph asx4 R 0:34 4 bg[612x713]
a a a a b b d d
a a a a b b c c
a a a a b b c c
diff --git a/doc/html/selectplugins.shtml b/doc/html/selectplugins.shtml
index b583a356c40..5b3e9d9b2c6 100644
--- a/doc/html/selectplugins.shtml
+++ b/doc/html/selectplugins.shtml
@@ -157,7 +157,7 @@ SLURM_NO_CHANGE_IN_DATA if data has not changed since last packed, otherwise SLU
<h3>Job-Specific Node Selection Functions</h3>
<p class="commandline">int select_p_job_test (struct job_record *job_ptr,
-bitstr_t *bitmap, int min_nodes, int max_nodes);</p>
+bitstr_t *bitmap, int min_nodes, int max_nodes, bool test_only);</p>
<p style="margin-left:.2in"><b>Description</b>: Given a job's scheduling requirement
specification and a set of nodes which might be used to satisfy the request, identify
the nodes which "best" satify the request. Note that nodes being considered for allocation
@@ -186,6 +186,9 @@ and partition specifications.<br>
<span class="commandline"> max_nodes</span> (input)
maximum number of nodes to allocate to this job. Note this reflects both job
and partition specifications.<br>
+<span class="commandline"> test_only</span> (input)
+if set then we only want to test our ability to run the job at some time,
+not necesarily now with currently available resources.<br>
</p>
<p style="margin-left:.2in"><b>Returns</b>: SLURM_SUCCESS if successful. On failure,
the plugin should return SLURM_ERROR and future attempts may be made to schedule
@@ -323,6 +326,6 @@ to maintain data format compatibility across different versions of the plugin.</
<p class="footer"><a href="#top">top</a></p>
-<p style="text-align:center;">Last modified 27 December 2005</p>
+<p style="text-align:center;">Last modified 23 February 2006</p>
<!--#include virtual="footer.txt"-->
diff --git a/doc/man/man1/sacct.1 b/doc/man/man1/sacct.1
index 61ec6bb0531..b862c0ec53c 100644
--- a/doc/man/man1/sacct.1
+++ b/doc/man/man1/sacct.1
@@ -353,7 +353,7 @@ The format of this field's output is as follows:
.RS
.PD "0"
.HP
-\f2[DD:[hh:]]mm:ss.nn\fP
+\f2[DD-[hh:]]mm:ss.nn\fP
.PD
.RE
.IP
diff --git a/doc/man/man1/smap.1 b/doc/man/man1/smap.1
index 6c1d19e7267..8c021422941 100644
--- a/doc/man/man1/smap.1
+++ b/doc/man/man1/smap.1
@@ -47,7 +47,7 @@ Displays information about jobs running on system.
Displays information about slurm partitions on the system
.TP
.I "b"
-Displays information about BGL partitions on the system
+Displays information about BG partitions on the system
.TP
.I "c"
Displays current node states and allows users to configure the system.
@@ -84,8 +84,8 @@ scroll the window containing the text information.
\fBAVAIL\fR
Partition state: \fBup\fR or \fBdown\fR.
.TP
-\fBBGL_BLOCK\fR
-BGL Block Name\fR.
+\fBBG_BLOCK\fR
+BlueGene Block Name\fR.
.TP
\fBCONN\fR
Connection Type: \fBTORUS\fR or \fBMESH\fR or \fBSMALL\fR (for small blocks).
@@ -187,7 +187,7 @@ blocks.
\fBOUTPUT\fR
.TP
-.I "BGL_BLOCK" BGL Block Name.
+.I "BG_BLOCK" BlueGene Block Name.
