diff --git a/NEWS b/NEWS
index 1c0119ac9a4c921be2bd02de1b8be6b4cf4ae046..0bc32ee872d2d7bb494b2d141b8365b2d93b384e 100644
--- a/NEWS
+++ b/NEWS
@@ -19,13 +19,16 @@ documents those changes that are of interest to users and admins.
Based upon work by Kumar Krishna (HP, India).
-- Add multi-core options to salloc and sbatch commands (sbatch.patch and
cleanup.patch from Chris Holmes, HP).
- -- In select/cons_res properly release resources allocated to job being suspended
- (rmbreak.patch, from Chris Holmes, HP).
+ -- In select/cons_res properly release resources allocated to job being
+ suspended (rmbreak.patch, from Chris Holmes, HP).
-- Removed database and jobacct plugin replaced with jobacct_storage
and jobacct_gather for easier hooks for further expansion of the
jobacct plugin.
-- Job step launch in srun is now done from the slurm api's all futher
modifications to job launch should be done there.
+ -- Add new partition configuration parameter Priority and replace Shared
+ with MaxShare.
+ -- Add new configuration parameter SchedulerTimeSlice.
* Changes in SLURM 1.3.0-pre2
=============================
diff --git a/RELEASE_NOTES b/RELEASE_NOTES
index 928f99b01fc915dcfbe92358894da9c2db28289e..a4093c8df66f9fb5f8806ff77b5b68f7798df848 100644
--- a/RELEASE_NOTES
+++ b/RELEASE_NOTES
@@ -53,6 +53,20 @@ CONFIGURATION FILE CHANGES
specify how SLURM will talk to the database.
* Added new paramter "CryptoType" to specify digital signature plugin to
be used. Currenly only supports crypto/openssl.
+* Added new parameter "SchedulerTimeSlice" to control the lenght of gang scheduler
+ time slices.
+* Added new partition parameter "Priority". A job's scheduling priority is based
+ upon two factors. First the priority of its partition and the job's priority.
+ Since nodes can be configured in multiple partitions, this can be used to configure
+ high priority partitions (queues).
+* The partition parameter "Shared" has been replace by "MaxShare" which controls the
+ number of jobs which can be executed on a given set of resources. Here are equivalent
+ values of the parameters:
+ Shared=EXCLUSIVE MaxShare=0
+ Shared=NO MaxShare=1
+ Shared=YES MaxShare=4 (or some other small number, two or higher)
+ Shared=FORCE MaxShare=65535
+
* See "man slurm.conf" for more information.
OTHER CHANGES
diff --git a/doc/html/configurator.html.in b/doc/html/configurator.html.in
index 2f1c04e180bea93f139c3ee19b181c75ad71661f..121a7df36408fe4dee8683f06a2e7bff5576ff68 100644
--- a/doc/html/configurator.html.in
+++ b/doc/html/configurator.html.in
@@ -189,6 +189,7 @@ function displayfile()
"# SCHEDULING<br>" +
"SchedulerType=sched/" + get_radio_value(document.config.sched_type) + "<br>" +
get_field("SchedulerPort",document.config.scheduler_port) + "<br>" +
+ "#SchedulerTimeSlice= <br>" +
"#SchedulerRootFilter= <br>" +
"SelectType=select/" + get_radio_value(document.config.select_type) + "<br>" +
get_select_type_params() + "<br>" +
@@ -507,7 +508,7 @@ individual processors and memory<BR>
when CR_CPU or CR_CPU_MEMORY is selected.
They are considered to compute the total number of
tasks when -n is not specified
- <DD> Note: CR_MEMORY assumes Shared=Yes
+ <DD> Note: CR_MEMORY assumes MaxShare value of one of higher
<DT> <input type="radio" name="cons_res_params" value="CR_CPU" checked
onClick="javascript:set_select_type(this, 'cons_res')">
<B>CR_CPU</B>: (default)
@@ -525,7 +526,7 @@ individual processors and memory<BR>
<DT> <input type="radio" name="cons_res_params" value="CR_Memory"
onClick="javascript:set_select_type(this)">
<B>CR_Memory</B>: Memory as a consumable resource.
- <DD> Note: CR_Memory assumes Shared=Yes
+ <DD> Note: CR_Memory assumes MaxShare value of one of higher
<DT> <input type="radio" name="cons_res_params" value="CR_CPU_Memory"
onClick="javascript:set_select_type(this)">
<B>CR_CPU_Memory</B>:
@@ -723,6 +724,6 @@ before terminating all remaining tasks. A value of zero indicates unlimited wait
</FORM>
<HR>
<p class="footer">UCRL-WEB-225274<br>
-Last modified 11 Sep 2007</p>
+Last modified 21 September 2007</p>
</BODY>
diff --git a/doc/man/man5/slurm.conf.5 b/doc/man/man5/slurm.conf.5
index be46f68245470c08c8cfd87d4f11383244032cb5..a5a619fbe78ff49d41485c4a622df47dfb3424d7 100644
--- a/doc/man/man5/slurm.conf.5
+++ b/doc/man/man5/slurm.conf.5
@@ -1,4 +1,4 @@
-.TH "slurm.conf" "5" "July 2007" "slurm.conf 1.3" "Slurm configuration file"
+.TH "slurm.conf" "5" "September 2007" "slurm.conf 1.3" "Slurm configuration file"
.SH "NAME"
slurm.conf \- Slurm configuration file
.SH "DESCRIPTION"
@@ -572,14 +572,6 @@ default value is 0, which means that a node will remain in the
DOWN state until a system administrator explicitly changes its state
(even if the slurmd daemon registers and resumes communications).
-.TP
-\fBSchedulerRootFilter\fR
-If set to '1' then scheduler will filter and avoid \fBRootOnly\fR
-partitions (let root user or process schedule these partitions).
-Otherwise scheduler will treat \fBRootOnly\fR partitions as any
-other standard partition.
-Currently only supported by sched/backfill schedululer plugin.
-
.TP
\fBSchedulerPort\fR
The port number on which slurmctld should listen for connection requests.
@@ -595,12 +587,18 @@ partitions are exempt from any external scheduling activities. The
default value is 1. Currently only used by the built\-in backfill
scheduling module "sched/backfill" (see \fBSchedulerType\fR).
+.TP
+\fBSchedulerTimeSlice\fR
+Number of seconds in each time slice when \fBSchedulerType=sched/gang\fR.
+The default value is 30.
+
.TP
\fBSchedulerType\fR
Identifies the type of scheduler to be used. Acceptable values include
"sched/builtin" for the built\-in FIFO scheduler,
"sched/backfill" for a backfill scheduling module to augment
the default FIFO scheduling,
+"sched/gang" for gang scheduler (time\-slicing of parallel jobs),
"sched/hold" to hold all newly arriving jobs if a file "/etc/slurm.hold"
exists otherwise use the built\-in FIFO scheduler, and
"sched/wiki" for the Wiki interface to the Maui Scheduler.
@@ -636,8 +634,8 @@ This is the default value for non\-BlueGene systems.
The resources within a node are individually allocated as
consumable resources.
Note that whole nodes can be allocated to jobs for selected
-partitions by using the \fIShared=EXCLUSIVE\fR option.
-See the partition \fBShared\fR parameter for more information.
+partitions by using the \fIMaxShare=\fR option.
