diff --git a/META b/META
index aef01630c7ce89b4abcef4a3b264c89a26cd79c4..d9d11cc0e5c9e0efa9ed81e75338f82b856c7c75 100644
--- a/META
+++ b/META
@@ -3,9 +3,9 @@
Api_revision: 0
Major: 2
Meta: 1
- Micro: 9
+ Micro: 10
Minor: 1
Name: slurm
Release: 1
Release_tags: dist
- Version: 2.1.9
+ Version: 2.1.10
diff --git a/NEWS b/NEWS
index 6e070ec89239bca28ae108403926626790a46008..bb56de769ba045a01b91a06df5edfb1cfc363bcf 100644
--- a/NEWS
+++ b/NEWS
@@ -1,6 +1,65 @@
This file describes changes in recent versions of SLURM. It primarily
documents those changes that are of interest to users and admins.
+
+* Changes in SLURM 2.1.11
+=========================
+
+* Changes in SLURM 2.1.10
+=========================
+ -- Fix memory leak in sched/builtin plugin.
+ -- Fixed sbatch to work correctly when no nodes are specified, but
+ --ntasks-per-node is.
+ -- Make sure account and wckey for a job are lower case before inserting into
+ accounting.
+ -- Added note to squeue documentation about --jobs option displaying jobs
+ even if they are on hidden partitions.
+ -- Fix srun to work correctly with --uid when getting an allocation
+ and creating a step.
+ -- Fix for when removing a limit from a users association inside the
+ fairshare tree the parents limit is now inherited automatically in
+ the slurmctld. Previously the slurmctld would have to be restarted.
+ This problem only exists when setting a users association limit to -1.
+ -- Patch from Matthieu Hautreux (CEA) dealing with possible overflows that
+ could come up with the select/cons_res plugin with uint32_t's being treated
+ as uint16_t's.
+ -- Correct logic for creating a reservation with a Duration=Infinite (used to
+ set reservation end time in the past).
+ -- Correct logic for creating a reservation that properly handles the OVERLAP
+ and IGNORE_JOBS flags (flags were ignored under some conditions).
+ -- Fixed a fair-share calculation bug in the priority/multifactor plugin.
+ -- Make sure a user entry in the database that was previously deleted is
+ restored clean when added back, i.e. remove admin privileges previously
+ given.
+ -- BLUEGENE - Future start time is set correctly when eligible time for a job
+ is in the future, but the job can physically run earlier.
+ -- Updated Documentation for sacctmgr for Wall and CPUMin options stating when
+ the limit is reached running jobs will be killed.
+ -- Fix deadlock issue in the slurmctld when lowering limits in accounting to
+ lower than that of pending jobs.
+ -- Fix bug in salloc, sbatch and srun that could under some conditions process
+ the --threads-per-core, --cores-per-socket and --sockets-per-node options
+ improperly.
+ -- Fix bug in select/cons_res with memory management plus job preemption with
+ job removal (e.g. requeue) which under some conditions failed to preempt
+ jobs.
+ -- Fix deadlock potential when using qos and associations in the slurmctld.
+ -- Update documentation to state --ntasks-per-* is for a maximum value
+ instead of an absolute.
+ -- Get ReturnToService=2 working for front-end configurations (e.g. Cray or
+ BlueGene).
+ -- Fix issue when deleting a QOS on a pending/running job which previously
+ would create possible memory corruption.
+ -- Do not make a non-responding node available for use after running
+ "scontrol update nodename=<name> state=resume". Wait for node to respond
+ before use.
+ -- Added slurm_xlator.h to jobacct_gather plugins so they resolve symbols
+ correctly when linking to the slurm api.
+ -- You can now update a jobs QOS from scontrol. Previously you could only do
+ this from sview.
+ -- BLUEGENE - Fixed bug where if running in non-dynamic mode sometimes the
+ start time returned for a job when using test-only would not be correct.
+
* Changes in SLURM 2.1.9
========================
-- In select/linear - Fix logic to prevent over-subscribing memory with shared
@@ -36,7 +95,7 @@ documents those changes that are of interest to users and admins.
slurmctld's running for a single cluster, which should rarely if ever
happen.
-- Fixed sacct -c option.
- -- Critical bug fix in sched/backfill plugin that caused in memory corruption.
+ -- Critical bug fix in sched/backfill plugin that caused memory corruption.
* Changes in SLURM 2.1.8
========================
@@ -5074,4 +5133,4 @@ 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 20444 2010-06-08 22:20:32Z jette $
+$Id: NEWS 20750 2010-07-16 21:32:31Z da $
diff --git a/doc/man/man1/sacctmgr.1 b/doc/man/man1/sacctmgr.1
index 7431f49c81335c1ba04e680e1c4880035e926b7f..f908de3bb8c62f81332726bdc4229393ab3c97fa 100644
--- a/doc/man/man1/sacctmgr.1
+++ b/doc/man/man1/sacctmgr.1
@@ -223,15 +223,16 @@ Maximum number of CPU minutes running jobs are able to be allocated in
aggregate for this association and all association which are children
of this association.
To clear a previously set value use the modify command with a new
-value of \-1. (NOTE: This limit is not enforced if set on the root
+value of \-1.
+NOTE: This limit is not enforced if set on the root
association of a cluster. So even though it may appear in sacctmgr
output it will not be enforced.
NOTE: This limit only applys when using the Priority Multifactor
plugin. The time is decayed using the value of PriorityDecayHalfLife
-or PriorityUsageResetPeriod as set in the slurm.conf. Currently when
-this limit is reached jobs will be delayed until they are able to run
-inside the limit. No jobs will be killed if this limit is reached,
-this will change in future versions of SLURM.)
+or PriorityUsageResetPeriod as set in the slurm.conf. When this limit
+is reached all associated jobs running will be killed and all future jobs
+submitted with associations in the group will be delayed until they
+are able to run inside the limit.
.TP
\fIGrpCPUs\fP=<max cpus>
@@ -266,15 +267,15 @@ To clear a previously set value use the modify command with a new value of \-1.
Maximum wall clock time running jobs are able to be allocated in aggregate for
this association and all association which are children of this association.
To clear a previously set value use the modify command with a new value of \-1.
-(NOTE: This limit is not enforced if set on the root
+NOTE: This limit is not enforced if set on the root
association of a cluster. So even though it may appear in sacctmgr
output it will not be enforced.
NOTE: This limit only applys when using the Priority Multifactor
plugin. The time is decayed using the value of PriorityDecayHalfLife
-or PriorityUsageResetPeriod as set in the slurm.conf. Currently when
-this limit is reached jobs will be delayed until they are able to run
-inside the limit. No jobs will be killed if this limit is reached,
-this will change in future versions of SLURM.)
+or PriorityUsageResetPeriod as set in the slurm.conf. When this limit
+is reached all associated jobs running will be killed and all future jobs
+submitted with associations in the group will be delayed until they
+are able to run inside the limit.
.TP
\fIMaxCPUMins\fP=<max cpu minutes>
@@ -743,10 +744,10 @@ To clear a previously set value use the modify command with a new
value of \-1.
NOTE: This limit only applys when using the Priority Multifactor
plugin. The time is decayed using the value of PriorityDecayHalfLife
-or PriorityUsageResetPeriod as set in the slurm.conf. Currently when
-this limit is reached jobs will be delayed until they are able to run
-inside the limit. No jobs will be killed if this limit is reached,
-this will change in future versions of SLURM.)
+or PriorityUsageResetPeriod as set in the slurm.conf. When this limit
+is reached all associated jobs running will be killed and all future jobs
+submitted with this QOS will be delayed until they are able to run
+inside the limit.
.TP
\fIGrpCPUs\fP
@@ -781,10 +782,10 @@ this QOS.
To clear a previously set value use the modify command with a new value of \-1.
NOTE: This limit only applys when using the Priority Multifactor
plugin. The time is decayed using the value of PriorityDecayHalfLife
-or PriorityUsageResetPeriod as set in the slurm.conf. Currently when
-this limit is reached jobs will be delayed until they are able to run
-inside the limit. No jobs will be killed if this limit is reached,
-this will change in future versions of SLURM.)
+or PriorityUsageResetPeriod as set in the slurm.conf. When this limit
+is reached all associated jobs running will be killed and all future jobs
+submitted with this QOS will be delayed until they are able to run
+inside the limit.
.TP
\fIMaxCPUMins\fP
diff --git a/doc/man/man1/salloc.1 b/doc/man/man1/salloc.1
index 414d438181413483dcc276b75eedc0d47ab183cc..87c25b64b8cfb3db1b1b3278cb64e7ba2b734220 100644
--- a/doc/man/man1/salloc.1
+++ b/doc/man/man1/salloc.1
@@ -657,7 +657,7 @@ ignored if \fISchedulerType=sched/wiki\fR or
.TP
\fB\-\-ntasks\-per\-core\fR=<\fIntasks\fR>
-Request that \fIntasks\fR be invoked on each core.
+Request the maximum \fIntasks\fR be invoked on each core.
Meant to be used with the \fB\-\-ntasks\fR option.
Related to \fB\-\-ntasks\-per\-node\fR except at the core level
instead of the node level. Masks will automatically be generated
@@ -669,7 +669,7 @@ NOTE: This option is not supported unless
.TP
\fB\-\-ntasks\-per\-socket\fR=<\fIntasks\fR>
-Request that \fIntasks\fR be invoked on each socket.
+Request the maximum \fIntasks\fR be invoked on each socket.
Meant to be used with the \fB\-\-ntasks\fR option.
Related to \fB\-\-ntasks\-per\-node\fR except at the socket level
instead of the node level. Masks will automatically be generated
@@ -681,7 +681,7 @@ NOTE: This option is not supported unless
.TP
\fB\-\-ntasks\-per\-node\fR=<\fIntasks\fR>
-Request that \fIntasks\fR be invoked on each node.
+Request the maximum \fIntasks\fR be invoked on each node.
Meant to be used with the \fB\-\-nodes\fR option.
This is related to \fB\-\-cpus\-per\-task\fR=\fIncpus\fR,
but does not require knowledge of the actual number of cpus on
diff --git a/doc/man/man1/sbatch.1 b/doc/man/man1/sbatch.1
index c62b53894694dd48faf047c40a2bd58cc175ce67..3b68213542f048566b410baa5074542eda1be2f3 100644
--- a/doc/man/man1/sbatch.1
+++ b/doc/man/man1/sbatch.1
@@ -696,7 +696,7 @@ behavior on the cluster.
.TP
\fB\-\-ntasks\-per\-core\fR=<\fIntasks\fR>
-Request that \fIntasks\fR be invoked on each core.
+Request the maximum \fIntasks\fR be invoked on each core.
Meant to be used with the \fB\-\-ntasks\fR option.
Related to \fB\-\-ntasks\-per\-node\fR except at the core level
instead of the node level. Masks will automatically be generated
@@ -708,7 +708,7 @@ NOTE: This option is not supported unless
.TP
\fB\-\-ntasks\-per\-socket\fR=<\fIntasks\fR>
-Request that \fIntasks\fR be invoked on each socket.
+Request the maximum \fIntasks\fR be invoked on each socket.
Meant to be used with the \fB\-\-ntasks\fR option.
Related to \fB\-\-ntasks\-per\-node\fR except at the socket level
instead of the node level. Masks will automatically be generated
@@ -720,7 +720,7 @@ NOTE: This option is not supported unless
.TP
\fB\-\-ntasks\-per\-node\fR=<\fIntasks\fR>
-Request that \fIntasks\fR be invoked on each node.
+Request the maximum \fIntasks\fR be invoked on each node.
Meant to be used with the \fB\-\-nodes\fR option.
This is related to \fB\-\-cpus\-per\-task\fR=\fIncpus\fR,
but does not require knowledge of the actual number of cpus on
diff --git a/doc/man/man1/squeue.1 b/doc/man/man1/squeue.1
index d528284d20acaa63c1dd88995c475fe46f73abf8..8b7adc7bd422c7ed12438a7a2cdd5bc83637fdef 100644
--- a/doc/man/man1/squeue.1
+++ b/doc/man/man1/squeue.1
@@ -48,6 +48,8 @@ By default, prints a time stamp with the header.
Requests a comma separated list of job ids to display. Defaults to all jobs.
The \fB\-\-jobs=<job_id_list>\fR option may be used in conjunction with the
\fB\-\-steps\fR option to print step information about specific jobs.
+Note: If a list of job ids is provided, the jobs are displayed even if
+they are on hidden partitions.
.TP
\fB\-l\fR, \fB\-\-long\fR
diff --git a/doc/man/man1/srun.1 b/doc/man/man1/srun.1
index 9524159ef7416294a22f0a9a36f6aba55e3e94d3..100f124599f5b0f09a400af163b9c42ab9f9a1e4 100644
--- a/doc/man/man1/srun.1
+++ b/doc/man/man1/srun.1
@@ -759,7 +759,7 @@ ignored if \fISchedulerType=sched/wiki\fR or
.TP
\fB\-\-ntasks\-per\-core\fR=<\fIntasks\fR>
-Request that \fIntasks\fR be invoked on each core.