.TP
.I "CONN"
Connection Type: \fBTORUS\fR or \fBMESH\fR or \fBSMALL\fR (for small blocks).
diff --git a/slurm.spec b/slurm.spec
index 188bd7ab3b4..b45d0f61182 100644
--- a/slurm.spec
+++ b/slurm.spec
@@ -157,7 +157,7 @@ rm -f $RPM_BUILD_ROOT/%{_libdir}/slurm/*.{a,la}
LIST=./slurm.files
touch $LIST
if [ -d /etc/init.d ]; then
- echo "%config(noreplace) /etc/init.d/slurm" >> $LIST
+ echo "/etc/init.d/slurm" >> $LIST
fi
test -f $RPM_BUILD_ROOT/%{_libdir}/slurm/task_affinity.so &&
echo %{_libdir}/slurm/task_affinity.so >> $LIST
diff --git a/slurm/slurm.h.in b/slurm/slurm.h.in
index 6ab7fdb8e78..8cfa63e3eaf 100644
--- a/slurm/slurm.h.in
+++ b/slurm/slurm.h.in
@@ -193,8 +193,7 @@ enum select_data_type {
SELECT_DATA_NODE_USE, /* data-> uint16_t node_use */
SELECT_DATA_CONN_TYPE, /* data-> uint16_t connection_type */
SELECT_DATA_BLOCK_ID, /* data-> char bg_block_id */
- SELECT_DATA_QUARTER, /* data-> uint32_t quarter */
- SELECT_DATA_CHECKED /* data-> uint16_t checked */
+ SELECT_DATA_QUARTER /* data-> uint32_t quarter */
};
enum select_print_mode {
diff --git a/src/common/node_select.c b/src/common/node_select.c
index 046ab54c5db..b0811a6c839 100644
--- a/src/common/node_select.c
+++ b/src/common/node_select.c
@@ -10,7 +10,7 @@
*
* $Id$
*****************************************************************************
- * Copyright (C) 2002 The Regents of the University of California.
+ * Copyright (C) 2002-2006 The Regents of the University of California.
* Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER).
* Written by Morris Jette <jette1@llnl.gov>.
* UCRL-CODE-217948.
@@ -66,7 +66,7 @@ typedef struct slurm_select_ops {
int (*part_init) ( List part_list );
int (*job_test) ( struct job_record *job_ptr,
bitstr_t *bitmap, int min_nodes,
- int max_nodes );
+ int max_nodes, bool test_only );
int (*job_begin) ( struct job_record *job_ptr );
int (*job_ready) ( struct job_record *job_ptr );
int (*job_fini) ( struct job_record *job_ptr );
@@ -109,12 +109,8 @@ struct select_jobinfo {
uint16_t node_use; /* see enum node_use_type */
char *bg_block_id; /* Blue Gene partition ID */
uint16_t magic; /* magic number */
- int32_t quarter; /* for bg to tell which quarter of a small
+ int32_t quarter; /* for bg to tell which quarter of a small
partition the job is running */
- uint32_t checked; /* for bg to tell plugin it already
- checked and all partitions were full
- looking for best choice now */
-
};
#endif
@@ -407,16 +403,17 @@ extern int select_g_get_info_from_plugin (enum select_data_info cr_info, void *d
* IN/OUT bitmap - map of nodes being considered for allocation on input,
* map of nodes actually to be assigned on output
* IN min_nodes - minimum number of nodes to allocate to job
- * IN max_nodes - maximum number of nodes to allocate to job
+ * IN max_nodes - maximum number of nodes to allocate to job
+ * IN test_only - if true, only test if ever could run, not necessarily now
*/
extern int select_g_job_test(struct job_record *job_ptr, bitstr_t *bitmap,
- int min_nodes, int max_nodes)
+ int min_nodes, int max_nodes, bool test_only)
{
if (slurm_select_init() < 0)
return SLURM_ERROR;
return (*(g_select_context->ops.job_test))(job_ptr, bitmap,