+See the partition \fBMaxShare\fR parameter for more information.
.TP
\fBselect/bluegene\fR
for a three\-dimentional BlueGene system.
@@ -674,7 +672,7 @@ Memory and CPUs are consumable resources.
.TP
\fBCR_Memory\fR
Memory is a consumable resource.
-NOTE: This implies \fIShared=Yes\fR for all partitions.
+NOTE: This implies \fIMaxShare\fR is non\-zero for all partitions.
.RE
.TP
@@ -1311,6 +1309,25 @@ may be allocated to any single job.
The default value is "UNLIMITED", which is represented internally as \-1.
This limit does not apply to jobs executed by SlurmUser or user root.
+.TP
+\fBMaxShare\fR
+Ability of the partition to execute more than one job at a
+time on each node. Shared nodes will offer unpredictable performance
+for application programs, but can provide higher system utilization
+and responsiveness than otherwise possible.
+Accepts a numeric value indicating the number of jobs which can
+be allocated each resources in the partition.
+For example, a value of two indicates that two jobs can simultaneously
+be allocated each resources.
+If \fBSchedulerType=sched/gang\fR then those jobs can be time sliced.
+Other values of \fBSchedulerType\fR will rely upon the operating
+system to perform timesharing of the resouces to active jobs.
+The default value is 1 (resources are not shared).
+A value of 0 is treated as a special case to allocate entire
+nodes to jobs even with \fBSelectType=select/cons_res\fR.
+A value of 65535 is treated as a special case to make all nodes
+available for sharing without any user means of disabling it.
+
.TP
\fBMaxTime\fR
Maximum wall\-time limit for any job in minutes. The default
@@ -1342,24 +1359,18 @@ with that record will apply to subsequent partition specifications
unless explicitly set to other values in that partition record or
replaced with a different set of default values.
+.TP
+\fBPriority\fR
+Jobs submitted to a higher priority partition will be dispatched
+before pending jobs in lower priority partitions and if possible
+they will preempt running jobs from lower priority partitions.
+Note that a partition's priority takes precedence over a job's
+priority.
+The value may not exceed 65533.
+
.TP
\fBShared\fR
-Ability of the partition to execute more than one job at a
-time on each node. Shared nodes will offer unpredictable performance
-for application programs, but can provide higher system utilization
-and responsiveness than otherwise possible.
-Possible values are "EXCLUSIVE", "FORCE", "YES", and "NO".
-"EXCLUSIVE" allocates entire nodes to jobs even with
-select/cons_res configured.
-This can be used to allocate whole nodes in some partitions
-and individual processors in other partitions.
-"FORCE" makes all nodes in the partition available for sharing
-without user means of disabling it.
-"YES" makes nodes in the partition available for sharing if and
-only if the individual jobs permit sharing (see the srun
-"\-\-share" option).
-"NO" makes nodes unavailable for sharing under all circumstances.
-The default value is "NO".
+This parameter has been replace by \fBMaxShare\fR.
.TP
\fBState\fR
diff --git a/slurm/slurm.h.in b/slurm/slurm.h.in
index a716b0f223df2f3cf2886520702548076a8df273..291eab2076ee1daec76dd1cf600256e30fdc0ab8 100644
--- a/slurm/slurm.h.in
+++ b/slurm/slurm.h.in
@@ -425,13 +425,6 @@ typedef enum select_type_plugin_info {
#define TASK_PARAM_CPUSETS 0x0001
#define TASK_PARAM_SCHED 0x0002
-enum part_shared {
- SHARED_NO = 0, /* Nodes never shared in partition */
- SHARED_YES, /* Nodes possible to share in partition */
- SHARED_FORCE, /* Nodes always shares in partition */
- SHARED_EXCLUSIVE /* Nodes never shared even with cons_res */
-};
-
/*****************************************************************************\
* PROTOCOL DATA STRUCTURE DEFINITIONS
\*****************************************************************************/
@@ -819,8 +812,9 @@ typedef struct partition_info {
uint16_t node_scaling; /* select plugin node scaling factor */
uint16_t default_part; /* 1 if this is default partition */
uint16_t hidden; /* 1 if partition is hidden by default */
+ uint16_t priority; /* scheduling priority for jobs */
uint16_t root_only; /* 1 if allocate must come for user root */
- uint16_t shared; /* See part_shared above */
+ uint16_t max_share; /* number of jobs to gang schedule */
uint16_t state_up; /* 1 if state is up, 0 if down */
char *nodes; /* list names of nodes in partition */
int *node_inx; /* list index pairs into node_table:
@@ -920,6 +914,7 @@ typedef struct slurm_ctl_conf {
* registration */
uint16_t resume_rate; /* nodes to make full power, per minute */
char *resume_program; /* program to make nodes full power */
+ uint16_t sched_time_slice; /* gang scheduler slice time, secs */
char *schedtype; /* type of scheduler to use */
uint16_t schedport; /* port for scheduler connection */
uint16_t schedrootfltr; /* 1 if rootOnly partitions should be
diff --git a/src/api/config_info.c b/src/api/config_info.c
index 4f47f608f4a30f9d4ddec14dacd2a1ae955ea219..cce832844c52faaf16aed93273528898e6c36003 100644
--- a/src/api/config_info.c
+++ b/src/api/config_info.c
@@ -221,6 +221,8 @@ void slurm_print_ctl_conf ( FILE* out,
slurm_ctl_conf_ptr->schedport);
fprintf(out, "SchedulerRootFilter = %u\n",
slurm_ctl_conf_ptr->schedrootfltr);
+ fprintf(out, "SchedulerTimeSlice = %u\n",
+ slurm_ctl_conf_ptr->sched_time_slice);
fprintf(out, "SchedulerType = %s\n",
slurm_ctl_conf_ptr->schedtype);
fprintf(out, "SelectType = %s\n",
diff --git a/src/api/init_msg.c b/src/api/init_msg.c
index f06d15829cdfaab22380311246d5acb2cbc2dbc0..7d593eff4175a7abd294f160ba77ceaf92da6afe 100644
--- a/src/api/init_msg.c
+++ b/src/api/init_msg.c
@@ -150,7 +150,8 @@ void slurm_init_part_desc_msg (update_part_msg_t * update_part_msg)
update_part_msg->hidden = (uint16_t) NO_VAL;
update_part_msg->default_part = (uint16_t) NO_VAL;
update_part_msg->root_only = (uint16_t) NO_VAL;
- update_part_msg->shared = (uint16_t) NO_VAL;
+ update_part_msg->max_share = (uint16_t) NO_VAL;
+ update_part_msg->priority = (uint16_t) NO_VAL;
update_part_msg->state_up = (uint16_t) NO_VAL;
}
diff --git a/src/api/partition_info.c b/src/api/partition_info.c
index 2eff5fa6c22658bfc106736945e75d679ccce25f..f6a0b4e80e3a5305ad944a9d22d323986f89f0f2 100644
--- a/src/api/partition_info.c
+++ b/src/api/partition_info.c
@@ -2,7 +2,7 @@
* partition_info.c - get/print the partition state information of slurm
* $Id$
*****************************************************************************
- * Copyright (C) 2002-2006 The Regents of the University of California.
+ * Copyright (C) 2002-2007 The Regents of the University of California.
* Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER).