+Request the maximum \fIntasks\fR be invoked on each core.
Meant to be used with the \fB\-\-ntasks\fR option.
Related to \fB\-\-ntasks\-per\-node\fR except at the core level
instead of the node level. Masks will automatically be generated
@@ -771,7 +771,7 @@ NOTE: This option is not supported unless
.TP
\fB\-\-ntasks\-per\-socket\fR=<\fIntasks\fR>
-Request that \fIntasks\fR be invoked on each socket.
+Request the maximum \fIntasks\fR be invoked on each socket.
Meant to be used with the \fB\-\-ntasks\fR option.
Related to \fB\-\-ntasks\-per\-node\fR except at the socket level
instead of the node level. Masks will automatically be generated
@@ -783,7 +783,7 @@ NOTE: This option is not supported unless
.TP
\fB\-\-ntasks\-per\-node\fR=<\fIntasks\fR>
-Request that \fIntasks\fR be invoked on each node.
+Request the maximum \fIntasks\fR be invoked on each node.
Meant to be used with the \fB\-\-nodes\fR option.
This is related to \fB\-\-cpus\-per\-task\fR=\fIncpus\fR,
but does not require knowledge of the actual number of cpus on
@@ -928,7 +928,7 @@ with the \fB\-\-relative\fR option, a warning message will be
printed and the \fB\-\-relative\fR option will take precedence.
.TP
-\fB\-\-resv-ports\fR
+\fB\-\-resv\-ports\fR
Reserve communication ports for this job.
Used for OpenMPI.
@@ -1415,6 +1415,9 @@ Same as \fB\-l, \-\-label\fR
\fBSLURM_MEM_BIND\fR
Same as \fB\-\-mem_bind\fR
.TP
+\fBSLURM_MPI_TYPE\fR
+Same as \fB\-\-mpi\fR
+.TP
\fBSLURM_NETWORK\fR
Same as \fB\-\-network\fR
.TP
@@ -1457,6 +1460,9 @@ Same as \fB\-D, \-\-chdir=\fR
\fBSLURM_RESTART_DIR\fR
Same as \fB\-\-restart\-dir\fR
.TP
+\fBSLURM_RESV_PORTS\fR
+Same as \fB\-\-resv\-ports\fR
+.TP
\fBSLURM_SIGNAL\fR
Same as \fB\-\-signal\fR
.TP
diff --git a/slurm.spec b/slurm.spec
index 40eb411235de739f3309577b550ea8d210eeccc2..dd82601f3c64fb835b8b6177cfd815e1c0650035 100644
--- a/slurm.spec
+++ b/slurm.spec
@@ -83,14 +83,14 @@
%endif
Name: slurm
-Version: 2.1.9
+Version: 2.1.10
Release: 1%{?dist}
Summary: Simple Linux Utility for Resource Management
License: GPL
Group: System Environment/Base
-Source: slurm-2.1.9.tar.bz2
+Source: slurm-2.1.10.tar.bz2
BuildRoot: %{_tmppath}/%{name}-%{version}-%{release}
URL: https://computing.llnl.gov/linux/slurm/
@@ -352,7 +352,7 @@ Gives the ability for SLURM to use Berkeley Lab Checkpoint/Restart
#############################################################################
%prep
-%setup -n slurm-2.1.9
+%setup -n slurm-2.1.10
%build
%configure --program-prefix=%{?_program_prefix:%{_program_prefix}} \
diff --git a/src/common/assoc_mgr.c b/src/common/assoc_mgr.c
index 6a8ac27e018da988248cc23fec2218162380d39c..96cd17121d81c7825ce7264442ea6b578965597e 100644
--- a/src/common/assoc_mgr.c
+++ b/src/common/assoc_mgr.c
@@ -61,6 +61,7 @@ static char *assoc_mgr_cluster_name = NULL;
static int setup_childern = 0;
void (*remove_assoc_notify) (acct_association_rec_t *rec) = NULL;
+void (*remove_qos_notify) (acct_qos_rec_t *rec) = NULL;
pthread_mutex_t assoc_mgr_association_lock = PTHREAD_MUTEX_INITIALIZER;
@@ -802,6 +803,8 @@ extern int assoc_mgr_init(void *db_conn, assoc_init_args_t *args)
enforce = args->enforce;
if(args->remove_assoc_notify)
remove_assoc_notify = args->remove_assoc_notify;
+ if(args->remove_qos_notify)
+ remove_qos_notify = args->remove_qos_notify;
cache_level = args->cache_level;
assoc_mgr_refresh_lists(db_conn, args);
}
@@ -1685,6 +1688,10 @@ extern int assoc_mgr_update_assocs(acct_update_object_t *update)
if(!assoc_mgr_association_list)
return SLURM_SUCCESS;
+ /* Since we could possibly need the qos lock handle it
+ now to avoid deadlock. Always do QOS first.
+ */
+ slurm_mutex_lock(&assoc_mgr_qos_lock);
slurm_mutex_lock(&assoc_mgr_association_lock);
itr = list_iterator_create(assoc_mgr_association_list);
while((object = list_pop(update->objects))) {
@@ -1814,7 +1821,11 @@ extern int assoc_mgr_update_assocs(acct_update_object_t *update)
rec->valid_qos);
rec->valid_qos =
bit_alloc(g_qos_count);
- }
+ } else
+ bit_nclear(rec->valid_qos, 0,
+ (bit_size(rec->
+ valid_qos)
+ - 1));
set_qos_bitstr_from_list(
rec->valid_qos, rec->qos_list);
}
@@ -1822,9 +1833,7 @@ extern int assoc_mgr_update_assocs(acct_update_object_t *update)
if(!slurmdbd_conf && !parents_changed) {
debug("updating assoc %u", rec->id);
- slurm_mutex_lock(&assoc_mgr_qos_lock);
log_assoc_rec(rec, assoc_mgr_qos_list);
- slurm_mutex_unlock(&assoc_mgr_qos_lock);
}
break;
case ACCT_ADD_ASSOC:
@@ -1942,18 +1951,17 @@ extern int assoc_mgr_update_assocs(acct_update_object_t *update)
}
if(setup_childern) {
/* Now normalize the static shares */
- slurm_mutex_lock(&assoc_mgr_qos_lock);
list_iterator_reset(itr);
while((object = list_next(itr))) {
_normalize_assoc_shares(object);
log_assoc_rec(object, assoc_mgr_qos_list);
}
- slurm_mutex_unlock(&assoc_mgr_qos_lock);
}
}
list_iterator_destroy(itr);
slurm_mutex_unlock(&assoc_mgr_association_lock);
+ slurm_mutex_unlock(&assoc_mgr_qos_lock);
/* This needs to happen outside of the
assoc_mgr_association_lock */
@@ -2173,11 +2181,16 @@ extern int assoc_mgr_update_qos(acct_update_object_t *update)
acct_association_rec_t *assoc = NULL;
int rc = SLURM_SUCCESS;
bool resize_qos_bitstr = 0;
+ List remove_list = NULL;
if(!assoc_mgr_qos_list)
return SLURM_SUCCESS;
slurm_mutex_lock(&assoc_mgr_qos_lock);
+ /* Since we could possibly need the association lock handle it
+ now to avoid deadlock. Always do QOS first.
+ */
+ slurm_mutex_lock(&assoc_mgr_association_lock);
itr = list_iterator_create(assoc_mgr_qos_list);
while((object = list_pop(update->objects))) {
list_iterator_reset(itr);
@@ -2269,7 +2282,22 @@ extern int assoc_mgr_update_qos(acct_update_object_t *update)
break;
case ACCT_REMOVE_QOS:
- if(rec)
+ if(!rec) {
+ //rc = SLURM_ERROR;
+ break;
+ }
+
+ if(remove_qos_notify) {
+ /* since there are some deadlock
+ issues while inside our lock here
+ we have to process a notify later
+ */
+ if(!remove_list)
+ remove_list = list_create(
+ destroy_acct_qos_rec);
+ list_remove(itr);
+ list_append(remove_list, rec);
+ } else
list_delete_item(itr);
if(!assoc_mgr_association_list)
@@ -2277,7 +2305,6 @@ extern int assoc_mgr_update_qos(acct_update_object_t *update)
/* Remove this qos from all the associations
on this cluster.
*/
- slurm_mutex_lock(&assoc_mgr_association_lock);
assoc_itr = list_iterator_create(
assoc_mgr_association_list);
while((assoc = list_next(assoc_itr))) {
@@ -2288,7 +2315,6 @@ extern int assoc_mgr_update_qos(acct_update_object_t *update)
bit_clear(assoc->valid_qos, object->id);
}
list_iterator_destroy(assoc_itr);
- slurm_mutex_unlock(&assoc_mgr_association_lock);
break;
default:
@@ -2310,7 +2336,6 @@ extern int assoc_mgr_update_qos(acct_update_object_t *update)
g_qos_count);
}
if(assoc_mgr_association_list) {
- slurm_mutex_lock(&assoc_mgr_association_lock);
assoc_itr = list_iterator_create(
assoc_mgr_association_list);
while((assoc = list_next(assoc_itr))) {
@@ -2321,13 +2346,25 @@ extern int assoc_mgr_update_qos(acct_update_object_t *update)
g_qos_count);
}
list_iterator_destroy(assoc_itr);
- slurm_mutex_unlock(&assoc_mgr_association_lock);
}
}
list_iterator_destroy(itr);
+ slurm_mutex_unlock(&assoc_mgr_association_lock);
slurm_mutex_unlock(&assoc_mgr_qos_lock);
+ /* This needs to happen outside of the
+ assoc_mgr_association_lock */
+ if(remove_list) {
+ itr = list_iterator_create(remove_list);
+
+ while((rec = list_next(itr)))
+ remove_qos_notify(rec);
+
+ list_iterator_destroy(itr);
+ list_destroy(remove_list);
+ }
+
return rc;
}
@@ -2708,13 +2745,16 @@ extern int load_assoc_usage(char *state_save_location)
if(assoc->id == assoc_id)
break;
+ /* We want to do this all the way up to and including
+ root. This way we can keep track of how much usage
+ has occured on the entire system and use that to
+ normalize against.
+ */
while(assoc) {
assoc->grp_used_wall += grp_used_wall;
assoc->usage_raw += (long double)usage_raw;
assoc = assoc->parent_assoc_ptr;
- if(assoc == assoc_mgr_root_assoc)
- break;
}
list_iterator_reset(itr);
}
diff --git a/src/common/assoc_mgr.h b/src/common/assoc_mgr.h
index a57f5a36abfe7893957cfe8adadd39ddea5465bb..1e2a723d0835ea660fac3a64594b1cf24c1cff29 100644
--- a/src/common/assoc_mgr.h
+++ b/src/common/assoc_mgr.h
@@ -61,6 +61,7 @@ typedef struct {
uint16_t cache_level;
uint16_t enforce;
void (*remove_assoc_notify) (acct_association_rec_t *rec);
+ void (*remove_qos_notify) (acct_qos_rec_t *rec);
} assoc_init_args_t;
extern List assoc_mgr_association_list;
diff --git a/src/common/slurm_accounting_storage.c b/src/common/slurm_accounting_storage.c
index 372e3d45d661c8eba14626985c09e5a332d4fc23..be473b46d9d9b92c9874e476a8d8c58f08a5e772 100644
--- a/src/common/slurm_accounting_storage.c
+++ b/src/common/slurm_accounting_storage.c
@@ -7910,6 +7910,12 @@ extern void log_assoc_rec(acct_association_rec_t *assoc_ptr, List qos_list)
if(temp_char) {
debug2(" Qos : %s", temp_char);
xfree(temp_char);
+ if(assoc_ptr->valid_qos) {
+ temp_char = get_qos_complete_str_bitstr(
+ qos_list, assoc_ptr->valid_qos);
+ debug3(" Valid Qos : %s", temp_char);
+ xfree(temp_char);
+ }
}
} else {
debug2(" Qos : %s", "Normal");
diff --git a/src/common/slurm_protocol_defs.c b/src/common/slurm_protocol_defs.c
index fb382893ef533bba8d2a730cfa0579ab2be4e398..2e13bdaf0c011fec459f520a8cbaf75d8da5a37b 100644
--- a/src/common/slurm_protocol_defs.c
+++ b/src/common/slurm_protocol_defs.c
@@ -76,7 +76,6 @@ static void _slurm_free_reserve_info_members (reserve_info_t * part);
static void _free_all_step_info (job_step_info_response_msg_t *msg);
static void _slurm_free_job_step_info_members (job_step_info_t * msg);
-static void _make_lower(char *change);
/*
* slurm_msg_t_init - initialize a slurm message
@@ -172,7 +171,7 @@ extern int slurm_addto_char_list(List char_list, char *names)
else
count++;
- _make_lower(name);
+ xstrtolower(name);
list_append(char_list, name);
list_iterator_reset(itr);
@@ -206,7 +205,7 @@ extern int slurm_addto_char_list(List char_list, char *names)
else
count++;
- _make_lower(name);
+ xstrtolower(name);
list_append(char_list, name);
}
list_iterator_destroy(itr);
@@ -2007,20 +2006,6 @@ void inline slurm_free_job_notify_msg(job_notify_msg_t * msg)
}
}
-/* make everything lowercase should not be called on static char *'s */
-static void _make_lower(char *change)
-{
- if(change) {
- int j = 0;
- while(change[j]) {
- char lower = tolower(change[j]);
- if(lower != change[j])
- change[j] = lower;
- j++;
- }
- }
-}
-
/*
* Sanitize spank_job_env by prepending "SPANK_" to all entries,
* thus rendering them harmless in environment of scripts and
diff --git a/src/common/xstring.c b/src/common/xstring.c
index 7ae5df560f6ddef97594b552a32543f15b50fe9d..675c105324324976c5bac3b2c3fdcc8d0e81678c 100644
--- a/src/common/xstring.c
+++ b/src/common/xstring.c
@@ -86,6 +86,7 @@ strong_alias(_xstrsubstitute, slurm_xstrsubstitute);
strong_alias(xstrstrip, slurm_xstrstrip);
strong_alias(xshort_hostname, slurm_xshort_hostname);
strong_alias(xstring_is_whitespace, slurm_xstring_is_whitespace);
+strong_alias(xstrtolower, slurm_xstrtolower);
/*
* Ensure that a string has enough space to add 'needed' characters.