- min_nodes, max_nodes);
+ min_nodes, max_nodes, test_only);
}
/*
@@ -566,9 +563,6 @@ extern int select_g_set_jobinfo (select_jobinfo_t jobinfo,
case SELECT_DATA_QUARTER:
jobinfo->quarter = *tmp_32;
break;
- case SELECT_DATA_CHECKED:
- jobinfo->checked = *tmp_16;
- break;
default:
debug("select_g_set_jobinfo data_type %d invalid",
data_type);
@@ -620,9 +614,6 @@ extern int select_g_get_jobinfo (select_jobinfo_t jobinfo,
case SELECT_DATA_QUARTER:
*tmp_32 = jobinfo->quarter;
break;
- case SELECT_DATA_CHECKED:
- *tmp_16 = jobinfo->checked;
- break;
default:
debug("select_g_get_jobinfo data_type %d invalid",
data_type);
diff --git a/src/common/node_select.h b/src/common/node_select.h
index 90c10ebe34c..cd1102985f7 100644
--- a/src/common/node_select.h
+++ b/src/common/node_select.h
@@ -113,10 +113,11 @@ extern int select_g_job_init(List job_list);
* IN/OUT bitmap - map of nodes being considered for allocation on input,
* map of nodes actually to be assigned on output
* IN min_nodes - minimum number of nodes to allocate to job
- * IN max_nodes - maximum number of nodes to allocate to job
+ * IN max_nodes - maximum number of nodes to allocate to job
+ * IN test_only - if true, only test if ever could run, not necessarily now
*/
extern int select_g_job_test(struct job_record *job_ptr, bitstr_t *bitmap,
- int min_nodes, int max_nodes);
+ int min_nodes, int max_nodes, bool test_only);
/*
* Note initiation of job is about to begin. Called immediately
diff --git a/src/plugins/select/bluegene/plugin/bg_job_place.c b/src/plugins/select/bluegene/plugin/bg_job_place.c
index 07c7130c584..6029c707f28 100644
--- a/src/plugins/select/bluegene/plugin/bg_job_place.c
+++ b/src/plugins/select/bluegene/plugin/bg_job_place.c
@@ -44,7 +44,8 @@ _STMT_START { \
static int _find_best_block_match(struct job_record* job_ptr,
bitstr_t* slurm_block_bitmap,
int min_nodes, int max_nodes,
- int spec, bg_record_t** found_bg_record);
+ int spec, bg_record_t** found_bg_record,
+ bool test_only);
static void _rotate_geo(uint16_t *req_geometry, int rot_cnt);
/* Rotate a 3-D geometry array through its six permutations */
@@ -87,47 +88,22 @@ static void _rotate_geo(uint16_t *req_geometry, int rot_cnt)
*/
static int _find_best_block_match(struct job_record* job_ptr,
bitstr_t* slurm_block_bitmap, int min_nodes, int max_nodes,
- int spec, bg_record_t** found_bg_record)
+ int spec, bg_record_t** found_bg_record, bool test_only)
{
ListIterator itr;
bg_record_t* record = NULL;
- int i;
+ int i, job_running = 0;
uint16_t req_geometry[BA_SYSTEM_DIMENSIONS];
- uint16_t conn_type, rotate, target_size = 1, checked;
+ uint16_t conn_type, rotate, target_size = 1;
uint32_t req_procs = job_ptr->num_procs;
int rot_cnt = 0;
uint32_t proc_cnt;
- int job_running = 0;
if(!bg_list) {
error("_find_best_block_match: There is no bg_list");
return SLURM_ERROR;
}
- select_g_get_jobinfo(job_ptr->select_jobinfo,
- SELECT_DATA_CHECKED, &checked);
-
- /* have to check checked to see which time the node
- scheduler is looking to see if it is runnable. If checked >=2
- we want to fall through to tell the scheduler that it is runnable
- just not right now.