* Written by Morris Jette <jette1@llnl.gov> et. al.
* UCRL-CODE-226842.
@@ -137,14 +137,8 @@ char *slurm_sprint_partition_info ( partition_info_t * part_ptr,
else
sprintf(tmp_line, "Default=NO ");
xstrcat(out, tmp_line);
- if (part_ptr->shared == SHARED_NO)
- sprintf(tmp_line, "Shared=NO ");
- else if (part_ptr->shared == SHARED_YES)
- sprintf(tmp_line, "Shared=YES ");
- else if (part_ptr->shared == SHARED_EXCLUSIVE)
- sprintf(tmp_line, "Shared=EXCLUSIVE ");
- else
- sprintf(tmp_line, "Shared=FORCE ");
+ sprintf(tmp_line, "MaxShare=%u Priority=%u ",
+ part_ptr->max_share, part_ptr->priority);
xstrcat(out, tmp_line);
if (part_ptr->state_up)
sprintf(tmp_line, "State=UP ");
diff --git a/src/common/read_config.c b/src/common/read_config.c
index 750b8db3118c573b3ab30bb33484be73d023813d..217eed6c4a508649570ae86e8d61824a3cd68134 100644
--- a/src/common/read_config.c
+++ b/src/common/read_config.c
@@ -175,6 +175,7 @@ s_p_options_t slurm_conf_options[] = {
{"SchedulerAuth", S_P_STRING},
{"SchedulerPort", S_P_UINT16},
{"SchedulerRootFilter", S_P_UINT16},
+ {"SchedulerTimeSlice", S_P_UINT16},
{"SchedulerType", S_P_STRING},
{"SelectType", S_P_STRING},
{"SelectTypeParameters", S_P_STRING},
@@ -422,8 +423,10 @@ static int parse_partitionname(void **dest, slurm_parser_enum_t type,
{"Hidden", S_P_BOOLEAN}, /* YES or NO */
{"MaxTime", S_P_UINT32}, /* INFINITE or a number */
{"MaxNodes", S_P_UINT32}, /* INFINITE or a number */
+ {"MaxShare", S_P_UINT16},
{"MinNodes", S_P_UINT32},
{"Nodes", S_P_STRING},
+ {"Priority", S_P_UINT16},
{"RootOnly", S_P_BOOLEAN}, /* YES or NO */
{"Shared", S_P_STRING}, /* YES, NO, or FORCE */
{"State", S_P_BOOLEAN}, /* UP or DOWN */
@@ -482,20 +485,26 @@ static int parse_partitionname(void **dest, slurm_parser_enum_t type,
&& !s_p_get_boolean(&p->root_only_flag, "RootOnly", dflt))
p->root_only_flag = false;
- if (!s_p_get_string(&tmp, "Shared", tbl)
- && !s_p_get_string(&tmp, "Shared", dflt)) {
- p->shared = SHARED_NO;
- } else {
+ if (!s_p_get_uint16(&p->priority, "Priority", tbl) &&
+ !s_p_get_uint16(&p->priority, "Priority", dflt))
+ p->priority = 1;
+
+ if (s_p_get_uint16(&p->max_share, "MaxShare", tbl) ||
+ s_p_get_uint16(&p->max_share, "MaxShare", dflt)) {
+ /* Use MaxShare and ignore Shared value */
+ ;
+ } else if (s_p_get_string(&tmp, "Shared", tbl) ||
+ s_p_get_string(&tmp, "Shared", dflt)) {
if (strcasecmp(tmp, "NO") == 0)
- p->shared = SHARED_NO;
+ p->max_share = 1;
#ifndef HAVE_XCPU
/* Only "Shared=NO" is valid on XCPU systems */
else if (strcasecmp(tmp, "YES") == 0)
- p->shared = SHARED_YES;
+ p->max_share = (uint16_t) 64;
else if (strcasecmp(tmp, "EXCLUSIVE") == 0)
- p->shared = SHARED_EXCLUSIVE;
+ p->max_share = 0;
else if (strcasecmp(tmp, "FORCE") == 0)
- p->shared = SHARED_FORCE;
+ p->max_share = (uint16_t) INFINITE;
#endif
else {
error("Bad value \"%s\" for Shared", tmp);
@@ -504,7 +513,11 @@ static int parse_partitionname(void **dest, slurm_parser_enum_t type,
xfree(tmp);
return -1;
}
+ } else {
+ /* No MaxShare or Shared specified */
+ p->min_nodes = 1;
}
+
xfree(tmp);
if (!s_p_get_boolean(&p->state_up_flag, "State", tbl)
@@ -1122,6 +1135,7 @@ init_slurm_conf (slurm_ctl_conf_t *ctl_conf_ptr)
xfree (ctl_conf_ptr->resume_program);
ctl_conf_ptr->resume_rate = (uint16_t) NO_VAL;
ctl_conf_ptr->ret2service = (uint16_t) NO_VAL;
+ ctl_conf_ptr->sched_time_slice = (uint16_t) NO_VAL;
ctl_conf_ptr->schedport = (uint16_t) NO_VAL;
ctl_conf_ptr->schedrootfltr = (uint16_t) NO_VAL;
xfree( ctl_conf_ptr->schedtype );
@@ -1640,6 +1654,10 @@ validate_and_set_defaults(slurm_ctl_conf_t *conf, s_p_hashtbl_t *hashtbl)
conf->schedport = DEFAULT_SCHEDULER_PORT;
}
+ if (!s_p_get_uint16(&conf->sched_time_slice, "SchedulerTimeSlice",
+ hashtbl))
+ conf->sched_time_slice = DEFAULT_SCHED_TIME_SLICE;
+
if (!s_p_get_uint16(&conf->schedrootfltr,
"SchedulerRootFilter", hashtbl))
conf->schedrootfltr = DEFAULT_SCHEDROOTFILTER;
diff --git a/src/common/read_config.h b/src/common/read_config.h
index 188774c2b36b5a2e119654c2d7135b2e3a48c038..ea9f91a6bfd1bfe1ad0a0b6bd687a9db4dbd65d2 100644
--- a/src/common/read_config.h
+++ b/src/common/read_config.h
@@ -91,6 +91,7 @@ extern char *default_plugstack;
#define DEFAULT_SAVE_STATE_LOC "/tmp"
#define DEFAULT_SCHEDROOTFILTER 1
#define DEFAULT_SCHEDULER_PORT 7321
+#define DEFAULT_SCHED_TIME_SLICE 30
#define DEFAULT_SCHEDTYPE "sched/builtin"
#ifdef HAVE_BG /* Blue Gene specific default configuration parameters */
# define DEFAULT_SELECT_TYPE "select/bluegene"
@@ -137,10 +138,10 @@ typedef struct slurm_conf_partition {
uint32_t min_nodes; /* per job */
uint32_t total_nodes; /* total number of nodes in the partition */
uint32_t total_cpus; /* total number of cpus in the partition */
+ uint16_t priority; /* scheduling priority for jobs */
bool root_only_flag;/* 1 if allocate/submit RPC can only be
issued by user root */
- uint16_t shared; /* 1 if job can share a node,
- 2 if sharing required */
+ uint16_t max_share; /* number of jobs to gang schedule */
bool state_up_flag; /* 1 if state is up, 0 if down */
char *nodes; /* comma delimited list names of nodes */
char *allow_groups; /* comma delimited list of groups,
diff --git a/src/common/slurm_protocol_pack.c b/src/common/slurm_protocol_pack.c
index 61795df2d5496bfcc191996e7da497c25b661929..af4c3b7ed89bf477eef86478cbcd130619fdf2d0 100644
--- a/src/common/slurm_protocol_pack.c
+++ b/src/common/slurm_protocol_pack.c
@@ -1423,9 +1423,10 @@ _pack_update_partition_msg(update_part_msg_t * msg, Buf buffer)
packstr(msg->name, buffer);
packstr(msg->nodes, buffer);
- pack16(msg-> hidden, buffer);
+ pack16(msg-> hidden, buffer);
+ pack16(msg-> max_share, buffer);
+ pack16(msg-> priority, buffer);
pack16(msg-> root_only, buffer);
- pack16(msg-> shared, buffer);
pack16(msg-> state_up, buffer);
}
@@ -1449,10 +1450,11 @@ _unpack_update_partition_msg(update_part_msg_t ** msg, Buf buffer)
safe_unpackstr_xmalloc(&tmp_ptr->name, &uint16_tmp, buffer);
safe_unpackstr_xmalloc(&tmp_ptr->nodes, &uint16_tmp, buffer);
