From 44e764dae5503680a92c74b706cd61c2e5127fc8 Mon Sep 17 00:00:00 2001
From: Danny Auble <da@llnl.gov>
Date: Tue, 30 Dec 2008 22:32:37 +0000
Subject: [PATCH] svn merge -r16104:16120
https://eris.llnl.gov/svn/slurm/branches/slurm-1.3
---
doc/man/man1/sreport.1 | 4 +
doc/man/man5/slurm.conf.5 | 54 ++++++++-----
src/common/slurm_protocol_pack.c | 1 -
.../mysql/accounting_storage_mysql.c | 76 ++++++++-----------
src/slurmctld/read_config.c | 6 +-
src/sreport/common.c | 24 ++++++
src/sreport/sreport.c | 2 +
src/sreport/sreport.h | 1 +
src/sreport/user_reports.c | 69 ++++++++++++++++-
9 files changed, 168 insertions(+), 69 deletions(-)
diff --git a/doc/man/man1/sreport.1 b/doc/man/man1/sreport.1
index e60fbeecbff..219ab933c26 100644
--- a/doc/man/man1/sreport.1
+++ b/doc/man/man1/sreport.1
@@ -239,6 +239,10 @@ List of partitions jobs ran on to include in report. Default is all.
When used with the Sizes report will print number of jobs ran instead
of time used.
.TP
+.B TopCount=<OPT>
+Used in the TopUsage report. Change the number of users displayed.
+Default is 10.
+.TP
.B Users=<OPT>
List of users jobs to include in report. Default is all.
.RE
diff --git a/doc/man/man5/slurm.conf.5 b/doc/man/man5/slurm.conf.5
index 4fc62e540f2..5ed417e62b7 100644
--- a/doc/man/man5/slurm.conf.5
+++ b/doc/man/man5/slurm.conf.5
@@ -941,25 +941,45 @@ The default value is 30 seconds.
.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.
-The default value is "sched/builtin".
+Identifies the type of scheduler to be used.
+Note the \fBslurmctld\fR daemon must be restarted for a change in
+scheduler type to become effective (reconfiguring a running daemon has
+no effect for this parameter).
+The \fBscontrol\fR command can be used to manually change job priorities
+if desired.
+Acceptable values include:
+.RS
+.TP
+\fBsched/builtin\fR
+for the built\-in FIFO (First In First Out) scheduler.
+This is the default.
+.TP
+\fBsched/backfill\fR
+for a backfill scheduling module to augment the default FIFO scheduling.
Backfill scheduling will initiate lower\-priority jobs if doing
-so does not delay the expected initiation time of any higher
+so does not delay the expected initiation time of any higher
priority job.
-When initially setting the value to "sched/wiki", any pending jobs
-must have their priority set to zero (held).
-When changing the value from "sched/wiki", all pending jobs
-should have their priority change from zero to some large number.
-The \fBscontrol\fR command can be used to change job priorities.
-The \fBslurmctld\fR daemon must be restarted for a change in
-scheduler type to become effective.
+Effectiveness of backfill scheduling is dependent upon users specifying
+job time limits, otherwise all jobs will have the same time limit and
+backfilling is impossible.
+.TP
+\fBsched/gang\fR
+for gang scheduler (time\-slicing of parallel jobs). This also supports
+preemption of jobs in lower priority partitions (queues).