- */
- slurm_mutex_lock(&block_state_mutex);
- if((full_system_block->job_running != -1) && (checked < 2)) {
- checked++;
- slurm_mutex_unlock(&block_state_mutex);
- select_g_set_jobinfo(job_ptr->select_jobinfo,
- SELECT_DATA_CHECKED, &checked);
-
- debug("_find_best_block_match none found "
- "full system running on block %s. %d",
- full_system_block->bg_block_id,
- full_system_block->job_running);
-
- return SLURM_ERROR;
- }
- slurm_mutex_unlock(&block_state_mutex);
-
select_g_get_jobinfo(job_ptr->select_jobinfo,
SELECT_DATA_CONN_TYPE, &conn_type);
select_g_get_jobinfo(job_ptr->select_jobinfo,
@@ -146,30 +122,16 @@ static int _find_best_block_match(struct job_record* job_ptr,
debug("number of blocks to check: %d", list_count(bg_list));
itr = list_iterator_create(bg_list);
while ((record = (bg_record_t*) list_next(itr))) {
- /* Check processor count */
- /* have to check checked to see which time the node
- scheduler is looking to see if it is runnable.
- If checked >=2 we want to fall through to tell the
- scheduler that it is runnable just not right now.
- */
- slurm_mutex_lock(&block_state_mutex);
- debug3("job_running = %d", record->job_running);
- if((record->job_running != -1) && (checked < 2)) {
+ if ((record->job_running != -1) && (!test_only)) {
job_running++;
- debug("block %s in use by %s",
- record->bg_block_id,
- record->user_name);
- slurm_mutex_unlock(&block_state_mutex);
continue;
}
- slurm_mutex_unlock(&block_state_mutex);
-
if(record->full_block && job_running) {
debug("Can't run on full system block "
- "another block has a job running.");
+ "another block has a job running.");
continue;
}
-
+
if (req_procs > record->cnodes_per_bp) {
/* We use the c-node count here. Job could start
* twice this count if VIRTUAL_NODE_MODE, but this
@@ -260,9 +222,6 @@ static int _find_best_block_match(struct job_record* job_ptr,
break;
}
list_iterator_destroy(itr);
- checked++;
- select_g_set_jobinfo(job_ptr->select_jobinfo,
- SELECT_DATA_CHECKED, &checked);
/* set the bitmap and do other allocation activities */
if (*found_bg_record) {
@@ -284,10 +243,11 @@ static int _find_best_block_match(struct job_record* job_ptr,
* be used
* IN min_nodes, max_nodes - minimum and maximum number of nodes to allocate
* to this job (considers slurm block limits)
+ * IN test_only - if true, only test if ever could run, not necessarily now
* RET - SLURM_SUCCESS if job runnable now, error code otherwise
*/
extern int submit_job(struct job_record *job_ptr, bitstr_t *slurm_block_bitmap,
- int min_nodes, int max_nodes)
+ int min_nodes, int max_nodes, bool test_only)
{
int spec = 1; /* this will be like, keep TYPE a priority, etc, */
bg_record_t* record = NULL;
@@ -301,7 +261,8 @@ extern int submit_job(struct job_record *job_ptr, bitstr_t *slurm_block_bitmap,
max_nodes);
if ((_find_best_block_match(job_ptr, slurm_block_bitmap, min_nodes,
- max_nodes, spec, &record)) == SLURM_ERROR) {
+ max_nodes, spec, &record, test_only))
+ == SLURM_ERROR) {
return SLURM_ERROR;
} else {
/* set the block id and quarter (if any) */
diff --git a/src/plugins/select/bluegene/plugin/bg_job_place.h b/src/plugins/select/bluegene/plugin/bg_job_place.h
index f68e857af2c..38d6190e4d9 100644
--- a/src/plugins/select/bluegene/plugin/bg_job_place.h
+++ b/src/plugins/select/bluegene/plugin/bg_job_place.h
@@ -2,7 +2,7 @@
* bg_job_place.h - header for blue gene job placement (e.g. base partition
* selection) functions.
*****************************************************************************
- * Copyright (C) 2004 The Regents of the University of California.
+ * Copyright (C) 2004-2006 The Regents of the University of California.
* Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER).
* Written by Dan Phung <phung4@llnl.gov> et. al.