- safe_unpack16(&tmp_ptr->hidden, buffer);
+ safe_unpack16(&tmp_ptr->hidden, buffer);
+ safe_unpack16(&tmp_ptr->max_share, buffer);
+ safe_unpack16(&tmp_ptr->priority, buffer);
safe_unpack16(&tmp_ptr->root_only, buffer);
- safe_unpack16(&tmp_ptr->shared, buffer);
- safe_unpack16(&tmp_ptr->state_up, buffer);
+ safe_unpack16(&tmp_ptr->state_up, buffer);
return SLURM_SUCCESS;
unpack_error:
@@ -1811,7 +1813,8 @@ _unpack_partition_info_members(partition_info_t * part, Buf buffer)
safe_unpack16(&part->default_part, buffer);
safe_unpack16(&part->hidden, buffer);
safe_unpack16(&part->root_only, buffer);
- safe_unpack16(&part->shared, buffer);
+ safe_unpack16(&part->max_share, buffer);
+ safe_unpack16(&part->priority, buffer);
safe_unpack16(&part->state_up, buffer);
safe_unpackstr_xmalloc(&part->allow_groups, &uint16_tmp, buffer);
@@ -2188,6 +2191,7 @@ _pack_slurm_ctl_conf_msg(slurm_ctl_conf_info_msg_t * build_ptr, Buf buffer)
pack16(build_ptr->schedport, buffer);
pack16(build_ptr->schedrootfltr, buffer);
+ pack16(build_ptr->sched_time_slice, buffer);
packstr(build_ptr->schedtype, buffer);
packstr(build_ptr->select_type, buffer);
pack16(build_ptr->select_type_param, buffer);
@@ -2328,6 +2332,7 @@ _unpack_slurm_ctl_conf_msg(slurm_ctl_conf_info_msg_t **
safe_unpack16(&build_ptr->schedport, buffer);
safe_unpack16(&build_ptr->schedrootfltr, buffer);
+ safe_unpack16(&build_ptr->sched_time_slice, buffer);
safe_unpackstr_xmalloc(&build_ptr->schedtype, &uint16_tmp, buffer);
safe_unpackstr_xmalloc(&build_ptr->select_type, &uint16_tmp, buffer);
safe_unpack16(&build_ptr->select_type_param, buffer);
diff --git a/src/plugins/sched/backfill/backfill.c b/src/plugins/sched/backfill/backfill.c
index 804d289cdf7bcd0b57712ce24e5c05a826b9173e..235b3e9f974dfa825785d53751d9b52c9376b3ca 100644
--- a/src/plugins/sched/backfill/backfill.c
+++ b/src/plugins/sched/backfill/backfill.c
@@ -198,7 +198,7 @@ backfill_agent(void *args)
part_iterator = list_iterator_create(part_list);
while ((part_ptr = (struct part_record *)
list_next(part_iterator))) {
- if ( ((part_ptr->shared) ||
+ if ( ((part_ptr->max_share > 1) ||
(part_ptr->state_up == 0)) )
continue; /* not under our control */
if ((part_ptr->root_only) && filter_root)
diff --git a/src/plugins/sched/wiki2/get_jobs.c b/src/plugins/sched/wiki2/get_jobs.c
index c8bd554429bd8716bf8fcb17200bf03fba0a2c1b..703da587dd356b06a3a81abec6046adcce170729 100644
--- a/src/plugins/sched/wiki2/get_jobs.c
+++ b/src/plugins/sched/wiki2/get_jobs.c
@@ -346,16 +346,15 @@ static void _get_job_comment(struct job_record *job_ptr,
/* SHARED NODES */
if (cr_enabled) { /* consumable resources */
- if (job_ptr->part_ptr->shared == SHARED_EXCLUSIVE)
+ if (job_ptr->part_ptr->max_share == 0) /* Exclusive use */
sharing = 0;
- else if (job_ptr->details && (job_ptr->details->shared != 0))
+ else if (job_ptr->details && job_ptr->details->shared)
sharing = 1;
- } else if (job_ptr->part_ptr) { /* partition with */
- if (job_ptr->part_ptr->shared == SHARED_FORCE)
- sharing = 1;
- else if ((job_ptr->part_ptr->shared == SHARED_YES)
- && (job_ptr->details) /* optional for partition */
- && (job_ptr->details->shared)) /* with job to share */
+ } else if (job_ptr->part_ptr) { /* partition level control */
+ if (job_ptr->part_ptr->max_share == (uint16_t) INFINITE)
+ sharing = 1; /* Sharing forced */
+ else if ((job_ptr->part_ptr->max_share > 1) &&
+ (job_ptr->details) && (job_ptr->details->shared))
sharing = 1;
}
if (sharing) {
diff --git a/src/plugins/select/bluegene/plugin/bg_job_place.c b/src/plugins/select/bluegene/plugin/bg_job_place.c
index 1651aff7cf0de8421c7da70610fa0be611472c6e..dc03a6250ea5e9a4b4ac02c34345b1b52f1ba2a3 100644
--- a/src/plugins/select/bluegene/plugin/bg_job_place.c
+++ b/src/plugins/select/bluegene/plugin/bg_job_place.c
@@ -911,7 +911,7 @@ extern int submit_job(struct job_record *job_ptr, bitstr_t *slurm_block_bitmap,
slurm_mutex_lock(&block_state_mutex);
if((record->ionodes)
- && (job_ptr->part_ptr->shared == 0))
+ && (job_ptr->part_ptr->max_share <= 1))
error("Small block used in "
"non-shared partition");
diff --git a/src/scontrol/update_part.c b/src/scontrol/update_part.c
index 4b21a854259338392f2fe18d72f275ee2ebabd68..d00f09c8fb8d3b5bd4d549ac219a1f251f5f1ce1 100644
--- a/src/scontrol/update_part.c
+++ b/src/scontrol/update_part.c
@@ -127,23 +127,14 @@ scontrol_update_part (int argc, char *argv[])
}
update_cnt++;
}
- else if (strncasecmp(argv[i], "Shared=", 7) == 0) {
- if (strcasecmp(&argv[i][7], "NO") == 0)
- part_msg.shared = SHARED_NO;
- else if (strcasecmp(&argv[i][7], "YES") == 0)
- part_msg.shared = SHARED_YES;
- else if (strcasecmp(&argv[i][7], "EXCLUSIVE") == 0)
- part_msg.shared = SHARED_EXCLUSIVE;
- else if (strcasecmp(&argv[i][7], "FORCE") == 0)
- part_msg.shared = SHARED_FORCE;
- else {
- exit_code = 1;
- fprintf (stderr, "Invalid input: %s\n",
- argv[i]);
- fprintf (stderr, "Acceptable Shared values "
- "are YES, NO and FORCE\n");
- return 0;
- }
+ else if (strncasecmp(argv[i], "MaxShare=", 9) == 0) {
+ part_msg.max_share = (uint16_t) strtol(&argv[i][9],
+ (char **) NULL, 10);
+ update_cnt++;
+ }
+ else if (strncasecmp(argv[i], "Priority=", 9) == 0) {
+ part_msg.priority = (uint16_t) strtol(&argv[i][9],
+ (char **) NULL, 10);
update_cnt++;
}
else if (strncasecmp(argv[i], "State=", 6) == 0) {
diff --git a/src/sinfo/opts.c b/src/sinfo/opts.c
index a01a349d487deb0b3c0ea1fd22c4978904217024..c91d81df385bc8dfbe69cb04b25717f994d0afa3 100644
--- a/src/sinfo/opts.c
+++ b/src/sinfo/opts.c
@@ -517,6 +517,12 @@ _parse_format( char* format )
field_size,
right_justify,
suffix );
+ } else if (field[0] == 'p') {
+ params.match_flags.priority_flag = true;
+ format_add_priority( params.format_list,
+ field_size,
+ right_justify,
+ suffix );
} else if (field[0] == 'P') {
params.match_flags.partition_flag = true;
format_add_partition( params.format_list,
@@ -692,6 +698,8 @@ void _print_options( void )
"true" : "false");
printf("partition_flag = %s\n", params.match_flags.partition_flag ?