+See the \fIPreemption\fR web page for details:
+.br
+\fIhttps://computing.llnl.gov/linux/slurm/preempt.html\fR
+.TP
+\fBsched/hold\fR
+to hold all newly arriving jobs if a file "/etc/slurm.hold"
+exists otherwise use the built\-in FIFO scheduler
+.TP
+\fBsched/wiki\fR
+for the Wiki interface to the Maui Scheduler
+.TP
+\fBsched/wiki2\fR
+for the Wiki interface to the Moab Cluster Suite
+.RE
.TP
\fBSelectType\fR
diff --git a/src/common/slurm_protocol_pack.c b/src/common/slurm_protocol_pack.c
index 2abbe679a2f..e15137dc66d 100644
--- a/src/common/slurm_protocol_pack.c
+++ b/src/common/slurm_protocol_pack.c
@@ -2034,7 +2034,6 @@ unpack_error:
return SLURM_ERROR;
}
-
static int
_unpack_partition_info_msg(partition_info_msg_t ** msg, Buf buffer)
{
diff --git a/src/plugins/accounting_storage/mysql/accounting_storage_mysql.c b/src/plugins/accounting_storage/mysql/accounting_storage_mysql.c
index be6373d80e2..49775bb4d46 100644
--- a/src/plugins/accounting_storage/mysql/accounting_storage_mysql.c
+++ b/src/plugins/accounting_storage/mysql/accounting_storage_mysql.c
@@ -1131,6 +1131,10 @@ static int _setup_association_cond_limits(acct_association_cond_t *assoc_cond,
}
list_iterator_destroy(itr);
xstrcat(*extra, ")");
+ } else if(assoc_cond->user_list) {
+ /* we want all the users, but no non-user associations */
+ set = 1;
+ xstrfmtcat(*extra, " && (%s.user!='')", prefix);
}
if(assoc_cond->partition_list
@@ -6790,7 +6794,6 @@ extern List acct_storage_p_get_users(mysql_conn_t *mysql_conn, uid_t uid,
MYSQL_ROW row;
uint16_t private_data = 0;
acct_user_rec_t user;
- acct_wckey_cond_t *wckey_cond = NULL;
/* if this changes you will need to edit the corresponding enum */
char *user_req_inx[] = {
@@ -6925,33 +6928,6 @@ empty:
user_list = list_create(destroy_acct_user_rec);
- if(user_cond && user_cond->with_assocs) {
- /* We are going to be freeing the inners of
- this list in the user->name so we don't
- free it here
- */
- if(!user_cond->assoc_cond)
- user_cond->assoc_cond = xmalloc(
- sizeof(acct_association_cond_t));
-
- if(user_cond->assoc_cond->user_list)
- list_destroy(user_cond->assoc_cond->user_list);
- user_cond->assoc_cond->user_list = list_create(NULL);
- }
-
- if(user_cond && user_cond->with_wckeys) {
- /* We are going to be freeing the inners of
- this list in the user->name so we don't
- free it here
- */
- wckey_cond = xmalloc(sizeof(acct_wckey_cond_t));
- wckey_cond->user_list = list_create(NULL);
-
- if(user_cond->assoc_cond && user_cond->assoc_cond->cluster_list)
- wckey_cond->cluster_list =
- user_cond->assoc_cond->cluster_list;
- }
-
while((row = mysql_fetch_row(result))) {
acct_user_rec_t *user = xmalloc(sizeof(acct_user_rec_t));
/* uid_t pw_uid; */
@@ -6978,23 +6954,26 @@ empty:
if(user_cond && user_cond->with_coords)
_get_user_coords(mysql_conn, user);
-
-
- if(user_cond && user_cond->with_assocs)
- list_append(user_cond->assoc_cond->user_list,
- user->name);
-
- if(user_cond && user_cond->with_wckeys)
- list_append(wckey_cond->user_list, user->name);
}
mysql_free_result(result);
- if(user_cond && user_cond->with_assocs
- && list_count(user_cond->assoc_cond->user_list)) {
+ if(user_cond && user_cond->with_assocs) {
ListIterator assoc_itr = NULL;