*
@@ -36,9 +36,10 @@
* be used
* IN min_nodes, max_nodes - minimum and maximum number of nodes to allocate
* to this job (considers slurm partition limits)
+ * IN test_only - if true, only test if ever could run, not necessarily now
* RET - SLURM_SUCCESS if job runnable now, error code otherwise
*/
extern int submit_job(struct job_record *job_ptr, bitstr_t *bitmap,
- int min_nodes, int max_nodes);
+ int min_nodes, int max_nodes, bool test_only);
#endif /* _BG_JOB_PLACE_H_ */
diff --git a/src/plugins/select/bluegene/plugin/bluegene.c b/src/plugins/select/bluegene/plugin/bluegene.c
index 3a0c250109e..faf2be73b35 100644
--- a/src/plugins/select/bluegene/plugin/bluegene.c
+++ b/src/plugins/select/bluegene/plugin/bluegene.c
@@ -1447,7 +1447,7 @@ static int _delete_old_blocks(void)
}
if(found_record == NULL) {
debug("deleting %s",
- found_record->bg_block_id);
+ init_record->bg_block_id);
slurm_attr_init(&attr_agent);
if (pthread_attr_setdetachstate(
&attr_agent,
diff --git a/src/plugins/select/bluegene/plugin/select_bluegene.c b/src/plugins/select/bluegene/plugin/select_bluegene.c
index 23dd477e2d9..7216f37c699 100644
--- a/src/plugins/select/bluegene/plugin/select_bluegene.c
+++ b/src/plugins/select/bluegene/plugin/select_bluegene.c
@@ -229,12 +229,13 @@ extern int select_p_node_init(struct node_record *node_ptr, int node_cnt)
* satisfy the request are cleared, other left set
* IN min_nodes - minimum count of nodes
* IN max_nodes - maximum count of nodes (0==don't care)
+ * IN test_only - if true, only test if ever could run, not necessarily now
* RET zero on success, EINVAL otherwise
* NOTE: bitmap must be a superset of req_nodes at the time that
* select_p_job_test is called
*/
extern int select_p_job_test(struct job_record *job_ptr, bitstr_t *bitmap,
- int min_nodes, int max_nodes)
+ int min_nodes, int max_nodes, bool test_only)
{
/* bg block test - is there a block where we have:
* 1) geometry requested
@@ -246,7 +247,7 @@ extern int select_p_job_test(struct job_record *job_ptr, bitstr_t *bitmap,
* as the SLURM block logic will handle access rights.
*/
- return submit_job(job_ptr, bitmap, min_nodes, max_nodes);
+ return submit_job(job_ptr, bitmap, min_nodes, max_nodes, test_only);
}
extern int select_p_job_begin(struct job_record *job_ptr)
diff --git a/src/plugins/select/cons_res/select_cons_res.c b/src/plugins/select/cons_res/select_cons_res.c
index ee246631715..859575fcf66 100644
--- a/src/plugins/select/cons_res/select_cons_res.c
+++ b/src/plugins/select/cons_res/select_cons_res.c
@@ -641,6 +641,8 @@ extern int select_p_block_init(List part_list)
* satisfy the request are cleared, other left set
* IN min_nodes - minimum count of nodes
* IN max_nodes - maximum count of nodes (0==don't care)
+ * IN test_only - if true, only test if ever could run, not necessarily now,
+ * not used in this implementation
* RET zero on success, EINVAL otherwise
* globals (passed via select_p_node_init):
* node_record_count - count of nodes configured
@@ -653,7 +655,7 @@ extern int select_p_block_init(List part_list)
* select_p_job_test is called
*/
extern int select_p_job_test(struct job_record *job_ptr, bitstr_t * bitmap,
- int min_nodes, int max_nodes)
+ int min_nodes, int max_nodes, bool test_only)