"true" : "false");
+ printf("priority_flag = %s\n", params.match_flags.priority_flag ?
+ "true" : "false");
printf("reason_flag = %s\n", params.match_flags.reason_flag ?
"true" : "false");
printf("root_flag = %s\n", params.match_flags.root_flag ?
diff --git a/src/sinfo/print.c b/src/sinfo/print.c
index 9a235a17627ba10e43023c9d567ea0dd83854ef0..3f16b822267535627345ccb10f31b5b025c78b10 100644
--- a/src/sinfo/print.c
+++ b/src/sinfo/print.c
@@ -1,7 +1,7 @@
/*****************************************************************************\
* print.c - sinfo print job functions
*****************************************************************************
- * Copyright (C) 2002-2006 The Regents of the University of California.
+ * Copyright (C) 2002-2007 The Regents of the University of California.
* Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER).
* Written by Joey Ekstrom <ekstrom1@llnl.gov> and
* Morris Jette <jette1@llnl.gov>
@@ -170,7 +170,8 @@ static int _print_secs(long time, int width, bool right, bool cut_output)
}
static int
-_build_min_max_16_string(char *buffer, int buf_size, uint16_t min, uint16_t max, bool range)
+_build_min_max_16_string(char *buffer, int buf_size, uint16_t min, uint16_t max,
+ bool range)
{
char tmp_min[7];
char tmp_max[7];
@@ -191,7 +192,8 @@ _build_min_max_16_string(char *buffer, int buf_size, uint16_t min, uint16_t max,
}
static int
-_build_min_max_32_string(char *buffer, int buf_size, uint32_t min, uint32_t max, bool range)
+_build_min_max_32_string(char *buffer, int buf_size, uint32_t min, uint32_t max,
+ bool range)
{
char tmp_min[7];
char tmp_max[7];
@@ -631,6 +633,24 @@ int _print_prefix(sinfo_data_t * job, int width, bool right_justify,
return SLURM_SUCCESS;
}
+int _print_priority(sinfo_data_t * sinfo_data, int width,
+ bool right_justify, char *suffix)
+{
+ char id[FORMAT_STRING_SIZE];
+
+ if (sinfo_data) {
+ _build_min_max_16_string(id, FORMAT_STRING_SIZE,
+ sinfo_data->part_info->priority,
+ sinfo_data->part_info->priority, true);
+ _print_str(id, width, right_justify, true);
+ } else
+ _print_str("PRIORITY", width, right_justify, true);
+
+ if (suffix)
+ printf("%s", suffix);
+ return SLURM_SUCCESS;
+}
+
int _print_reason(sinfo_data_t * sinfo_data, int width,
bool right_justify, char *suffix)
{
@@ -668,17 +688,15 @@ int _print_root(sinfo_data_t * sinfo_data, int width,
int _print_share(sinfo_data_t * sinfo_data, int width,
bool right_justify, char *suffix)
{
+ char id[FORMAT_STRING_SIZE];
+
if (sinfo_data) {
- if (sinfo_data->part_info == NULL)
- _print_str("n/a", width, right_justify, true);
- else if (sinfo_data->part_info->shared > 1)
- _print_str("force", width, right_justify, true);
- else if (sinfo_data->part_info->shared)
- _print_str("yes", width, right_justify, true);
- else
- _print_str("no", width, right_justify, true);
+ _build_min_max_16_string(id, FORMAT_STRING_SIZE,
+ sinfo_data->part_info->max_share,
+ sinfo_data->part_info->max_share, true);
+ _print_str(id, width, right_justify, true);
} else
- _print_str("SHARE", width, right_justify, true);
+ _print_str("MAX_SHARE", width, right_justify, true);
if (suffix)
printf("%s", suffix);
diff --git a/src/sinfo/print.h b/src/sinfo/print.h
index 989f25b7913d1b44f3a891493f36c8ae6fd0a355..c2c2645717ee5bbc7cd3f5552ce63ac6697a398b 100644
--- a/src/sinfo/print.h
+++ b/src/sinfo/print.h
@@ -100,6 +100,8 @@ int print_sinfo_list(List sinfo_list);
format_add_function(list,wid,right,suffix,_print_partition)
#define format_add_prefix(list,wid,right,suffix) \
format_add_function(list,wid,right,suffix,_print_prefix)
+#define format_add_priority(list,wid,right,suffix) \
+ format_add_function(list,wid,right,suffix,_print_priority)
#define format_add_reason(list,wid,right,suffix) \
format_add_function(list,wid,right,suffix,_print_reason)
#define format_add_root(list,wid,right,prefix) \
@@ -155,6 +157,8 @@ int _print_partition(sinfo_data_t * sinfo_data, int width,
bool right_justify, char *suffix);
int _print_prefix(sinfo_data_t * sinfo_data, int width,
bool right_justify, char *suffix);
+int _print_priority(sinfo_data_t * sinfo_data, int width,
+ bool right_justify, char *suffix);
int _print_reason(sinfo_data_t * sinfo_data, int width,
bool right_justify, char *suffix);
int _print_root(sinfo_data_t * sinfo_data, int width,
diff --git a/src/sinfo/sinfo.c b/src/sinfo/sinfo.c
index 599c98f4245e3d8bc61ed510425acdf8ea6b7a1b..701de69021a7ab11deac513f824f5d7511965850 100644
--- a/src/sinfo/sinfo.c
+++ b/src/sinfo/sinfo.c
@@ -3,7 +3,7 @@
*
* $Id$
*****************************************************************************
- * Copyright (C) 2002-2006 The Regents of the University of California.