@@ -487,3 +488,16 @@ bool xstring_is_whitespace(const char *str)
return true;
}
+/*
+ * If str make everything lowercase. Should not be called on static char *'s
+ */
+void xstrtolower(char *str)
+{
+ if(str) {
+ int j = 0;
+ while(str[j]) {
+ str[j] = tolower(str[j]);
+ j++;
+ }
+ }
+}
diff --git a/src/common/xstring.h b/src/common/xstring.h
index 7c7147f615134cb9c0cb0d271692ba6b384a92cd..458e5909c02435e2c6a255231c8ec4d7b067c7f4 100644
--- a/src/common/xstring.h
+++ b/src/common/xstring.h
@@ -149,4 +149,9 @@ char *xshort_hostname(void);
*/
bool xstring_is_whitespace(const char *str);
+/*
+ * If str make everything lowercase. Should not be called on static char *'s
+ */
+void xstrtolower(char *str);
+
#endif /* !_XSTRING_H */
diff --git a/src/plugins/accounting_storage/mysql/accounting_storage_mysql.c b/src/plugins/accounting_storage/mysql/accounting_storage_mysql.c
index c47073e7b4a28a96f6b5271dc3cf90c21f5b1af9..e460e546b241dec16700590182a1b17e81c41310 100644
--- a/src/plugins/accounting_storage/mysql/accounting_storage_mysql.c
+++ b/src/plugins/accounting_storage/mysql/accounting_storage_mysql.c
@@ -3919,14 +3919,17 @@ extern int acct_storage_p_add_users(mysql_conn_t *mysql_conn, uint32_t uid,
xstrfmtcat(vals, ", %u", object->admin_level);
xstrfmtcat(extra, ", admin_level=%u",
object->admin_level);
- }
+ } else
+ xstrfmtcat(extra, ", admin_level=%u",
+ ACCT_ADMIN_NONE);
if(object->default_wckey) {
xstrcat(cols, ", default_wckey");
xstrfmtcat(vals, ", \"%s\"", object->default_wckey);
xstrfmtcat(extra, ", default_wckey=\"%s\"",
object->default_wckey);
- }
+ } else
+ xstrcat(extra, ", default_wckey=\"\"");
query = xstrdup_printf(
"insert into %s (%s) values (%s) "
@@ -5540,8 +5543,8 @@ extern List acct_storage_p_modify_associations(
time_t now = time(NULL);
char *user_name = NULL;
int set = 0, i = 0, is_admin=0;
- MYSQL_RES *result = NULL;
- MYSQL_ROW row;
+ MYSQL_RES *result = NULL, *result2 = NULL;
+ MYSQL_ROW row, row2;
acct_user_rec_t user;
char *tmp_char1=NULL, *tmp_char2=NULL;
int set_qos_vals = 0;
@@ -5572,6 +5575,18 @@ extern List acct_storage_p_modify_associations(
MASSOC_COUNT
};
+ enum {
+ ASSOC2_REQ_PARENT_ID,
+ ASSOC2_REQ_MJ,
+ ASSOC2_REQ_MSJ,
+ ASSOC2_REQ_MCPJ,
+ ASSOC2_REQ_MNPJ,
+ ASSOC2_REQ_MWPJ,
+ ASSOC2_REQ_MCMPJ,
+ ASSOC2_REQ_QOS,
+ ASSOC2_REQ_DELTA_QOS,
+ };
+
if(!assoc_cond || !assoc) {
error("we need something to change");
return NULL;
@@ -5677,24 +5692,26 @@ extern List acct_storage_p_modify_associations(
*/
uint32_t lft = atoi(row[MASSOC_LFT]);
uint32_t rgt = atoi(row[MASSOC_RGT]);
+ char *account;
+
+ /* Here we want to see if the person
+ * is a coord of the parent account
+ * since we don't want him to be able
+ * to alter the limits of the account
+ * he is directly coord of. They
+ * should be able to alter the
+ * sub-accounts though. If no parent account
+ * that means we are talking about a user
+ * association so account is really the parent
+ * of the user a coord can change that all day long.
+ */
+ if(row[MASSOC_PACCT][0])
+ account = row[MASSOC_PACCT];
+ else
+ account = row[MASSOC_ACCT];
if(!is_admin) {
acct_coord_rec_t *coord = NULL;
- char *account = row[MASSOC_ACCT];
-
- /* Here we want to see if the person
- * is a coord of the parent account
- * since we don't want him to be able
- * to alter the limits of the account
- * he is directly coord of. They
- * should be able to alter the
- * sub-accounts though. If no parent account
- * that means we are talking about a user
- * association so account is really the parent
- * of the user a coord can change that all day long.
- */
- if(row[MASSOC_PACCT][0])
- account = row[MASSOC_PACCT];
if(!user.coord_accts) { // This should never
// happen
@@ -5800,6 +5817,50 @@ extern List acct_storage_p_modify_associations(
xstrfmtcat(name_char, " || id=%s", row[MASSOC_ID]);
}
+ /* If there is a variable cleared here we need to make
+ sure we get the parent's information, if any. */
+ query = xstrdup_printf(
+ "call get_parent_limits('%s', "
+ "'%s', '%s', %u);"
+ "select @par_id, @mj, @msj, @mcpj, "
+ "@mnpj, @mwpj, @mcmpj, @qos, @delta_qos;",
+ assoc_table, account,
+ row[MASSOC_CLUSTER], 0);
+ debug4("%d(%d) query\n%s",
+ mysql_conn->conn, __LINE__, query);
+ if(!(result2 = mysql_db_query_ret(
+ mysql_conn->db_conn, query, 1))) {
+ xfree(query);
+ break;
+ }
+ xfree(query);
+
+ if((row2 = mysql_fetch_row(result2))) {
+ if((assoc->max_jobs == INFINITE) && row2[ASSOC2_REQ_MJ])
+ assoc->max_jobs = atoi(row2[ASSOC2_REQ_MJ]);
+ if((assoc->max_submit_jobs == INFINITE)
+ && row2[ASSOC2_REQ_MSJ])
+ assoc->max_submit_jobs =
+ atoi(row2[ASSOC2_REQ_MSJ]);
+ if((assoc->max_cpus_pj == INFINITE)
+ && row2[ASSOC2_REQ_MCPJ])
+ assoc->max_cpus_pj =
+ atoi(row2[ASSOC2_REQ_MCPJ]);
+ if((assoc->max_nodes_pj == INFINITE)
+ && row2[ASSOC2_REQ_MNPJ])
+ assoc->max_nodes_pj =
+ atoi(row2[ASSOC2_REQ_MNPJ]);
+ if((assoc->max_wall_pj == INFINITE)
+ && row2[ASSOC2_REQ_MWPJ])
+ assoc->max_wall_pj =
+ atoi(row2[ASSOC2_REQ_MWPJ]);
+ if((assoc->max_cpu_mins_pj == INFINITE)
+ && row2[ASSOC2_REQ_MCMPJ])
+ assoc->max_cpu_mins_pj =
+ atoi(row2[ASSOC2_REQ_MCMPJ]);
+ }
+ mysql_free_result(result2);
+
mod_assoc = xmalloc(sizeof(acct_association_rec_t));
init_acct_association_rec(mod_assoc);
mod_assoc->id = atoi(row[MASSOC_ID]);
@@ -8587,11 +8648,15 @@ empty:
parent_id = atoi(row2[ASSOC2_REQ_PARENT_ID]);
if(!without_parent_limits) {
- if(row2[ASSOC2_REQ_MCMPJ])
- parent_mcmpj =
- atoi(row2[ASSOC2_REQ_MCMPJ]);
+ if(row2[ASSOC2_REQ_MJ])
+ parent_mj = atoi(row2[ASSOC2_REQ_MJ]);
else
- parent_mcmpj = INFINITE;
+ parent_mj = INFINITE;
+
+ if(row2[ASSOC2_REQ_MSJ])
+ parent_msj = atoi(row2[ASSOC2_REQ_MSJ]);
+ else
+ parent_msj = INFINITE;
if(row2[ASSOC2_REQ_MCPJ])
parent_mcpj =
@@ -8599,11 +8664,6 @@ empty:
else
parent_mcpj = INFINITE;
- if(row2[ASSOC2_REQ_MJ])
- parent_mj = atoi(row2[ASSOC2_REQ_MJ]);
- else
- parent_mj = INFINITE;
-
if(row2[ASSOC2_REQ_MNPJ])
parent_mnpj =
atoi(row2[ASSOC2_REQ_MNPJ]);
@@ -8631,15 +8691,10 @@ empty:
xfree(parent_delta_qos);
if(row2[ASSOC2_REQ_DELTA_QOS][0])
- xstrcat(parent_delta_qos,
+ parent_delta_qos = xstrdup(
row2[ASSOC2_REQ_DELTA_QOS]);
else
parent_delta_qos = NULL;
-
- if(row2[ASSOC2_REQ_MSJ])
- parent_msj = atoi(row2[ASSOC2_REQ_MSJ]);
- else
- parent_msj = INFINITE;
}
last_acct = parent_acct;
last_cluster = row[ASSOC_REQ_CLUSTER];
diff --git a/src/plugins/jobacct_gather/aix/jobacct_gather_aix.c b/src/plugins/jobacct_gather/aix/jobacct_gather_aix.c
index 32a1e0304dd372ae4cf74b4fec56c410b5399a73..aa50224e24eb258a0090e2df5f59a76afe632798 100644
--- a/src/plugins/jobacct_gather/aix/jobacct_gather_aix.c
+++ b/src/plugins/jobacct_gather/aix/jobacct_gather_aix.c
@@ -41,6 +41,7 @@
\*****************************************************************************/
#include <signal.h>
+#include "src/common/slurm_xlator.h"
#include "src/common/jobacct_common.h"
#include "src/common/slurm_protocol_api.h"
#include "src/common/slurm_protocol_defs.h"
diff --git a/src/plugins/jobacct_gather/linux/jobacct_gather_linux.c b/src/plugins/jobacct_gather/linux/jobacct_gather_linux.c
index c4b42643bcb2fe13d5651c4bf48cef0c24cf2ba4..d2c63108212f468fafbe4755c16b0add752efd0d 100644
--- a/src/plugins/jobacct_gather/linux/jobacct_gather_linux.c
+++ b/src/plugins/jobacct_gather/linux/jobacct_gather_linux.c
@@ -41,6 +41,7 @@
#include <fcntl.h>
#include <signal.h>
+#include "src/common/slurm_xlator.h"
#include "src/common/jobacct_common.h"
#include "src/common/slurm_protocol_api.h"
#include "src/common/slurm_protocol_defs.h"
diff --git a/src/plugins/jobacct_gather/none/jobacct_gather_none.c b/src/plugins/jobacct_gather/none/jobacct_gather_none.c
index 1ced7a60fb598bd30827045a77fb31c9f7bc14fb..0cda67945302b8aba772a592aeaa179a3dd1dd76 100644
--- a/src/plugins/jobacct_gather/none/jobacct_gather_none.c
+++ b/src/plugins/jobacct_gather/none/jobacct_gather_none.c
@@ -39,6 +39,7 @@
* Copyright (C) 2002 The Regents of the University of California.
\*****************************************************************************/
+#include "src/common/slurm_xlator.h"
#include "src/common/slurm_jobacct_gather.h"
/*
diff --git a/src/plugins/priority/multifactor/priority_multifactor.c b/src/plugins/priority/multifactor/priority_multifactor.c
index 2d19542596c9514ddc792e58612f1bfa7957d6cd..db468ef5712d0beb4b1a6b885ed8b5f9b9115924 100644
--- a/src/plugins/priority/multifactor/priority_multifactor.c
+++ b/src/plugins/priority/multifactor/priority_multifactor.c
@@ -65,6 +65,11 @@
#define SECS_PER_DAY (24 * 60 * 60)
#define SECS_PER_WEEK (7 * 24 * 60 * 60)
+/* These are defined here so when we link with something other than
+ * the slurmctld we will have these symbols defined. They will get
+ * overwritten when linking with the slurmctld.