{
int i, index, error_code = EINVAL, sufficient;
int *consec_nodes; /* how many nodes we can add from this
diff --git a/src/plugins/select/linear/select_linear.c b/src/plugins/select/linear/select_linear.c
index 75729b4f8d9..f9a56960db8 100644
--- a/src/plugins/select/linear/select_linear.c
+++ b/src/plugins/select/linear/select_linear.c
@@ -170,6 +170,8 @@ extern int select_p_block_init(List part_list)
* satisfy the request are cleared, other left set
* IN min_nodes - minimum count of nodes
* IN max_nodes - maximum count of nodes (0==don't care)
+ * IN test_only - if true, only test if ever could run, not necessarily now,
+ * not used in this implementation of plugin
* RET zero on success, EINVAL otherwise
* globals (passed via select_p_node_init):
* node_record_count - count of nodes configured
@@ -182,7 +184,7 @@ extern int select_p_block_init(List part_list)
* select_p_job_test is called
*/
extern int select_p_job_test(struct job_record *job_ptr, bitstr_t *bitmap,
- int min_nodes, int max_nodes)
+ int min_nodes, int max_nodes, bool test_only)
{
int i, index, error_code = EINVAL, sufficient;
int *consec_nodes; /* how many nodes we can add from this
diff --git a/src/plugins/switch/federation/federation.c b/src/plugins/switch/federation/federation.c
index 8e95379b557..b4643699ab7 100644
--- a/src/plugins/switch/federation/federation.c
+++ b/src/plugins/switch/federation/federation.c
@@ -1542,6 +1542,7 @@ _window_state_set(int adapter_cnt, fed_tableinfo_t *tableinfo,
fed_window_t *window;
NTBL *table;
int i, j;
+ bool adapter_found;
assert(tableinfo);
assert(hostname);
@@ -1568,13 +1569,18 @@ _window_state_set(int adapter_cnt, fed_tableinfo_t *tableinfo,
return SLURM_ERROR;
}
+ adapter_found = false;
/* Find the adapter that matches the one in tableinfo */
for (j = 0; j < node->adapter_count; j++) {
adapter = &node->adapter_list[j];
- if (adapter->lid == table->lid)
+ if (strcasecmp(adapter->name,
+ tableinfo[i].adapter_name) == 0
+ && adapter->lid == table->lid) {
+ adapter_found = true;
break;
+ }
}
- if (adapter->lid != table->lid) {
+ if (!adapter_found) {
if (table->lid != 0)
error("Did not find the correct adapter: "
"%hu vs. %hu",
diff --git a/src/sacct/sacct.c b/src/sacct/sacct.c
index 6df1dd9506d..a786c53dba8 100644
--- a/src/sacct/sacct.c
+++ b/src/sacct/sacct.c
@@ -1900,7 +1900,7 @@ void helpMsg(void)
" * Elapsed time fields are presented as 2 fields, integral\n"
" seconds and integral microseconds\n"
" * If --dump is not specified, elapsed time fields are presented\n"
- " as [[days:]hours:]minutes:seconds.hundredths\n"
+ " as [[days-]hours:]minutes:seconds.hundredths\n"
" * The default input file is the file named in the \"jobacct_loc\"\n"
" parameter in " SLURM_CONFIG_FILE ".\n"
"\n"
diff --git a/src/slurmctld/node_scheduler.c b/src/slurmctld/node_scheduler.c
index 5ec62daee57..cacdf43fae5 100644
--- a/src/slurmctld/node_scheduler.c
+++ b/src/slurmctld/node_scheduler.c
@@ -289,13 +289,13 @@ _pick_best_load(struct job_record *job_ptr, bitstr_t * bitmap,
bit_or(bitmap, job_ptr->details->req_node_bitmap);
error_code = select_g_job_test(job_ptr, bitmap,
- min_nodes, max_nodes);
+ min_nodes, max_nodes, false);
/* now try to use idle and lightly loaded nodes */