+ * Copyright (C) 2002-2007 The Regents of the University of California.
* Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER).
* Written by Joey Ekstrom <ekstrom1@llnl.gov>, Morris Jette <jette1@llnl.gov>
* UCRL-CODE-226842.
@@ -629,7 +629,11 @@ static bool _match_part_data(sinfo_data_t *sinfo_ptr,
return false;
if (params.match_flags.share_flag &&
- (part_ptr->shared != sinfo_ptr->part_info->shared))
+ (part_ptr->max_share != sinfo_ptr->part_info->max_share))
+ return false;
+
+ if (params.match_flags.priority_flag &&
+ (part_ptr->priority != sinfo_ptr->part_info->priority))
return false;
return true;
diff --git a/src/sinfo/sinfo.h b/src/sinfo/sinfo.h
index b1e6d3d1064fcb99380be5c5d54f882d9bdb2069..1c9212cfc9998c9de2feaffd9f1498deb8f7316a 100644
--- a/src/sinfo/sinfo.h
+++ b/src/sinfo/sinfo.h
@@ -105,7 +105,7 @@ typedef struct {
hostlist_t ionodes;
#endif
/* part_info contains partition, avail, max_time, job_size,
- * root, share, groups */
+ * root, share, groups, priority */
partition_info_t* part_info;
uint16_t part_inx;
} sinfo_data_t;
@@ -126,6 +126,7 @@ struct sinfo_match_flags {
bool max_time_flag;
bool memory_flag;
bool partition_flag;
+ bool priority_flag;
bool reason_flag;
bool root_flag;
bool share_flag;
diff --git a/src/sinfo/sort.c b/src/sinfo/sort.c
index 08303757f509a1cf1890b53e00d7a53681e3d82d..a98141dc05f3ee723f55279d181daec9be1cee07 100644
--- a/src/sinfo/sort.c
+++ b/src/sinfo/sort.c
@@ -63,6 +63,7 @@ static int _sort_by_node_list(void *void1, void *void2);
static int _sort_by_nodes_ai(void *void1, void *void2);
static int _sort_by_nodes(void *void1, void *void2);
static int _sort_by_partition(void *void1, void *void2);
+static int _sort_by_priority(void *void1, void *void2);
static int _sort_by_reason(void *void1, void *void2);
static int _sort_by_reason_time(void *void1, void *void2);
static int _sort_by_root(void *void1, void *void2);
@@ -122,6 +123,8 @@ void sort_sinfo_list(List sinfo_list)
list_sort(sinfo_list, _sort_by_memory);
else if (params.sort[i] == 'N')
list_sort(sinfo_list, _sort_by_node_list);
+ else if (params.sort[i] == 'p')
+ list_sort(sinfo_list, _sort_by_priority);
else if (params.sort[i] == 'P')
list_sort(sinfo_list, _sort_by_partition);
else if (params.sort[i] == 'r')
@@ -527,9 +530,27 @@ static int _sort_by_share(void *void1, void *void2)
int val1 = 0, val2 = 0;
if (sinfo1->part_info)
- val1 = sinfo1->part_info->shared;
+ val1 = sinfo1->part_info->max_share;
if (sinfo2->part_info)
- val2 = sinfo2->part_info->shared;
+ val2 = sinfo2->part_info->max_share;
+ diff = val1 - val2;
+
+ if (reverse_order)
+ diff = -diff;
+ return diff;
+}
+
+static int _sort_by_priority(void *void1, void *void2)
+{
+ int diff;
+ sinfo_data_t *sinfo1 = (sinfo_data_t *) void1;
+ sinfo_data_t *sinfo2 = (sinfo_data_t *) void2;
+ int val1 = 0, val2 = 0;
+
+ if (sinfo1->part_info)
+ val1 = sinfo1->part_info->priority;
+ if (sinfo2->part_info)
+ val2 = sinfo2->part_info->priority;
diff = val1 - val2;
if (reverse_order)
diff --git a/src/slurmctld/job_mgr.c b/src/slurmctld/job_mgr.c
index 70bee45dc9882f91d5366096a0d493f358612739..d53ee1dcb3048df824d81ea2a6d5f4660f89d6de 100644
--- a/src/slurmctld/job_mgr.c
+++ b/src/slurmctld/job_mgr.c
@@ -3315,9 +3315,9 @@ void reset_job_priority(void)
}
/*
- * _top_priority - determine if any other job for this partition has a
- * higher priority than specified job
- * IN job_ptr - pointer to selected partition
+ * _top_priority - determine if any other job has a higher priority than the
+ * specified job
+ * IN job_ptr - pointer to selected job
* RET true if selected job has highest priority
*/
static bool _top_priority(struct job_record *job_ptr)
@@ -3347,8 +3347,11 @@ static bool _top_priority(struct job_record *job_ptr)
continue;
if (!job_independent(job_ptr2))
continue;
- if ((job_ptr2->priority > job_ptr->priority) &&
- (job_ptr2->part_ptr == job_ptr->part_ptr)) {
+ if ((job_ptr2->part_ptr->priority >
+ job_ptr ->part_ptr->priority) ||
+ ((job_ptr2->part_ptr->priority ==
+ job_ptr ->part_ptr->priority) &&
+ (job_ptr2->priority > job_ptr->priority))) {
top = false;
break;
}
diff --git a/src/slurmctld/job_scheduler.c b/src/slurmctld/job_scheduler.c
index a56f8f99a90705ae12f68cfe068b30df53c7ba30..7dec58858fb18c67b2d1d2ba98acd88d0cb9e3cb 100644
--- a/src/slurmctld/job_scheduler.c
+++ b/src/slurmctld/job_scheduler.c
@@ -2,7 +2,7 @@
* job_scheduler.c - manage the scheduling of pending jobs in priority order
* Note there is a global job list (job_list)
*****************************************************************************
- * Copyright (C) 2002-2006 The Regents of the University of California.
+ * Copyright (C) 2002-2007 The Regents of the University of California.
* Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER).
* Written by Morris Jette <jette1@llnl.gov>
* UCRL-CODE-226842.