+ */
+time_t last_job_update;
/*
* These variables are required by the generic plugin interface. If they
@@ -141,9 +146,10 @@ static int _apply_decay(double decay_factor)
slurm_mutex_lock(&assoc_mgr_association_lock);
itr = list_iterator_create(assoc_mgr_association_list);
+ /* We want to do this to all associations including
+ root. All usage_raws are calculated from the bottom up.
+ */
while((assoc = list_next(itr))) {
- if (assoc == assoc_mgr_root_assoc)
- continue;
assoc->usage_raw *= decay_factor;
assoc->grp_used_wall *= decay_factor;
}
@@ -180,9 +186,10 @@ static int _reset_usage()
slurm_mutex_lock(&assoc_mgr_association_lock);
itr = list_iterator_create(assoc_mgr_association_list);
+ /* We want to do this to all associations including
+ root. All usage_raws are calculated from the bottom up.
+ */
while((assoc = list_next(itr))) {
- if (assoc == assoc_mgr_root_assoc)
- continue;
assoc->usage_raw = 0;
assoc->grp_used_wall = 0;
}
@@ -781,25 +788,24 @@ static void *_decay_thread(void *no_data)
}
slurm_mutex_lock(&assoc_mgr_association_lock);
+ /* We want to do this all the way up
+ to and including root. This way we
+ can keep track of how much usage
+ has occured on the entire system
+ and use that to normalize against.
+ */
while(assoc) {
- /* we don't want to make the
- root assoc responsible for
- keeping track of time
- */
- if (assoc == assoc_mgr_root_assoc)
- break;
assoc->grp_used_wall += run_decay;
assoc->usage_raw +=
(long double)real_decay;
debug4("adding %f new usage to "
"assoc %u (user='%s' acct='%s') "
"raw usage is now %Lf. Group "
- "wall added %d making it %f.",
+ "wall added %f making it %f.",
real_decay, assoc->id,
assoc->user, assoc->acct,
assoc->usage_raw, run_decay,
assoc->grp_used_wall);
-
assoc = assoc->parent_assoc_ptr;
}
slurm_mutex_unlock(&assoc_mgr_association_lock);
@@ -1033,10 +1039,10 @@ extern int priority_p_set_max_cluster_usage(uint32_t procs, uint32_t half_life)
last_procs = procs;
last_half_life = half_life;
- /* get the total decay for the entire cluster */
- assoc_mgr_root_assoc->usage_raw =
- (long double)procs * (long double)half_life * (long double)2;
- assoc_mgr_root_assoc->usage_norm = 1.0;
+ /* This should always be 1 and it doesn't get calculated later
+ so set it now. usage_raw and usage_norm get calculated the
+ same way the other associations do. */
+ assoc_mgr_root_assoc->usage_efctv = 1.0;
debug3("Total possible cpu usage for half_life of %d secs "
"on the system is %.0Lf",
half_life, assoc_mgr_root_assoc->usage_raw);
@@ -1057,10 +1063,18 @@ extern void priority_p_set_assoc_usage(acct_association_rec_t *assoc)
}
xassert(assoc_mgr_root_assoc);
- xassert(assoc_mgr_root_assoc->usage_raw);
xassert(assoc->parent_assoc_ptr);
- assoc->usage_norm = assoc->usage_raw / assoc_mgr_root_assoc->usage_raw;
+ if(assoc_mgr_root_assoc->usage_raw)
+ assoc->usage_norm =
+ assoc->usage_raw / assoc_mgr_root_assoc->usage_raw;
+ else
+ /* This should only happen when no usage has occured
+ at all so no big deal, the other usage should be 0
+ as well here.
+ */
+ assoc->usage_norm = 0;
+
debug4("Normalized usage for %s %s off %s %Lf / %Lf = %Lf",
child, child_str, assoc->parent_assoc_ptr->acct,
assoc->usage_raw, assoc_mgr_root_assoc->usage_raw,
diff --git a/src/plugins/proctrack/linuxproc/kill_tree.c b/src/plugins/proctrack/linuxproc/kill_tree.c
index 4f32c3827e7f588e087428b8609c1b90548dca6d..88a8211f9583f6e5f9c1fea6824b3ca3d8c5d8d9 100644
--- a/src/plugins/proctrack/linuxproc/kill_tree.c
+++ b/src/plugins/proctrack/linuxproc/kill_tree.c
@@ -54,9 +54,10 @@
#include <limits.h>
#include "slurm/slurm.h"
+#include "slurm/slurm_errno.h"
+#include "src/common/log.h"
#include "src/common/xmalloc.h"
#include "src/common/xstring.h"
-#include "src/common/log.h"
#include "kill_tree.h"
typedef struct xpid_s {
@@ -163,12 +164,14 @@ static xppid_t **_build_hashtbl(void)
hashtbl = (xppid_t **)xmalloc(HASH_LEN * sizeof(xppid_t *));
+ slurm_seterrno(0);
while ((de = readdir(dir)) != NULL) {
num = de->d_name;
if ((num[0] < '0') || (num[0] > '9'))
continue;
ret_l = strtol(num, &endptr, 10);
- if(errno == ERANGE) {
+ if ((ret_l == LONG_MIN) || (ret_l == LONG_MAX) ||
+ (errno == ERANGE)) {
error("couldn't do a strtol on str %s(%d): %m",
num, ret_l);
}
diff --git a/src/plugins/sched/builtin/builtin_wrapper.c b/src/plugins/sched/builtin/builtin_wrapper.c
index 30a28eb2c7629a9a45e92d6894d6a2684b689154..b7ea5acf10bba286223c916d9b0811adfb1c0a54 100644
--- a/src/plugins/sched/builtin/builtin_wrapper.c
+++ b/src/plugins/sched/builtin/builtin_wrapper.c
@@ -177,6 +177,7 @@ void slurm_sched_plugin_job_is_pending( void )
FREE_NULL_BITMAP(avail_bitmap);
}
+ xfree(job_queue);
}
/**************************************************************************/
diff --git a/src/plugins/select/bluegene/plugin/bg_job_place.c b/src/plugins/select/bluegene/plugin/bg_job_place.c
index 8ff04f89178960bb6bac15eb49107a0eb9589d64..9b111fb16544b170be12ad4e02bf93a4d2b6faaa 100644
--- a/src/plugins/select/bluegene/plugin/bg_job_place.c
+++ b/src/plugins/select/bluegene/plugin/bg_job_place.c
@@ -314,8 +314,8 @@ static bg_record_t *_find_matching_block(List block_list,
continue;
} else if((bg_record->job_running != NO_JOB_RUNNING)
&& (bg_record->job_running != job_ptr->job_id)
- && (bg_conf->layout_mode == LAYOUT_DYNAMIC
- || (SELECT_IS_MODE_RUN_NOW(query_mode)
+ && ((bg_conf->layout_mode == LAYOUT_DYNAMIC)
+ || (!SELECT_IS_CHECK_FULL_SET(query_mode)
&& bg_conf->layout_mode != LAYOUT_DYNAMIC))) {
debug("block %s in use by %s job %d",
bg_record->bg_block_id,
@@ -1525,6 +1525,10 @@ preempt:
} else if(bg_record->job_running == BLOCK_ERROR_STATE)
starttime = INFINITE;
+ /* make sure the job is eligible to run */
+ if(job_ptr->details->begin_time > starttime)
+ starttime = job_ptr->details->begin_time;
+
job_ptr->start_time = starttime;
select_g_select_jobinfo_set(job_ptr->select_jobinfo,
diff --git a/src/plugins/select/cons_res/job_test.c b/src/plugins/select/cons_res/job_test.c
index af31c8f1c6a1420cf4c9e3bc4141986b442cebf2..fabf2575c9e2718381690f0b910ca505504ff88c 100644
--- a/src/plugins/select/cons_res/job_test.c
+++ b/src/plugins/select/cons_res/job_test.c
@@ -133,9 +133,10 @@ uint16_t _allocate_sockets(struct job_record *job_ptr, bitstr_t *core_map,
uint16_t si, cps, avail_cpus = 0, num_tasks = 0;
uint32_t core_begin = cr_get_coremap_offset(node_i);
uint32_t core_end = cr_get_coremap_offset(node_i+1);
+ uint32_t c;
uint16_t cpus_per_task = job_ptr->details->cpus_per_task;
uint16_t *used_cores, *free_cores, free_core_count = 0;
- uint16_t i, c, sockets = select_node_record[node_i].sockets;
+ uint16_t i, j, sockets = select_node_record[node_i].sockets;
uint16_t cores_per_socket = select_node_record[node_i].cores;
uint16_t threads_per_core = select_node_record[node_i].vpus;
uint16_t min_cores = 1, min_sockets = 1, ntasks_per_socket = 0;
@@ -220,7 +221,7 @@ uint16_t _allocate_sockets(struct job_record *job_ptr, bitstr_t *core_map,
used_cores = xmalloc(sockets * sizeof(uint16_t));
for (c = core_begin; c < core_end; c++) {
- i = (c - core_begin) / cores_per_socket;
+ i = (uint16_t) (c - core_begin) / cores_per_socket;
if (bit_test(core_map, c)) {
free_cores[i]++;
free_core_count++;
@@ -241,7 +242,7 @@ uint16_t _allocate_sockets(struct job_record *job_ptr, bitstr_t *core_map,
used_cores = NULL;
/* Step 2: check min_cores per socket and min_sockets per node */
- c = 0;
+ j = 0;
for (i = 0; i < sockets; i++) {
if (free_cores[i] < min_cores) {
/* cannot use this socket */
@@ -250,16 +251,16 @@ uint16_t _allocate_sockets(struct job_record *job_ptr, bitstr_t *core_map,
continue;
}
/* count this socket as usable */
- c++;
+ j++;
}
- if (c < min_sockets) {
+ if (j < min_sockets) {
/* cannot use this node */
num_tasks = 0;
goto fini;
}
/* check max_cores and max_sockets */
- c = 0;
+ j = 0;
for (i = 0; i < sockets; i++) {
if (free_cores[i] > max_cores) {
/* remove extra cores from this socket */
@@ -268,8 +269,8 @@ uint16_t _allocate_sockets(struct job_record *job_ptr, bitstr_t *core_map,
free_cores[i] -= tmp;
}
if (free_cores[i] > 0)
- c++;
- if (free_cores[i] && c > max_sockets) {
+ j++;
+ if (free_cores[i] && j > max_sockets) {
/* remove extra sockets from use */
free_core_count -= free_cores[i];
free_cores[i] = 0;
@@ -310,8 +311,8 @@ uint16_t _allocate_sockets(struct job_record *job_ptr, bitstr_t *core_map,
avail_cpus = num_tasks;
cps = num_tasks;
} else {
- c = avail_cpus / cpus_per_task;
- num_tasks = MIN(num_tasks, c);
+ j = avail_cpus / cpus_per_task;