acct_user_rec_t *user = NULL;
acct_association_rec_t *assoc = NULL;
- List assoc_list = acct_storage_p_get_associations(
+ List assoc_list = NULL;
+
+ /* Make sure we don't get any non-user associations
+ * this is done by at least having a user_list
+ * defined */
+ if(!user_cond->assoc_cond)
+ user_cond->assoc_cond =
+ xmalloc(sizeof(acct_association_cond_t));
+
+ if(!user_cond->assoc_cond->user_list)
+ user_cond->assoc_cond->user_list = list_create(NULL);
+
+ assoc_list = acct_storage_p_get_associations(
mysql_conn, uid, user_cond->assoc_cond);
if(!assoc_list) {
@@ -7024,15 +7003,22 @@ empty:
}
get_wckeys:
- if(wckey_cond) {
+ if(user_cond && user_cond->with_wckeys) {
ListIterator wckey_itr = NULL;
acct_user_rec_t *user = NULL;
acct_wckey_rec_t *wckey = NULL;
- List wckey_list = acct_storage_p_get_wckeys(
- mysql_conn, uid, wckey_cond);
-
- wckey_cond->cluster_list = NULL;
- destroy_acct_wckey_cond(wckey_cond);
+ List wckey_list = NULL;
+ acct_wckey_cond_t wckey_cond;
+
+ memset(&wckey_cond, 0, sizeof(acct_wckey_cond_t));
+ if(user_cond->assoc_cond) {
+ wckey_cond.user_list =
+ user_cond->assoc_cond->user_list;
+ wckey_cond.cluster_list =
+ user_cond->assoc_cond->cluster_list;
+ }
+ wckey_list = acct_storage_p_get_wckeys(
+ mysql_conn, uid, &wckey_cond);
if(!wckey_list) {
error("no wckeys");
diff --git a/src/slurmctld/read_config.c b/src/slurmctld/read_config.c
index 58e5d65134b..e7ce99fbc3e 100644
--- a/src/slurmctld/read_config.c
+++ b/src/slurmctld/read_config.c
@@ -314,7 +314,7 @@ static int _state_str2int(const char *state_str)
return state_val;
}
-#ifdef HAVE_BG
+#ifdef HAVE_3D
/* Used to get the general name of the machine, used primarily
* for bluegene systems. Not in general use because some systems
* have multiple prefix's such as foo[1-1000],bar[1-1000].
@@ -392,7 +392,7 @@ static int _build_single_nodeline_info(slurm_conf_node_t *node_ptr,
goto cleanup;
}
-#ifdef HAVE_BG
+#ifdef HAVE_3D
if (conf->node_prefix == NULL)
_set_node_prefix(node_ptr->nodenames, conf);
#endif
@@ -574,7 +574,7 @@ static int _build_all_nodeline_info(slurm_ctl_conf_t *conf)
_build_single_nodeline_info(node, config_ptr, conf);
}
xfree(highest_node_name);
-#ifdef HAVE_BG
+#ifdef HAVE_3D
{
char *node_000 = NULL;
struct node_record *node_rec = NULL;
diff --git a/src/sreport/common.c b/src/sreport/common.c
index e6fb85f3e75..4a3c691e484 100644
--- a/src/sreport/common.c
+++ b/src/sreport/common.c
@@ -433,3 +433,27 @@ extern int sort_assoc_dec(sreport_assoc_rec_t *assoc_a,
return 0;
}
+extern int get_uint(char *in_value, uint32_t *out_value, char *type)
+{
+ char *ptr = NULL, *meat = NULL;
+ long num;
+
+ if(!(meat = strip_quotes(in_value, NULL))) {
+ error("Problem with strip_quotes");
+ return SLURM_ERROR;
+ }
+ num = strtol(meat, &ptr, 10);
+ if ((num == 0) && ptr && ptr[0]) {
+ error("Invalid value for %s (%s)", type, meat);
+ xfree(meat);
+ return SLURM_ERROR;
+ }
+ xfree(meat);
+
+ if (num < 0)
+ *out_value = INFINITE; /* flag to clear */
+ else
+ *out_value = (uint32_t) num;
+ return SLURM_SUCCESS;
+}
+
diff --git a/src/sreport/sreport.c b/src/sreport/sreport.c
index 6a461afe317..4136d412754 100644
--- a/src/sreport/sreport.c
+++ b/src/sreport/sreport.c