@@ -61,8 +61,9 @@
#define MAX_RETRIES 10
struct job_queue {
- int priority;
struct job_record *job_ptr;
+ uint32_t job_priority;
+ uint16_t part_priority;
};
static int _build_job_queue(struct job_queue **job_queue);
@@ -104,7 +105,9 @@ static int _build_job_queue(struct job_queue **job_queue)
sizeof(struct job_queue));
}
my_job_queue[job_queue_size].job_ptr = job_ptr;
- my_job_queue[job_queue_size].priority = job_ptr->priority;
+ my_job_queue[job_queue_size].job_priority = job_ptr->priority;
+ my_job_queue[job_queue_size].part_priority =
+ job_ptr->part_ptr->priority;
job_queue_size++;
}
list_iterator_destroy(job_iterator);
@@ -283,26 +286,42 @@ int schedule(void)
static void _sort_job_queue(struct job_queue *job_queue, int job_queue_size)
{
int i, j, top_prio_inx;
- int tmp_prio, top_prio;
struct job_record *tmp_job_ptr;
+ uint32_t top_job_prio, tmp_job_prio;
+ uint16_t top_part_prio, tmp_part_prio;
for (i = 0; i < job_queue_size; i++) {
- top_prio = job_queue[i].priority;
- top_prio_inx = i;
+ top_prio_inx = i;
+ top_job_prio = job_queue[i].job_priority;
+ top_part_prio = job_queue[i].part_priority;
+
for (j = (i + 1); j < job_queue_size; j++) {
- if (top_prio >= job_queue[j].priority)
+ if (top_part_prio > job_queue[j].part_priority)
+ continue;
+ if ((top_part_prio == job_queue[j].part_priority) &&
+ (top_job_prio >= job_queue[j].job_priority))
continue;
- top_prio = job_queue[j].priority;
- top_prio_inx = j;
+
+ top_prio_inx = j;
+ top_job_prio = job_queue[j].job_priority;
+ top_part_prio = job_queue[j].part_priority;
}
if (top_prio_inx == i)
- continue;
- tmp_prio = job_queue[i].priority;
- tmp_job_ptr = job_queue[i].job_ptr;
- job_queue[i].priority = job_queue[top_prio_inx].priority;
- job_queue[i].job_ptr = job_queue[top_prio_inx].job_ptr;
- job_queue[top_prio_inx].priority = tmp_prio;
- job_queue[top_prio_inx].job_ptr = tmp_job_ptr;
+ continue; /* in correct order */
+
+ /* swap records at top_prio_inx and i */
+ tmp_job_ptr = job_queue[i].job_ptr;
+ tmp_job_prio = job_queue[i].job_priority;
+ tmp_part_prio = job_queue[i].part_priority;
+
+ job_queue[i].job_ptr = job_queue[top_prio_inx].job_ptr;
+ job_queue[i].job_priority = job_queue[top_prio_inx].job_priority;
+ job_queue[i].part_priority = job_queue[top_prio_inx].part_priority;
+
+ job_queue[top_prio_inx].job_ptr = tmp_job_ptr;
+ job_queue[top_prio_inx].job_priority = tmp_job_prio;
+ job_queue[top_prio_inx].part_priority = tmp_part_prio;
+
}
}
diff --git a/src/slurmctld/node_scheduler.c b/src/slurmctld/node_scheduler.c
index 53353434cbea44f85fc3e28d60bfd7d5f81126fe..690660987a42bcaf89fe8f8da784c2dd5969e54e 100644
--- a/src/slurmctld/node_scheduler.c
+++ b/src/slurmctld/node_scheduler.c
@@ -351,13 +351,13 @@ _job_count_bitmap(bitstr_t * bitmap, bitstr_t * jobmap, int job_cnt)
*
* IN user_flag - may be 0 (do not share nodes), 1 (node sharing allowed),
* or any other number means "don't care"
- * IN part_enum - current partition's node sharing policy
+ * IN part_max_share - current partition's node sharing policy
* IN cons_res_flag - 1 if the consumable resources flag is enable, 0 otherwise
*
* RET - 1 if nodes can be shared, 0 if nodes cannot be shared
*/
static int
-_resolve_shared_status(uint16_t user_flag, uint16_t part_enum,
+_resolve_shared_status(uint16_t user_flag, uint16_t part_max_share,
int cons_res_flag)
{
int shared;
@@ -368,7 +368,7 @@ _resolve_shared_status(uint16_t user_flag, uint16_t part_enum,
* the partition or user has to explicitly disable sharing to
* get exclusive nodes.
*/
- if ((part_enum == SHARED_EXCLUSIVE) || (user_flag == 0))
+ if ((part_max_share == 0) || (user_flag == 0))
shared = 0;
else
shared = 1;
@@ -376,9 +376,9 @@ _resolve_shared_status(uint16_t user_flag, uint16_t part_enum,
/* The partition sharing option is only used if
* the consumable resources plugin is NOT in use.
*/
- if (part_enum == SHARED_FORCE) /* shared=force */
+ if (part_max_share == (uint16_t) INFINITE) /* shared=force */
shared = 1;
- else if (part_enum == SHARED_NO) /* can't share */
+ else if (part_max_share <= 1) /* can't share */
shared = 0;
else
shared = (user_flag == 1) ? 1 : 0;
@@ -592,7 +592,7 @@ _pick_best_nodes(struct node_set *node_set_ptr, int node_set_size,
return error_code;
shared = _resolve_shared_status(job_ptr->details->shared,
- part_ptr->shared, cr_enabled);
+ part_ptr->max_share, cr_enabled);
job_ptr->details->shared = shared;
if (cr_enabled) {
diff --git a/src/slurmctld/partition_mgr.c b/src/slurmctld/partition_mgr.c
index 86035697615490e15f532b063430c837ca2d3284..78e6718ca373c884eb87a7cf039488bcddf18068 100644
--- a/src/slurmctld/partition_mgr.c
+++ b/src/slurmctld/partition_mgr.c
@@ -215,13 +215,14 @@ struct part_record *create_part_record(void)
xassert (part_ptr->magic = PART_MAGIC); /* set value */
strcpy(part_ptr->name, "DEFAULT");
- part_ptr->hidden = default_part.hidden;
+ part_ptr->hidden = default_part.hidden;
part_ptr->max_time = default_part.max_time;
part_ptr->max_nodes = default_part.max_nodes;
part_ptr->min_nodes = default_part.min_nodes;
part_ptr->root_only = default_part.root_only;
part_ptr->state_up = default_part.state_up;
- part_ptr->shared = default_part.shared;
+ part_ptr->max_share = default_part.max_share;
+ part_ptr->priority = default_part.priority;
part_ptr->node_bitmap = NULL;
if (default_part.allow_groups)
@@ -362,19 +363,20 @@ static void _dump_part_state(struct part_record *part_ptr, Buf buffer)
else
default_part_flag = 0;
- packstr(part_ptr->name, buffer);
- pack32(part_ptr->max_time, buffer);
+ packstr(part_ptr->name, buffer);
+ pack32(part_ptr->max_time, buffer);
pack32(part_ptr->max_nodes, buffer);
pack32(part_ptr->min_nodes, buffer);
- pack16(default_part_flag, buffer);
- pack16(part_ptr->hidden, buffer);
+ pack16(default_part_flag, buffer);
+ pack16(part_ptr->hidden, buffer);
pack16(part_ptr->root_only, buffer);
- pack16(part_ptr->shared, buffer);
+ pack16(part_ptr->max_share, buffer);
+ pack16(part_ptr->priority, buffer);
- pack16(part_ptr->state_up, buffer);
+ pack16(part_ptr->state_up, buffer);
packstr(part_ptr->allow_groups, buffer);