+ num_tasks = MIN(num_tasks, j);
avail_cpus = num_tasks * cpus_per_task;
}
if (job_ptr->details->ntasks_per_node &&
@@ -334,7 +335,7 @@ uint16_t _allocate_sockets(struct job_record *job_ptr, bitstr_t *core_map,
for (c = core_begin; c < core_end && avail_cpus > 0; c++) {
if (bit_test(core_map, c) == 0)
continue;
- i = (c - core_begin) / cores_per_socket;
+ i = (uint16_t) (c - core_begin) / cores_per_socket;
if (free_cores[i] > 0) {
/* this socket has free cores, but make sure
* we don't use more than are needed for
@@ -390,9 +391,10 @@ uint16_t _allocate_cores(struct job_record *job_ptr, bitstr_t *core_map,
uint16_t cpu_count = 0, avail_cpus = 0, num_tasks = 0;
uint32_t core_begin = cr_get_coremap_offset(node_i);
uint32_t core_end = cr_get_coremap_offset(node_i+1);
+ uint32_t c;
uint16_t cpus_per_task = job_ptr->details->cpus_per_task;
uint16_t *free_cores, free_core_count = 0;
- uint16_t i, c, sockets = select_node_record[node_i].sockets;
+ uint16_t i, j, sockets = select_node_record[node_i].sockets;
uint16_t cores_per_socket = select_node_record[node_i].cores;
uint16_t threads_per_core = select_node_record[node_i].vpus;
uint16_t min_cores = 1, min_sockets = 1;
@@ -471,7 +473,7 @@ uint16_t _allocate_cores(struct job_record *job_ptr, bitstr_t *core_map,
free_cores = xmalloc(sockets * sizeof(uint16_t));
for (c = core_begin; c < core_end; c++) {
- i = (c - core_begin) / cores_per_socket;
+ i = (uint16_t) (c - core_begin) / cores_per_socket;
if (bit_test(core_map, c)) {
free_cores[i]++;
free_core_count++;
@@ -479,7 +481,7 @@ uint16_t _allocate_cores(struct job_record *job_ptr, bitstr_t *core_map,
}
/* Step 2a: check min_cores per socket and min_sockets per node */
- c = 0;
+ j = 0;
for (i = 0; i < sockets; i++) {
if (free_cores[i] < min_cores) {
/* cannot use this socket */
@@ -488,16 +490,16 @@ uint16_t _allocate_cores(struct job_record *job_ptr, bitstr_t *core_map,
continue;
}
/* count this socket as usable */
- c++;
+ j++;
}
- if (c < min_sockets) {
+ if (j < min_sockets) {
/* cannot use this node */
num_tasks = 0;
goto fini;
}
/* Step 2b: check max_cores per socket and max_sockets per node */
- c = 0;
+ j = 0;
for (i = 0; i < sockets; i++) {
if (free_cores[i] > max_cores) {
/* remove extra cores from this socket */
@@ -506,8 +508,8 @@ uint16_t _allocate_cores(struct job_record *job_ptr, bitstr_t *core_map,
free_cores[i] -= tmp;
}
if (free_cores[i] > 0)
- c++;
- if (free_cores[i] && (c > max_sockets)) {
+ j++;
+ if (free_cores[i] && (j > max_sockets)) {
/* remove extra sockets from use */
free_core_count -= free_cores[i];
free_cores[i] = 0;
@@ -548,8 +550,8 @@ uint16_t _allocate_cores(struct job_record *job_ptr, bitstr_t *core_map,
if (cpus_per_task < 2) {
avail_cpus = num_tasks;
} else {
- c = avail_cpus / cpus_per_task;
- num_tasks = MIN(num_tasks, c);
+ j = avail_cpus / cpus_per_task;
+ num_tasks = MIN(num_tasks, j);
avail_cpus = num_tasks * cpus_per_task;
}
if (job_ptr->details->ntasks_per_node &&
@@ -563,7 +565,7 @@ uint16_t _allocate_cores(struct job_record *job_ptr, bitstr_t *core_map,
for (c = core_begin; c < core_end && avail_cpus > 0; c++) {
if (bit_test(core_map, c) == 0)
continue;
- i = (c - core_begin) / cores_per_socket;
+ i = (uint16_t) (c - core_begin) / cores_per_socket;
if (free_cores[i] == 0)
bit_clear(core_map, c);
else {
diff --git a/src/plugins/select/cons_res/select_cons_res.c b/src/plugins/select/cons_res/select_cons_res.c
index 7278eea888b3f7916b396f7da14257fc4abe9608..cc6c9643af5ab39fb2165cf4fe3c13f1f067a2b8 100644
--- a/src/plugins/select/cons_res/select_cons_res.c
+++ b/src/plugins/select/cons_res/select_cons_res.c
@@ -101,7 +101,6 @@
# endif
#endif
-#include "src/common/slurm_xlator.h"
#include "select_cons_res.h"
#include "dist_tasks.h"
#include "job_test.h"
@@ -1182,7 +1181,7 @@ static int _run_now(struct job_record *job_ptr, bitstr_t *bitmap,
preemptee_candidates)) {
/* Remove preemptable job now */
_rm_job_from_res(future_part, future_usage,
- tmp_job_ptr, 2);
+ tmp_job_ptr, 0);
bit_or(bitmap, orig_map);
rc = cr_job_test(job_ptr, bitmap, min_nodes,
max_nodes, req_nodes,
diff --git a/src/salloc/opt.c b/src/salloc/opt.c
index a45c5c0a5fc31e1ce4fdb2878f84ce328a4d3ca7..a013d2b2307a4e61f68f920a1937f3a32413201e 100644
--- a/src/salloc/opt.c
+++ b/src/salloc/opt.c
@@ -980,6 +980,7 @@ void set_options(const int argc, char **argv)
opt.qos = xstrdup(optarg);
break;
case LONG_OPT_SOCKETSPERNODE:
+ max_val = 0;
get_resource_arg_range( optarg, "sockets-per-node",
&opt.min_sockets_per_node,
&max_val, true );
@@ -988,6 +989,7 @@ void set_options(const int argc, char **argv)
opt.min_sockets_per_node = NO_VAL;
break;
case LONG_OPT_CORESPERSOCKET:
+ max_val = 0;
get_resource_arg_range( optarg, "cores-per-socket",
&opt.min_cores_per_socket,
&max_val, true );
@@ -996,6 +998,7 @@ void set_options(const int argc, char **argv)
opt.min_cores_per_socket = NO_VAL;
break;
case LONG_OPT_THREADSPERCORE:
+ max_val = 0;
get_resource_arg_range( optarg, "threads-per-core",
&opt.min_threads_per_core,
&max_val, true );
@@ -1218,7 +1221,7 @@ static bool _opt_verify(void)
/* check for realistic arguments */
if (opt.nprocs <= 0) {
- error("invalid number of processes (-n %d)",
+ error("invalid number of tasks (-n %d)",
opt.nprocs);
verified = false;
}
diff --git a/src/salloc/salloc.c b/src/salloc/salloc.c
index 09d63983904777293538e0ab5402a1a3fd7bbc0f..24ace1341cee96640b7d4958ecee7ec6ed53794b 100644
--- a/src/salloc/salloc.c
+++ b/src/salloc/salloc.c
@@ -339,9 +339,9 @@ int main(int argc, char *argv[])
return 1;
} else {
allocation_state = GRANTED;
+ pthread_mutex_unlock(&allocation_state_lock);
command_pid = pid = _fork_command(command_argv);
}
- pthread_mutex_unlock(&allocation_state_lock);
/*
* Wait for command to exit, OR for waitpid to be interrupted by a
diff --git a/src/sbatch/opt.c b/src/sbatch/opt.c
index 1f9155a03aa392e36dcbeead39af7af11ddd530f..d7f1c4940be336654c505edc0756e86b26c8ec5f 100644
--- a/src/sbatch/opt.c
+++ b/src/sbatch/opt.c
@@ -283,7 +283,7 @@ static void _opt_default()
opt.nprocs_set = false;
opt.cpus_per_task = 1;
opt.cpus_set = false;
- opt.min_nodes = 0;
+ opt.min_nodes = 1;
opt.max_nodes = 0;
opt.nodes_set = false;
opt.min_sockets_per_node = NO_VAL; /* requested min sockets */
@@ -1421,6 +1421,7 @@ static void _set_options(int argc, char **argv)
opt.qos = xstrdup(optarg);
break;
case LONG_OPT_SOCKETSPERNODE:
+ max_val = 0;
get_resource_arg_range( optarg, "sockets-per-node",
&opt.min_sockets_per_node,
&max_val, true );
@@ -1429,6 +1430,7 @@ static void _set_options(int argc, char **argv)
opt.min_sockets_per_node = NO_VAL;
break;
case LONG_OPT_CORESPERSOCKET:
+ max_val = 0;
get_resource_arg_range( optarg, "cores-per-socket",
&opt.min_cores_per_socket,
&max_val, true );
@@ -1437,6 +1439,7 @@ static void _set_options(int argc, char **argv)
opt.min_cores_per_socket = NO_VAL;
break;
case LONG_OPT_THREADSPERCORE:
+ max_val = 0;
get_resource_arg_range( optarg, "threads-per-core",
&opt.min_threads_per_core,
&max_val, true );
@@ -1996,7 +1999,7 @@ static bool _opt_verify(void)
/* check for realistic arguments */
if (opt.nprocs <= 0) {
- error("invalid number of processes (-n %d)", opt.nprocs);
+ error("invalid number of tasks (-n %d)", opt.nprocs);
verified = false;
}
diff --git a/src/scontrol/update_job.c b/src/scontrol/update_job.c
index 9cc255b13c4fe910220730f7aecd0c08b4fa8022..3c4502f4cbce3d9c1a11c9ce670668f72357bb6d 100644
--- a/src/scontrol/update_job.c
+++ b/src/scontrol/update_job.c
@@ -405,6 +405,10 @@ scontrol_update_job (int argc, char *argv[])
job_msg.partition = val;
update_cnt++;
}
+ else if (strncasecmp(tag, "QOS", MAX(taglen, 2)) == 0) {
+ job_msg.qos = val;
+ update_cnt++;
+ }
else if (strncasecmp(tag, "ReservationName",
MAX(taglen, 3)) == 0) {
job_msg.reservation = val;
diff --git a/src/slurmctld/acct_policy.c b/src/slurmctld/acct_policy.c
index 0fcaba9601a8578acad8df836ffa40c0bea07b43..c09bc3850ee439dc18924ee91978b3b865033a63 100644
--- a/src/slurmctld/acct_policy.c
+++ b/src/slurmctld/acct_policy.c
@@ -367,6 +367,7 @@ extern bool acct_policy_job_runnable(struct job_record *job_ptr)
bool rc = true;
uint64_t usage_mins;
uint32_t wall_mins;
+ bool cancel_job = 0;
int parent = 0; /*flag to tell us if we are looking at the
* parent or not
*/
@@ -391,6 +392,10 @@ extern bool acct_policy_job_runnable(struct job_record *job_ptr)
job_ptr->state_reason = WAIT_NO_REASON;
slurm_mutex_lock(&assoc_mgr_qos_lock);
+ /* Since we could possibly need the association lock handle it
+ now to avoid deadlock. Always do QOS first.