if (error_code) {
bit_or(bitmap, light_load_bit);
error_code = select_g_job_test(job_ptr, bitmap,
- min_nodes, max_nodes);
+ min_nodes, max_nodes, false);
}
FREE_NULL_BITMAP(light_load_bit);
@@ -303,7 +303,7 @@ _pick_best_load(struct job_record *job_ptr, bitstr_t * bitmap,
if (error_code) {
bit_or(bitmap, heavy_load_bit);
error_code = select_g_job_test(job_ptr, bitmap,
- min_nodes, max_nodes);
+ min_nodes, max_nodes, false);
}
FREE_NULL_BITMAP(heavy_load_bit);
@@ -402,9 +402,7 @@ _pick_best_nodes(struct node_set *node_set_ptr, int node_set_size,
bool runable_ever = false; /* Job can ever run */
bool runable_avail = false; /* Job can run with available nodes */
int cr_enabled = 0;
-#ifdef HAVE_BG
- uint16_t checked = 0;
-#endif
+
if (node_set_size == 0) {
info("_pick_best_nodes: empty node set for selection");
return ESLURM_REQUESTED_NODE_CONFIG_UNAVAILABLE;
@@ -607,7 +605,7 @@ _pick_best_nodes(struct node_set *node_set_ptr, int node_set_size,
pick_code = select_g_job_test(job_ptr,
avail_bitmap,
min_nodes,
- max_nodes);
+ max_nodes, false);
if (pick_code == SLURM_SUCCESS) {
if ((node_lim != INFINITE) &&
@@ -630,7 +628,7 @@ _pick_best_nodes(struct node_set *node_set_ptr, int node_set_size,
(avail_nodes >= min_nodes) &&
(avail_nodes < max_nodes)) {
pick_code = select_g_job_test(job_ptr, avail_bitmap,
- min_nodes, max_nodes);
+ min_nodes, max_nodes, false);
if ((pick_code == SLURM_SUCCESS) &&
((node_lim == INFINITE) ||
(bit_set_count(avail_bitmap) <= node_lim))) {
@@ -663,7 +661,7 @@ _pick_best_nodes(struct node_set *node_set_ptr, int node_set_size,
pick_code = select_g_job_test(job_ptr,
avail_bitmap,
min_nodes,
- max_nodes);
+ max_nodes, true);
if (cr_enabled)
job_ptr->cr_enabled = 1;
if (pick_code == SLURM_SUCCESS) {
@@ -680,7 +678,7 @@ _pick_best_nodes(struct node_set *node_set_ptr, int node_set_size,
pick_code = select_g_job_test(job_ptr,
total_bitmap,
min_nodes,
- max_nodes);
+ max_nodes, true);
if (cr_enabled)
job_ptr->cr_enabled = 1;
if (pick_code == SLURM_SUCCESS)
@@ -704,10 +702,6 @@ _pick_best_nodes(struct node_set *node_set_ptr, int node_set_size,
error_code = ESLURM_REQUESTED_NODE_CONFIG_UNAVAILABLE;
info("_pick_best_nodes: job never runnable");
}
-#ifdef HAVE_BG
- select_g_set_jobinfo(job_ptr->select_jobinfo,
- SELECT_DATA_CHECKED, &checked);
-#endif
if (error_code == SLURM_SUCCESS)
error_code = ESLURM_NODES_BUSY;
return error_code;
diff --git a/src/squeue/print.c b/src/squeue/print.c
index bde6da2565c..978f4615e15 100644
--- a/src/squeue/print.c
+++ b/src/squeue/print.c
@@ -562,7 +562,7 @@ int _print_job_reason_list(job_info_t * job, int width, bool right,
_print_nodes(job->nodes, width, right, false);
if(quarter != -1) {
- sprintf(tmp_char,"0.%d",quarter);
+ sprintf(tmp_char,".%d",quarter);
_print_str(tmp_char, width, right, false);
}
}
diff --git a/src/srun/srun.c b/src/srun/srun.c
index 416b404b998..407d21867a1 100644
--- a/src/srun/srun.c
+++ b/src/srun/srun.c
@@ -238,8 +238,10 @@ long start, end;
_print_job_information(resp);
job = job_create_allocation(resp);
- if (create_job_step(job, resp) < 0)
+ if (create_job_step(job, resp) < 0) {
+ srun_job_destroy(job, 0);
exit(1);
+ }
slurm_free_resource_allocation_response_msg(resp);
}
--
GitLab