- packstr(part_ptr->nodes, buffer);
+ packstr(part_ptr->nodes, buffer);
}
/*
@@ -388,7 +390,8 @@ int load_all_part_state(void)
char *part_name, *allow_groups, *nodes, *state_file, *data = NULL;
uint32_t max_time, max_nodes, min_nodes;
time_t time;
- uint16_t name_len, def_part_flag, hidden, root_only, shared, state_up;
+ uint16_t name_len, def_part_flag, hidden, root_only;
+ uint16_t max_share, priority, state_up;
struct part_record *part_ptr;
uint32_t data_size = 0;
int data_allocated, data_read = 0, error_code = 0, part_cnt = 0;
@@ -439,9 +442,10 @@ int load_all_part_state(void)
safe_unpack32(&min_nodes, buffer);
safe_unpack16(&def_part_flag, buffer);
- safe_unpack16(&hidden, buffer);
+ safe_unpack16(&hidden, buffer);
safe_unpack16(&root_only, buffer);
- safe_unpack16(&shared, buffer);
+ safe_unpack16(&max_share, buffer);
+ safe_unpack16(&priority, buffer);
safe_unpack16(&state_up, buffer);
safe_unpackstr_xmalloc(&allow_groups, &name_len, buffer);
@@ -450,10 +454,10 @@ int load_all_part_state(void)
/* validity test as possible */
if ((def_part_flag > 1) ||
(root_only > 1) || (hidden > 1) ||
- (shared > SHARED_EXCLUSIVE) || (state_up > 1)) {
+ (state_up > 1)) {
error("Invalid data for partition %s: def_part_flag=%u, "
- "hidden=%u root_only=%u, shared=%u, state_up=%u",
- part_name, def_part_flag, hidden, root_only, shared,
+ "hidden=%u root_only=%u, state_up=%u",
+ part_name, def_part_flag, hidden, root_only,
state_up);
error("No more partition data will be processed from "
"the checkpoint file");
@@ -477,8 +481,9 @@ int load_all_part_state(void)
default_part_loc = part_ptr;
}
part_ptr->root_only = root_only;
- part_ptr->shared = shared;
- part_ptr->state_up = state_up;
+ part_ptr->max_share = max_share;
+ part_ptr->priority = priority;
+ part_ptr->state_up = state_up;
xfree(part_ptr->allow_groups);
part_ptr->allow_groups = allow_groups;
xfree(part_ptr->nodes);
@@ -533,7 +538,8 @@ int init_part_conf(void)
default_part.min_nodes = 1;
default_part.root_only = 0;
default_part.state_up = 1;
- default_part.shared = SHARED_NO;
+ default_part.max_share = 1;
+ default_part.priority = 1;
default_part.total_nodes = 0;
default_part.total_cpus = 0;
xfree(default_part.nodes);
@@ -732,10 +738,11 @@ void pack_part(struct part_record *part_ptr, Buf buffer)
&node_scaling);
pack16(node_scaling, buffer);
pack32(part_ptr->total_cpus, buffer);
- pack16(default_part_flag, buffer);
- pack16(part_ptr->hidden, buffer);
- pack16(part_ptr->root_only, buffer);
- pack16(part_ptr->shared, buffer);
+ pack16(default_part_flag, buffer);
+ pack16(part_ptr->hidden, buffer);
+ pack16(part_ptr->root_only, buffer);
+ pack16(part_ptr->max_share, buffer);
+ pack16(part_ptr->priority, buffer);
pack16(part_ptr->state_up, buffer);
packstr(part_ptr->allow_groups, buffer);
@@ -782,8 +789,8 @@ int update_part(update_part_msg_t * part_desc)
last_part_update = time(NULL);
if (part_desc->hidden != (uint16_t) NO_VAL) {
- info("update_part: setting hidden to %u for partition %s",
- part_desc->hidden, part_desc->name);
+ info("update_part: setting hidden to %u for partition %s",
+ part_desc->hidden, part_desc->name);
part_ptr->hidden = part_desc->hidden;
}
@@ -817,10 +824,16 @@ int update_part(update_part_msg_t * part_desc)
part_ptr->state_up = part_desc->state_up;
}
- if (part_desc->shared != (uint16_t) NO_VAL) {
- info("update_part: setting shared to %u for partition %s",
- part_desc->shared, part_desc->name);
- part_ptr->shared = part_desc->shared;
+ if (part_desc->max_share != (uint16_t) NO_VAL) {
+ info("update_part: setting max_share to %u for partition %s",
+ part_desc->max_share, part_desc->name);
+ part_ptr->max_share = part_desc->max_share;
+ }
+
+ if (part_desc->priority != (uint16_t) NO_VAL) {
+ info("update_part: setting priority to %u for partition %s",
+ part_desc->priority, part_desc->name);
+ part_ptr->priority = part_desc->priority;
}
if ((part_desc->default_part == 1) &&
diff --git a/src/slurmctld/proc_req.c b/src/slurmctld/proc_req.c
index e56c452a8b9e975b40b60639a87b15611c3e57b3..3244da8f3c6bcf60aa63074285a237fc61d61a39 100644
--- a/src/slurmctld/proc_req.c
+++ b/src/slurmctld/proc_req.c
@@ -356,6 +356,7 @@ void _fill_ctld_conf(slurm_ctl_conf_t * conf_ptr)
conf_ptr->ret2service = conf->ret2service;
conf_ptr->schedport = conf->schedport;
conf_ptr->schedrootfltr = conf->schedrootfltr;
+ conf_ptr->sched_time_slice = conf->sched_time_slice;
conf_ptr->schedtype = xstrdup(conf->schedtype);
conf_ptr->select_type = xstrdup(conf->select_type);
conf_ptr->select_type_param = conf->select_type_param;
diff --git a/src/slurmctld/read_config.c b/src/slurmctld/read_config.c
index f626e87cf12ba05326db2b1e2090439ee11b37a2..394e90ff0095b3294da6e875c68e2b320f06fc83 100644
--- a/src/slurmctld/read_config.c
+++ b/src/slurmctld/read_config.c
@@ -607,11 +607,12 @@ static int _build_single_partitionline_info(slurm_conf_partition_t *part)
}
part_ptr->hidden = part->hidden_flag ? 1 : 0;
part_ptr->max_time = part->max_time;
+ part_ptr->max_share = part->max_share;
part_ptr->max_nodes = part->max_nodes;
part_ptr->min_nodes = part->min_nodes;
+ part_ptr->priority = part->priority;
part_ptr->root_only = part->root_only_flag ? 1 : 0;
part_ptr->state_up = part->state_up_flag ? 1 : 0;
- part_ptr->shared = part->shared;
if (part->allow_groups) {
xfree(part_ptr->allow_groups);
part_ptr->allow_groups = xstrdup(part->allow_groups);
diff --git a/src/slurmctld/slurmctld.h b/src/slurmctld/slurmctld.h
index 756593eaa28f9d29dd65d8d2d453d19cba040cfa..d96eb033c4647179706da161600fa6ad6cc298e8 100644
--- a/src/slurmctld/slurmctld.h
+++ b/src/slurmctld/slurmctld.h
@@ -242,7 +242,8 @@ struct part_record {
uint32_t max_offset; /* select plugin max offset */
uint16_t root_only; /* 1 if allocate/submit RPC can only be
issued by user root */
- uint16_t shared; /* See part_shared in slurm.h */
+ uint16_t max_share; /* number of jobs to gang schedule */
+ uint16_t priority; /* scheduling priority for jobs */
uint16_t state_up; /* 1 if state is up, 0 if down */
char *nodes; /* comma delimited list names of nodes */
char *allow_groups; /* comma delimited list of groups,