+ */
+ slurm_mutex_lock(&assoc_mgr_association_lock);
qos_ptr = job_ptr->qos_ptr;
if(qos_ptr) {
usage_mins = (uint64_t)(qos_ptr->usage_raw / 60.0);
@@ -407,7 +412,7 @@ extern bool acct_policy_job_runnable(struct job_record *job_ptr)
qos_ptr->name, qos_ptr->grp_cpu_mins,
usage_mins);
rc = false;
- goto end_qos;
+ goto end_it;
}
/* NOTE: We can't enforce qos_ptr->grp_cpus at this
@@ -426,7 +431,7 @@ extern bool acct_policy_job_runnable(struct job_record *job_ptr)
qos_ptr->grp_used_jobs, qos_ptr->name);
rc = false;
- goto end_qos;
+ goto end_it;
}
if (qos_ptr->grp_nodes != INFINITE) {
@@ -438,7 +443,7 @@ extern bool acct_policy_job_runnable(struct job_record *job_ptr)
job_ptr->details->min_nodes,
qos_ptr->grp_nodes,
qos_ptr->name);
- _cancel_job(job_ptr);
+ cancel_job = 1;
} else if ((qos_ptr->grp_used_nodes +
job_ptr->details->min_nodes) >
qos_ptr->grp_nodes) {
@@ -456,7 +461,7 @@ extern bool acct_policy_job_runnable(struct job_record *job_ptr)
job_ptr->details->min_nodes,
qos_ptr->name);
rc = false;
- goto end_qos;
+ goto end_it;
}
}
@@ -475,7 +480,7 @@ extern bool acct_policy_job_runnable(struct job_record *job_ptr)
wall_mins, qos_ptr->name);
rc = false;
- goto end_qos;
+ goto end_it;
}
/* NOTE: We can't enforce qos_ptr->max_cpu_mins_pj at this
@@ -506,7 +511,7 @@ extern bool acct_policy_job_runnable(struct job_record *job_ptr)
qos_ptr->max_jobs_pu,
used_limits->jobs, qos_ptr->name);
rc = false;
- goto end_qos;
+ goto end_it;
}
}
@@ -519,9 +524,9 @@ extern bool acct_policy_job_runnable(struct job_record *job_ptr)
job_ptr->job_id,
job_ptr->details->min_nodes,
qos_ptr->max_nodes_pj);
- _cancel_job(job_ptr);
+ cancel_job = 1;
rc = false;
- goto end_qos;
+ goto end_it;
}
}
@@ -537,14 +542,13 @@ extern bool acct_policy_job_runnable(struct job_record *job_ptr)
"time limit %u exceeds account max %u",
job_ptr->job_id, job_ptr->time_limit,
time_limit);
- _cancel_job(job_ptr);
+ cancel_job = 1;
rc = false;
- goto end_qos;
+ goto end_it;
}
}
}
- slurm_mutex_lock(&assoc_mgr_association_lock);
assoc_ptr = job_ptr->assoc_ptr;
while(assoc_ptr) {
usage_mins = (uint64_t)(assoc_ptr->usage_raw / 60.0);
@@ -599,7 +603,7 @@ extern bool acct_policy_job_runnable(struct job_record *job_ptr)
job_ptr->job_id,
job_ptr->details->min_nodes,
assoc_ptr->grp_nodes, assoc_ptr->acct);
- _cancel_job(job_ptr);
+ cancel_job = 1;
} else if ((assoc_ptr->grp_used_nodes +
job_ptr->details->min_nodes) >
assoc_ptr->grp_nodes) {
@@ -686,7 +690,7 @@ extern bool acct_policy_job_runnable(struct job_record *job_ptr)
job_ptr->job_id,
job_ptr->details->min_nodes,
assoc_ptr->max_nodes_pj);
- _cancel_job(job_ptr);
+ cancel_job = 1;
rc = false;
goto end_it;
}
@@ -706,7 +710,7 @@ extern bool acct_policy_job_runnable(struct job_record *job_ptr)
"time limit %u exceeds account max %u",
job_ptr->job_id, job_ptr->time_limit,
time_limit);
- _cancel_job(job_ptr);
+ cancel_job = 1;
rc = false;
goto end_it;
}
@@ -717,7 +721,10 @@ extern bool acct_policy_job_runnable(struct job_record *job_ptr)
}
end_it:
slurm_mutex_unlock(&assoc_mgr_association_lock);
-end_qos:
slurm_mutex_unlock(&assoc_mgr_qos_lock);
+
+ if(cancel_job)
+ _cancel_job(job_ptr);
+
return rc;
}
diff --git a/src/slurmctld/controller.c b/src/slurmctld/controller.c
index b05e78e7a3bb43bc6e46069a9e86fd80a69e1629..0054135508835b5859b81e75c7bb09c00226659c 100644
--- a/src/slurmctld/controller.c
+++ b/src/slurmctld/controller.c
@@ -187,6 +187,7 @@ static void _init_pidfile(void);
static void _kill_old_slurmctld(void);
static void _parse_commandline(int argc, char *argv[]);
static void _remove_assoc(acct_association_rec_t *rec);
+static void _remove_qos(acct_qos_rec_t *rec);
inline static int _report_locks_set(void);
static void * _service_connection(void *arg);
static int _shutdown_backup_controller(int wait_time);
@@ -326,6 +327,7 @@ int main(int argc, char *argv[])
memset(&assoc_init_arg, 0, sizeof(assoc_init_args_t));
assoc_init_arg.enforce = accounting_enforce;
assoc_init_arg.remove_assoc_notify = _remove_assoc;
+ assoc_init_arg.remove_qos_notify = _remove_qos;
assoc_init_arg.cache_level = ASSOC_MGR_CACHE_ASSOC |
ASSOC_MGR_CACHE_USER | ASSOC_MGR_CACHE_QOS;
@@ -1127,6 +1129,18 @@ static void _remove_assoc(acct_association_rec_t *rec)
debug("Removed association id:%u user:%s", rec->id, rec->user);
}
+static void _remove_qos(acct_qos_rec_t *rec)
+{
+ int cnt = 0;
+
+ cnt = job_cancel_by_qos_id(rec->id);
+
+ if (cnt) {
+ info("Removed QOS:%s cancelled %u jobs", rec->name, cnt);
+ } else
+ debug("Removed QOS:%s", rec->name);
+}
+
/*
* _slurmctld_background - process slurmctld background activities
* purge defunct job records, save state, schedule jobs, and
diff --git a/src/slurmctld/job_mgr.c b/src/slurmctld/job_mgr.c
index 634119c59ba63b6b21d025b223dedc7b9b910c38..de213c6f030162de9f8e9ba91c86765ec078e3c5 100644
--- a/src/slurmctld/job_mgr.c
+++ b/src/slurmctld/job_mgr.c
@@ -934,6 +934,7 @@ static int _load_job_state(Buf buffer)
xfree(job_ptr->account);
job_ptr->account = account;
+ xstrtolower(job_ptr->account);
account = NULL; /* reused, nothing left to free */
xfree(job_ptr->alloc_node);
job_ptr->alloc_node = alloc_node;
@@ -965,6 +966,7 @@ static int _load_job_state(Buf buffer)
name = NULL; /* reused, nothing left to free */
xfree(job_ptr->wckey); /* in case duplicate record */
job_ptr->wckey = wckey;
+ xstrtolower(job_ptr->wckey);
wckey = NULL; /* reused, nothing left to free */
xfree(job_ptr->network);
job_ptr->network = network;
@@ -2461,6 +2463,11 @@ static int _job_create(job_desc_msg_t * job_desc, int allocate, int will_run,
return error_code;
}
+ /* Make sure anything that may be put in the database will be
+ lower case */
+ xstrtolower(job_desc->account);
+ xstrtolower(job_desc->wckey);
+
if ((error_code = _validate_job_desc(job_desc, allocate, submit_uid)))
return error_code;
@@ -3358,7 +3365,6 @@ _copy_job_desc_to_job_record(job_desc_msg_t * job_desc,
return ESLURM_INVALID_WCKEY;
}
}
- job_desc->wckey = xstrdup(job_desc->wckey);
} else if (accounting_enforce & ACCOUNTING_ENFORCE_WCKEYS) {
/* This should never happen */
info("_job_create: no wckey was given for "
@@ -3645,9 +3651,6 @@ void job_time_limit(void)
if (!IS_JOB_RUNNING(job_ptr))
continue;
- qos = (acct_qos_rec_t *)job_ptr->qos_ptr;
- assoc = (acct_association_rec_t *)job_ptr->assoc_ptr;
-
/* find out how many cpu minutes this job has been
* running for. */
job_cpu_usage_mins = (uint64_t)
@@ -3709,6 +3712,14 @@ void job_time_limit(void)
(list_count(job_ptr->step_list) > 0))
check_job_step_time_limit(job_ptr, now);
+ slurm_mutex_lock(&assoc_mgr_qos_lock);
+ /* Handle both locks here to avoid deadlock. Always do
+ * QOS first.
+ */
+ slurm_mutex_lock(&assoc_mgr_association_lock);
+ qos = (acct_qos_rec_t *)job_ptr->qos_ptr;
+ assoc = (acct_association_rec_t *)job_ptr->assoc_ptr;
+
/* The idea here is for qos to trump what an association
* has set for a limit, so if an association set of
* wall 10 mins and the qos has 20 mins set and the
@@ -3716,7 +3727,6 @@ void job_time_limit(void)
* until 20.
*/
if(qos) {
- slurm_mutex_lock(&assoc_mgr_qos_lock);
usage_mins = (uint64_t)(qos->usage_raw / 60.0);
wall_mins = qos->grp_used_wall / 60;
@@ -3730,7 +3740,6 @@ void job_time_limit(void)
qos->name, qos->grp_cpu_mins,
usage_mins);
job_ptr->state_reason = FAIL_TIMEOUT;
- slurm_mutex_unlock(&assoc_mgr_qos_lock);
goto job_failed;
}
@@ -3744,7 +3753,6 @@ void job_time_limit(void)
qos->name, qos->grp_wall,
wall_mins);
job_ptr->state_reason = FAIL_TIMEOUT;
- slurm_mutex_unlock(&assoc_mgr_qos_lock);
goto job_failed;
}
@@ -3758,14 +3766,11 @@ void job_time_limit(void)
qos->name, qos->max_cpu_mins_pj,
job_cpu_usage_mins);
job_ptr->state_reason = FAIL_TIMEOUT;
- slurm_mutex_unlock(&assoc_mgr_qos_lock);
goto job_failed;
}
- slurm_mutex_unlock(&assoc_mgr_qos_lock);
}
/* handle any association stuff here */
- slurm_mutex_lock(&assoc_mgr_association_lock);
while(assoc) {
usage_mins = (uint64_t)(assoc->usage_raw / 60.0);
wall_mins = assoc->grp_used_wall / 60;
@@ -3817,8 +3822,10 @@ void job_time_limit(void)
if(assoc == assoc_mgr_root_assoc)
break;
}
- slurm_mutex_unlock(&assoc_mgr_association_lock);
job_failed:
+ slurm_mutex_unlock(&assoc_mgr_association_lock);
+ slurm_mutex_unlock(&assoc_mgr_qos_lock);
+
if(job_ptr->state_reason == FAIL_TIMEOUT) {
last_job_update = now;
_job_timed_out(job_ptr);
@@ -4843,6 +4850,11 @@ int update_job(job_desc_msg_t * job_specs, uid_t uid)
&cpus_per_node);
#endif
+ /* Make sure anything that may be put in the database will be
+ lower case */
+ xstrtolower(job_specs->account);
+ xstrtolower(job_specs->wckey);
+
job_ptr = find_job_record(job_specs->job_id);
if (job_ptr == NULL) {
error("update_job: job_id %u does not exist.",
@@ -4940,32 +4952,41 @@ int update_job(job_desc_msg_t * job_specs, uid_t uid)
if (wiki_sched && strstr(job_ptr->comment, "QOS:")) {
acct_qos_rec_t qos_rec;
+ if (!IS_JOB_PENDING(job_ptr))
+ error_code = ESLURM_DISABLED;
+ else {
+ memset(&qos_rec, 0, sizeof(acct_qos_rec_t));
+
+ if (strstr(job_ptr->comment,
+ "FLAGS:PREEMPTOR"))
+ qos_rec.name = "expedite";
+ else if (strstr(job_ptr->comment,
+ "FLAGS:PREEMPTEE"))
+ qos_rec.name = "standby";
+
+ job_ptr->qos_ptr = _determine_and_validate_qos(
+ job_ptr->assoc_ptr, &qos_rec,
+ &error_code);
+ job_ptr->qos = qos_rec.id;
+ update_accounting = true;
+ }
+ }
+ } else if(job_specs->qos) {
+ acct_qos_rec_t qos_rec;
+ if (!IS_JOB_PENDING(job_ptr))
+ error_code = ESLURM_DISABLED;
+ else {
+ info("update_job: setting qos to %s for job_id %u",
+ job_specs->qos, job_specs->job_id);
memset(&qos_rec, 0, sizeof(acct_qos_rec_t));
-
- if (strstr(job_ptr->comment, "FLAGS:PREEMPTOR"))
- qos_rec.name = "expedite";
- else if (strstr(job_ptr->comment, "FLAGS:PREEMPTEE"))
- qos_rec.name = "standby";
+ qos_rec.name = job_specs->qos;
job_ptr->qos_ptr = _determine_and_validate_qos(
job_ptr->assoc_ptr, &qos_rec, &error_code);
job_ptr->qos = qos_rec.id;
update_accounting = true;
}
- } else if(job_specs->qos) {
- acct_qos_rec_t qos_rec;
-
- info("update_job: setting qos to %s for job_id %u",
- job_specs->qos, job_specs->job_id);
-
- memset(&qos_rec, 0, sizeof(acct_qos_rec_t));
- qos_rec.name = job_specs->qos;
-
- job_ptr->qos_ptr = _determine_and_validate_qos(
- job_ptr->assoc_ptr, &qos_rec, &error_code);
- job_ptr->qos = qos_rec.id;
- update_accounting = true;
}
if (!super_user && (accounting_enforce & ACCOUNTING_ENFORCE_LIMITS) &&
@@ -6965,7 +6986,9 @@ static bool _validate_acct_policy(job_desc_msg_t *job_desc,
xassert(limit_set_max_nodes);
//(*limit_set_max_nodes) = 0;
+ /* Handle both locks here to avoid deadlock. Always do QOS first. */
slurm_mutex_lock(&assoc_mgr_qos_lock);
+ slurm_mutex_lock(&assoc_mgr_association_lock);
if(qos_ptr) {
/* for validation we don't need to look at
* qos_ptr->grp_cpu_mins.
@@ -6990,7 +7013,7 @@ static bool _validate_acct_policy(job_desc_msg_t *job_desc,
qos_ptr->grp_nodes,
qos_ptr->name);
rc = false;
- goto end_qos;
+ goto end_it;
} else if (job_desc->max_nodes == 0
|| (*limit_set_max_nodes
&& (job_desc->max_nodes
@@ -7023,7 +7046,7 @@ static bool _validate_acct_policy(job_desc_msg_t *job_desc,
qos_ptr->grp_submit_jobs,
qos_ptr->name);
rc = false;
- goto end_qos;
+ goto end_it;
}
@@ -7054,7 +7077,7 @@ static bool _validate_acct_policy(job_desc_msg_t *job_desc,
job_desc->min_nodes,
qos_ptr->max_nodes_pj);
rc = false;
- goto end_qos;
+ goto end_it;
} else if (job_desc->max_nodes == 0
|| (*limit_set_max_nodes
&& (job_desc->max_nodes
@@ -7096,7 +7119,7 @@ static bool _validate_acct_policy(job_desc_msg_t *job_desc,
job_desc->user_id,
qos_ptr->max_submit_jobs_pu);
rc = false;
- goto end_qos;
+ goto end_it;
}
}
@@ -7120,13 +7143,12 @@ static bool _validate_acct_policy(job_desc_msg_t *job_desc,
job_desc->user_id,
job_desc->time_limit, time_limit);
rc = false;
- goto end_qos;
+ goto end_it;
}
}
}
- slurm_mutex_lock(&assoc_mgr_association_lock);
while(assoc_ptr) {
/* for validation we don't need to look at
* assoc_ptr->grp_cpu_mins.
@@ -7294,8 +7316,8 @@ static bool _validate_acct_policy(job_desc_msg_t *job_desc,
assoc_ptr = assoc_ptr->parent_assoc_ptr;
parent = 1;
}
+end_it:
slurm_mutex_unlock(&assoc_mgr_association_lock);
-end_qos:
slurm_mutex_unlock(&assoc_mgr_qos_lock);
return rc;
@@ -7347,6 +7369,50 @@ extern int job_cancel_by_assoc_id(uint32_t assoc_id)
return cnt;
}
+/*
+ * job_cancel_by_qos_id - Cancel all pending and running jobs with a given
+ * QOS ID. This happens when a QOS is deleted (e.g. when
+ * a QOS is removed from the association database).
+ * RET count of cancelled jobs
+ */
+extern int job_cancel_by_qos_id(uint16_t qos_id)
+{
+ int cnt = 0;
+ ListIterator job_iterator;
+ struct job_record *job_ptr;
+ /* Write lock on jobs */
+ slurmctld_lock_t job_write_lock =
+ { NO_LOCK, WRITE_LOCK, NO_LOCK, NO_LOCK };
+
+ if (!job_list)
+ return cnt;
+
+ lock_slurmctld(job_write_lock);
+ job_iterator = list_iterator_create(job_list);
+ while ((job_ptr = (struct job_record *) list_next(job_iterator))) {
+ if (job_ptr->qos != qos_id)
+ continue;
+
+ /* move up to the parent that should still exist */
+ if(job_ptr->qos_ptr)
+ job_ptr->qos_ptr = NULL;
+
+ if(IS_JOB_FINISHED(job_ptr))
+ continue;
+
+ info("QOS deleted, cancelling job %u",
+ job_ptr->job_id);
+ /* make sure the assoc_mgr_qos_lock isn't locked before this. */
+ job_signal(job_ptr->job_id, SIGKILL, 0, 0);
+ job_ptr->state_reason = FAIL_BANK_ACCOUNT;
+ xfree(job_ptr->state_desc);
+ cnt++;
+ }
+ list_iterator_destroy(job_iterator);
+ unlock_slurmctld(job_write_lock);
+ return cnt;
+}
+
/*
* Modify the account associated with a pending job
* IN module - where this is called from
diff --git a/src/slurmctld/node_mgr.c b/src/slurmctld/node_mgr.c
index ec70dd080a774a1579100e4f280b99bc6f236779..175a991a5ef806cf5fe18abf051027be0520c2d7 100644
--- a/src/slurmctld/node_mgr.c
+++ b/src/slurmctld/node_mgr.c
@@ -760,14 +760,16 @@ int update_node ( update_node_msg_t * update_node_msg )
} /* else already fully available */
node_ptr->node_state &= (~NODE_STATE_DRAIN);
node_ptr->node_state &= (~NODE_STATE_FAIL);
- bit_set (avail_node_bitmap, node_inx);
+ if (!IS_NODE_NO_RESPOND(node_ptr))
+ bit_set (avail_node_bitmap, node_inx);
bit_set (idle_node_bitmap, node_inx);
bit_set (up_node_bitmap, node_inx);
node_ptr->last_idle = now;
reset_job_priority();
} else if (state_val == NODE_STATE_ALLOCATED) {
if (!IS_NODE_DRAIN(node_ptr) &&
- !IS_NODE_FAIL(node_ptr))
+ !IS_NODE_FAIL(node_ptr) &&
+ !IS_NODE_NO_RESPOND(node_ptr))
bit_set(avail_node_bitmap, node_inx);
bit_set (up_node_bitmap, node_inx);
bit_clear (idle_node_bitmap, node_inx);
@@ -1600,7 +1602,11 @@ extern int validate_nodes_via_front_end(
node_ptr, now);
}
} else if (IS_NODE_DOWN(node_ptr) &&
- (slurmctld_conf.ret2service == 1)) {
+ ((slurmctld_conf.ret2service == 2) ||
+ ((slurmctld_conf.ret2service == 1) &&
+ (node_ptr->reason != NULL) &&
+ (strncmp(node_ptr->reason,
+ "Not responding", 14) == 0)))) {
updated_job = true;
if (jobs_on_node) {
node_ptr->node_state =
@@ -1684,7 +1690,7 @@ static void _sync_bitmaps(struct node_record *node_ptr, int job_count)
bit_set (share_node_bitmap, node_inx);
}
if (IS_NODE_DOWN(node_ptr) || IS_NODE_DRAIN(node_ptr) ||
- IS_NODE_FAIL(node_ptr))
+ IS_NODE_FAIL(node_ptr) || IS_NODE_NO_RESPOND(node_ptr))
bit_clear (avail_node_bitmap, node_inx);
else
bit_set (avail_node_bitmap, node_inx);
diff --git a/src/slurmctld/reservation.c b/src/slurmctld/reservation.c
index 46f05a2112e22dfd89d410b632f7a5abdfeb9828..0ddf8037b36e9e1c09cc0a87ad608977a6809fe8 100644
--- a/src/slurmctld/reservation.c
+++ b/src/slurmctld/reservation.c
@@ -1157,6 +1157,9 @@ extern int create_resv(resv_desc_msg_t *resv_desc_ptr)
rc = ESLURM_INVALID_TIME_VALUE;
goto bad_parse;
}
+ } else if (resv_desc_ptr->duration == INFINITE) {
+ resv_desc_ptr->end_time = resv_desc_ptr->start_time +
+ (365 * 24 * 60 * 60);
} else if (resv_desc_ptr->duration) {
resv_desc_ptr->end_time = resv_desc_ptr->start_time +
(resv_desc_ptr->duration * 60);
@@ -1224,7 +1227,8 @@ extern int create_resv(resv_desc_msg_t *resv_desc_ptr)
}
if (resv_desc_ptr->node_cnt == NO_VAL)
resv_desc_ptr->node_cnt = 0;
- if (_resv_overlap(resv_desc_ptr->start_time,
+ if (!(resv_desc_ptr->flags & RESERVE_FLAG_OVERLAP) &&
+ _resv_overlap(resv_desc_ptr->start_time,
resv_desc_ptr->end_time,
resv_desc_ptr->flags, node_bitmap,
NULL)) {
@@ -1233,7 +1237,8 @@ extern int create_resv(resv_desc_msg_t *resv_desc_ptr)
goto bad_parse;
}
resv_desc_ptr->node_cnt = bit_set_count(node_bitmap);
- if (_job_overlap(resv_desc_ptr->start_time,
+ if (!(resv_desc_ptr->flags & RESERVE_FLAG_IGN_JOBS) &&
+ _job_overlap(resv_desc_ptr->start_time,
resv_desc_ptr->flags, node_bitmap)) {
info("Reservation request overlaps jobs");
rc = ESLURM_NODES_BUSY;
@@ -2283,19 +2288,21 @@ static int _select_nodes(resv_desc_msg_t *resv_desc_ptr,
node_bitmap = bit_copy((*part_ptr)->node_bitmap);
/* Don't use node already reserved */
- iter = list_iterator_create(resv_list);
- if (!iter)
- fatal("malloc: list_iterator_create");
- while ((resv_ptr = (slurmctld_resv_t *) list_next(iter))) {
- if ((resv_ptr->node_bitmap == NULL) ||
- (resv_ptr->start_time >= resv_desc_ptr->end_time) ||
- (resv_ptr->end_time <= resv_desc_ptr->start_time))
- continue;
- bit_not(resv_ptr->node_bitmap);
- bit_and(node_bitmap, resv_ptr->node_bitmap);
- bit_not(resv_ptr->node_bitmap);
+ if (!(resv_desc_ptr->flags & RESERVE_FLAG_OVERLAP)) {
+ iter = list_iterator_create(resv_list);
+ if (!iter)
+ fatal("malloc: list_iterator_create");
+ while ((resv_ptr = (slurmctld_resv_t *) list_next(iter))) {
+ if ((resv_ptr->node_bitmap == NULL) ||
+ (resv_ptr->start_time >= resv_desc_ptr->end_time) ||
+ (resv_ptr->end_time <= resv_desc_ptr->start_time))
+ continue;
+ bit_not(resv_ptr->node_bitmap);
+ bit_and(node_bitmap, resv_ptr->node_bitmap);
+ bit_not(resv_ptr->node_bitmap);
+ }
+ list_iterator_destroy(iter);
}
- list_iterator_destroy(iter);
/* Satisfy feature specification */
if (resv_desc_ptr->features) {
diff --git a/src/slurmctld/slurmctld.h b/src/slurmctld/slurmctld.h
index 19159834633dabf1ff2031a9fa181ca62f11c17d..bfa1107330061b7eadb29a7406563eb0f5a6fd0a 100644
--- a/src/slurmctld/slurmctld.h
+++ b/src/slurmctld/slurmctld.h
@@ -815,6 +815,14 @@ extern int job_allocate(job_desc_msg_t * job_specs, int immediate,
*/
extern int job_cancel_by_assoc_id(uint32_t assoc_id);
+/*
+ * job_cancel_by_qos_id - Cancel all pending and running jobs with a given
+ * QOS ID. This happens when a QOS is deleted (e.g. when
+ * a QOS is removed from the association database).
+ * RET count of cancelled jobs
+ */
+extern int job_cancel_by_qos_id(uint16_t qos_id);
+
/* Perform checkpoint operation on a job */
extern int job_checkpoint(checkpoint_msg_t *ckpt_ptr, uid_t uid,
slurm_fd conn_fd);
diff --git a/src/srun/opt.c b/src/srun/opt.c
index 430588a1f8bdd92135f5e1dde30649e29e38e30a..489467458f3e40e3b75a0f65ac0da46f8a08f11b 100644
--- a/src/srun/opt.c
+++ b/src/srun/opt.c
@@ -1302,6 +1302,7 @@ static void set_options(const int argc, char **argv)
opt.qos = xstrdup(optarg);
break;
case LONG_OPT_SOCKETSPERNODE:
+ max_val = 0;
get_resource_arg_range( optarg, "sockets-per-node",
&opt.min_sockets_per_node,
&max_val, true );
@@ -1310,6 +1311,7 @@ static void set_options(const int argc, char **argv)
opt.min_sockets_per_node = NO_VAL;
break;
case LONG_OPT_CORESPERSOCKET:
+ max_val = 0;
get_resource_arg_range( optarg, "cores-per-socket",
&opt.min_cores_per_socket,
&max_val, true );
@@ -1318,6 +1320,7 @@ static void set_options(const int argc, char **argv)
opt.min_cores_per_socket = NO_VAL;
break;
case LONG_OPT_THREADSPERCORE:
+ max_val = 0;
get_resource_arg_range( optarg, "threads-per-core",
&opt.min_threads_per_core,
&max_val, true );
@@ -1810,15 +1813,11 @@ static bool _opt_verify(void)
opt.nprocs = opt.min_nodes;
/* 1 proc / min_[socket * core * thread] default */
- if (opt.min_sockets_per_node > 0) {
+ if ((opt.min_sockets_per_node != NO_VAL) &&
+ (opt.min_cores_per_socket != NO_VAL) &&
+ (opt.min_threads_per_core != NO_VAL)) {
opt.nprocs *= opt.min_sockets_per_node;
- opt.nprocs_set = true;
- }
- if (opt.min_cores_per_socket > 0) {
opt.nprocs *= opt.min_cores_per_socket;
- opt.nprocs_set = true;
- }
- if (opt.min_threads_per_core > 0) {
opt.nprocs *= opt.min_threads_per_core;
opt.nprocs_set = true;
}
diff --git a/src/srun/srun.c b/src/srun/srun.c
index 8182ba6078d2a3fc8bc83a9c40fbe44ca281bbca..35684307dd7bdccabe08f6d641420b36b3dfad7f 100644
--- a/src/srun/srun.c
+++ b/src/srun/srun.c
@@ -257,6 +257,12 @@ int srun(int ac, char **av)
/* Set up slurmctld message handler */
slurmctld_msg_init();
+ /*
+ * Become --uid user
+ */
+ if (_become_user () < 0)
+ info ("Warning: Unable to assume uid=%lu\n", opt.uid);
+
/* now global "opt" should be filled in and available,
* create a job from opt
*/
@@ -279,9 +285,9 @@ int srun(int ac, char **av)
job_id = resp->job_id;
if (opt.nodes_set_env && !opt.nodes_set_opt &&
(opt.min_nodes > resp->node_cnt)) {
- /* This signifies the job used the --no-kill option
+ /* This signifies the job used the --no-kill option
* and a node went DOWN or it used a node count range
- * specification, was checkpointed from one size and
+ * specification, was checkpointed from one size and
* restarted at a different size */
error("SLURM_NNODES environment varariable "
"conflicts with allocated node count (%u!=%u).",
@@ -294,7 +300,7 @@ int srun(int ac, char **av)
opt.nprocs = opt.min_nodes;
}
if (opt.alloc_nodelist == NULL)
- opt.alloc_nodelist = xstrdup(resp->node_list);
+ opt.alloc_nodelist = xstrdup(resp->node_list);
if (opt.exclusive)
_step_opt_exclusive();
_set_cpu_env_var(resp);
@@ -351,12 +357,6 @@ int srun(int ac, char **av)
slurm_free_resource_allocation_response_msg(resp);
}
- /*
- * Become --uid user
- */
- if (_become_user () < 0)
- info ("Warning: Unable to assume uid=%lu\n", opt.uid);
-
/*
* Enhance environment for job
*/
@@ -638,7 +638,7 @@ _print_job_information(resource_allocation_response_msg_t *resp)
for (i = 0; i < resp->num_cpu_groups; i++) {
xstrfmtcat(str, "%s%u(x%u)",
sep, resp->cpus_per_node[i],
- resp->cpu_count_reps[i]);
+ resp->cpu_count_reps[i]);
sep = ",";
}
verbose("%s", str);
@@ -1418,4 +1418,3 @@ static int _setup_signals()
return rc;
}
-
diff --git a/testsuite/expect/test1.40 b/testsuite/expect/test1.40
index edf53896a7c5cc9229c8d5b22acac072bdbcbce1..4353ed23f42c578faf661c78856e6ddc9cd67fe3 100755
--- a/testsuite/expect/test1.40
+++ b/testsuite/expect/test1.40
@@ -91,7 +91,7 @@ expect {
set job_id2 $expect_out(1,string)
exp_continue
}
- -re "Account=MY_ACCT" {
+ -re "Account=my_acct" {
set match_acct 1
exp_continue
}
@@ -115,7 +115,7 @@ if {$match_acct == 0} {
set match_acct 0
spawn $scontrol show job $job_id2
expect {
- -re "Account=QA_ACCT" {
+ -re "Account=qa_acct" {
set match_acct 1
exp_continue
}
diff --git a/testsuite/expect/test1.83 b/testsuite/expect/test1.83
index 51500e2cc0ea763d548999ce4d7398b108e04442..d51929014ad7125177473eb55c83c0e801675800 100755
--- a/testsuite/expect/test1.83
+++ b/testsuite/expect/test1.83
@@ -14,7 +14,7 @@
# Change tha node name parsing logic as needed for other formats.
############################################################################
# Copyright (C) 2002-2007 The Regents of the University of California.
-# Copyright (C) 2008-2009 Lawrence Livermore National Security.
+# Copyright (C) 2008-2010 Lawrence Livermore National Security.
# Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER).
# Written by Morris Jette <jette1@llnl.gov>
# CODE-OCEC-09-009. All rights reserved.
@@ -77,12 +77,9 @@ set host_2_num 0
set timeout $max_job_delay
set srun_pid [spawn $srun -N3-3 --contiguous -l -t1 $bin_printenv SLURMD_NODENAME]
expect {
- -re "Node count specification invalid" {
- send_user "\nWARNING: can't test srun task distribution\n"
- exit $exit_code
- }
- -re "Requested node configuration is not available" {
+ -re "Node count specification invalid|configuration not available" {
send_user "\nWARNING: can't test srun task distribution\n"
+ slow_kill $srun_pid
exit $exit_code
}
-re "($number): ($alpha_numeric_under)" {
diff --git a/testsuite/expect/test15.19 b/testsuite/expect/test15.19
index 69e0ea9577c75769147d2c6befc1495bb8054307..0a2b7482e9957ac078899f90c90585a313763892 100755
--- a/testsuite/expect/test15.19
+++ b/testsuite/expect/test15.19
@@ -10,7 +10,7 @@
# anything else indicates a failure mode that must be investigated.
############################################################################
# Copyright (C) 2002-2007 The Regents of the University of California.
-# Copyright (C) 2008 Lawrence Livermore National Security.
+# Copyright (C) 2008-2010 Lawrence Livermore National Security.
# Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER).
# Written by Morris Jette <jette1@llnl.gov>
# CODE-OCEC-09-009. All rights reserved.
@@ -43,7 +43,6 @@ print_header $test_id
#
# Submit a 1 node job and validate that we don't get more than one
#
-set can_not_run 0
set job_id 0
set host_0 ""
set task_cnt 0
@@ -54,12 +53,16 @@ expect {
set job_id $expect_out(1,string)
exp_continue
}
- -re "(Task count specification invalid|configuration is not available)" {
+ -re "(Task count specification invalid|configuration not available)" {
if { [test_front_end] } {
- set can_not_run 1
send_user "\nWARNING: error expected, Testing is incompatible with front-end systems\n"
+ } else {
+ send_user "\nWARNING: test not compatable with this configuration\n"
}
- exp_continue
+ if {$job_id != 0} {
+ cancel_job $job_id
+ }
+ exit $exit_code
}
-re "($number): ($alpha_numeric_under)" {
if {$task_cnt == 0} {
@@ -80,9 +83,6 @@ expect {
wait
}
}
-if {$can_not_run == 1} {
- exit 0
-}
if {[string compare $host_0 ""] == 0} {
send_user "\nFAILURE: Did not get SLURMD_NODENAME of task 0\n"
@@ -262,12 +262,11 @@ expect {
set job_id $expect_out(1,string)
exp_continue
}
- -re "More processors requested than permitted" {
- send_user "\nWARNING: can't test salloc task distribution\n"
- exit $exit_code
- }
- -re "Node count specification invalid" {
+ -re "More processors requested than permitted|Node count specification invalid|configuration not available" {
send_user "\nWARNING: can't test salloc task distribution\n"
+ if {$job_id != 0} {
+ cancel_job $job_id
+ }
exit $exit_code
}
-re "($number): ($alpha_numeric_under)" {
diff --git a/testsuite/expect/test15.25 b/testsuite/expect/test15.25
index 9f03b947a1ef8de9b8f3d1e746594b0827f23304..fc8623de1a18ca375440889d5ddcc4c8655ab198 100755
--- a/testsuite/expect/test15.25
+++ b/testsuite/expect/test15.25
@@ -38,6 +38,7 @@ set exit_code 0
set job_id1 0
set job_id2 0
set job_acct "TEST_ACCT"
+set job_acct_lc "test_acct"
print_header $test_id
@@ -92,7 +93,7 @@ expect {
set job_id2 $expect_out(1,string)
exp_continue
}
- -re "Account=$job_acct" {
+ -re "Account=$job_acct_lc" {
set match_acct 1
exp_continue
}
@@ -119,7 +120,7 @@ if {$match_acct == 0} {
set match_acct 0
spawn $scontrol show job $job_id2
expect {
- -re "Account=$job_acct" {
+ -re "Account=$job_acct_lc" {
set match_acct 1
exp_continue
}
diff --git a/testsuite/expect/test24.1 b/testsuite/expect/test24.1
index 95717691c0734c8cd457c7932d6f2afa629feb9d..4a20ff5b575ee57d1891a94c8a6abc9962c346bb 100755
--- a/testsuite/expect/test24.1
+++ b/testsuite/expect/test24.1
@@ -91,23 +91,31 @@ expect {
incr matches
exp_continue
}
- "AccountD||60|0.600000|25|0.250000|0.250000|" {
+ "AccountD||30|0.300000|25|0.250000|0.250000|" {
incr matches
exp_continue
}
- "AccountE||25|0.250000|25|0.250000|0.250000|" {
+ "AccountE||25|0.125000|25|0.250000|0.250000|" {
incr matches
exp_continue
}
- "AccountE|User4|1|0.250000|25|0.250000|0.250000|" {
+ "AccountE|User4|1|0.125000|25|0.250000|0.250000|" {
incr matches
exp_continue
}
- "AccountF||35|0.350000|0|0.000000|0.145833|" {
+ "AccountF||35|0.175000|0|0.000000|0.145833|" {
incr matches
exp_continue
}
- "AccountF|User5|1|0.350000|0|0.000000|0.145833|" {
+ "AccountF|User5|1|0.175000|0|0.000000|0.145833|" {
+ incr matches
+ exp_continue
+ }
+ "AccountG||30|0.300000|30|0.300000|0.300000|" {
+ incr matches
+ exp_continue
+ }
+ "AccountG|User6|1|0.300000|30|0.300000|0.300000|" {
incr matches
exp_continue
}
@@ -121,7 +129,7 @@ expect {
}
}
-if {$matches != 11} {
+if {$matches != 13} {
send_user "\nFAILURE: we didn't get the correct priorities from the plugin $matches\n"
set exit_code 1
}
diff --git a/testsuite/expect/test24.1.prog.c b/testsuite/expect/test24.1.prog.c
index 1eb7df76aa092aa8a733ed59754ae4afb720fb3e..48774465cb4cc4c5bc6ee2b39b12e07b3b47965e 100644
--- a/testsuite/expect/test24.1.prog.c
+++ b/testsuite/expect/test24.1.prog.c
@@ -71,6 +71,7 @@ int _setup_assoc_list()
{
acct_update_object_t update;
acct_association_rec_t *assoc = NULL;
+
/* make the main list */
assoc_mgr_association_list = list_create(destroy_acct_association_rec);
@@ -155,7 +156,7 @@ int _setup_assoc_list()
assoc = xmalloc(sizeof(acct_association_rec_t));
assoc->id = 3;
assoc->parent_id = 1;
- assoc->shares_raw = 60;
+ assoc->shares_raw = 30;
assoc->acct = xstrdup("AccountD");
list_push(update.objects, assoc);
@@ -195,8 +196,26 @@ int _setup_assoc_list()
assoc->user = xstrdup("User5");
list_push(update.objects, assoc);
+ assoc = xmalloc(sizeof(acct_association_rec_t));
+ assoc->id = 4;
+ assoc->parent_id = 1;
+ assoc->shares_raw = 30;
+ assoc->acct = xstrdup("AccountG");
+ list_push(update.objects, assoc);
+
+ /* sub of AccountG id 4 */
+ assoc = xmalloc(sizeof(acct_association_rec_t));
+ assoc->id = 41;
+ assoc->parent_id = 4;
+ assoc->shares_raw = 1;
+ assoc->usage_raw = 30;
+ assoc->acct = xstrdup("AccountG");
+ assoc->user = xstrdup("User6");
+ list_push(update.objects, assoc);
+
assoc_mgr_update_assocs(&update);
list_destroy(update.objects);
+
return SLURM_SUCCESS;
}
@@ -250,7 +269,7 @@ int main (int argc, char **argv)
if (slurm_priority_init() != SLURM_SUCCESS)
fatal("failed to initialize priority plugin");
/* on some systems that don't have multiple cores we need to
- sleep to make sure the tread get started. */
+ sleep to make sure the thread get started. */
sleep(1);
memset(&resp, 0, sizeof(shares_response_msg_t));
resp.assoc_shares_list = assoc_mgr_get_shares(NULL, 0, NULL, NULL);
diff --git a/testsuite/expect/test4.12 b/testsuite/expect/test4.12
index d7d8ae84b63cde7f00096a7404b9e2694889c292..b657d9ba4adf515c187ae516bc0a8058b0aa3163 100755
--- a/testsuite/expect/test4.12
+++ b/testsuite/expect/test4.12
@@ -283,29 +283,7 @@ proc allocate_and_quit { node proc_cnt total_procs } {
print_header $test_id
# find the default partition
-log_user 0
-set def_part ""
-set def_part_found 0
-spawn $sinfo
-expect {
- -re "($alpha_numeric_under)\\\* " {
- set def_part $expect_out(1,string)
- set def_part_found 1
- exp_continue
- }
- timeout {
- send_user "\nFAILURE: sinfo not responding\n"
- set exit_code 1
- }
- eof {
- wait
- }
-}
-log_user 1
-if {!$def_part_found} {
- send_user "\nFAILURE: could not default partition on the system\n"
- exit 1
-}
+set def_part [default_partition]
# find the nodes in the default partition
log_user 0
@@ -357,27 +335,26 @@ if {$host_cnt == 0} {
# find me an idle node in default partition
log_user 0
+set inode_name ""
+set inode_cores_per_socket 0
set inode_procs 0
-spawn $scontrol -o show node $def_hostlist
-expect {
- -re "NodeName=($alpha_numeric_under) .*CoresPerSocket=($number) .*CPUTot=($number)(K?).* Sockets=($number) State=IDLE ThreadsPerCore=($number)" {
- set inode_name $expect_out(1,string)
- set inode_cores_per_socket $expect_out(2,string)
- set inode_procs $expect_out(3,string)
- if {[string compare $expect_out(4,string) ""]} {
- set inode_procs [expr $inode_procs * 1024]
- }
- set inode_sockets $expect_out(5,string)
- set inode_threads_per_core $expect_out(6,string)
- }
- timeout {
- send_user "\nFAILURE: scontrol not responding\n"
- set exit_code 1
- }
- eof {
- wait
+set units ""
+set inode_sockets 0
+set inode_threads_per_core 0
+
+set fd [open "|$scontrol --oneliner show node $def_hostlist"]
+exp_internal 1
+while {[gets $fd line] != -1} {
+ if {[regexp {NodeName=(\w+).*CoresPerSocket=(\d+).*CPUTot=(\d+)(K?).*Sockets=(\d+) State=IDLE ThreadsPerCore=(\d+)} $line frag inode_name inode_cores_per_socket inode_procs units inode_sockets inode_threads_per_core] == 1} {
+ break
}
}
+exp_internal 0
+if {[string compare $units ""]} {
+ set inode_procs [expr $inode_procs * 1024]
+}
+catch {close $fd}
+
log_user 1
if {!$inode_procs} {