@@ -648,6 +648,8 @@ sreport [<OPTION>] [<COMMAND>] \n\
- Group - Group all accounts together for each user.\n\
Default is a separate entry for each user\n\
and account reference. \n\
+ - TopCount=<OPT> - Used in the TopUsage report. Change the \n\
+ number of users displayed. Default is 10.\n\
- Users=<OPT> - List of users jobs to include in report. \n\
Default is all. \n\
\n\
diff --git a/src/sreport/sreport.h b/src/sreport/sreport.h
index 05ff40cfe41..a1f2dc61775 100644
--- a/src/sreport/sreport.h
+++ b/src/sreport/sreport.h
@@ -153,5 +153,6 @@ extern int sort_cluster_dec(sreport_cluster_rec_t *cluster_a,
extern int sort_assoc_dec(sreport_assoc_rec_t *assoc_a,
sreport_assoc_rec_t *assoc_b);
+extern int get_uint(char *in_value, uint32_t *out_value, char *type);
#endif /* HAVE_SREPORT_H */
diff --git a/src/sreport/user_reports.c b/src/sreport/user_reports.c
index 56e6bf12fec..784cc63fa4c 100644
--- a/src/sreport/user_reports.c
+++ b/src/sreport/user_reports.c
@@ -49,7 +49,7 @@ enum {
static List print_fields_list = NULL; /* types are of print_field_t */
static bool group_accts = false;
-static int top_limit = 10;
+static uint32_t top_limit = 10;
static int _set_cond(int *start, int argc, char *argv[],
acct_user_cond_t *user_cond, List format_list)
@@ -131,6 +131,11 @@ static int _set_cond(int *start, int argc, char *argv[],
MAX(command_len, 1))) {
assoc_cond->usage_start = parse_time(argv[i]+end, 1);
set = 1;
+ } else if (!strncasecmp (argv[i], "TopCount",
+ MAX(command_len, 1))) {
+ if (get_uint(argv[i]+end, &top_limit, "TopCount")
+ != SLURM_SUCCESS)
+ exit_code = 1;
} else {
exit_code=1;
fprintf(stderr, " Unknown condition: %s\n"
@@ -240,17 +245,20 @@ extern int user_top(int argc, char *argv[])
{
int rc = SLURM_SUCCESS;
acct_user_cond_t *user_cond = xmalloc(sizeof(acct_user_cond_t));
+ acct_cluster_cond_t cluster_cond;
ListIterator itr = NULL;
ListIterator itr2 = NULL;
ListIterator itr3 = NULL;
ListIterator cluster_itr = NULL;
List format_list = list_create(slurm_destroy_char);
List user_list = NULL;
+ List usage_cluster_list = NULL;
List cluster_list = list_create(destroy_sreport_cluster_rec);
char *object = NULL;
int i=0;
acct_user_rec_t *user = NULL;
+ acct_cluster_rec_t *cluster = NULL;
acct_association_rec_t *assoc = NULL;
acct_accounting_rec_t *assoc_acct = NULL;
sreport_user_rec_t *sreport_user = NULL;
@@ -274,6 +282,58 @@ extern int user_top(int argc, char *argv[])
fprintf(stderr, " Problem with user query.\n");
goto end_it;
}
+ /* We have to get the clusters here or we will be unable to
+ get the correct total time for the cluster if associations
+ are not enforced.
+ */
+ memset(&cluster_cond, 0, sizeof(acct_cluster_cond_t));
+ cluster_cond.with_usage = 1;
+ cluster_cond.with_deleted = 1;
+ cluster_cond.usage_end = user_cond->assoc_cond->usage_end;
+ cluster_cond.usage_start = user_cond->assoc_cond->usage_start;
+ cluster_cond.cluster_list = user_cond->assoc_cond->cluster_list;
+ usage_cluster_list = acct_storage_g_get_clusters(
+ db_conn, my_uid, &cluster_cond);
+ if(!usage_cluster_list) {
+ exit_code=1;
+ fprintf(stderr, " Problem with cluster query.\n");
+ goto end_it;
+ }
+
+ itr = list_iterator_create(usage_cluster_list);
+ while((cluster = list_next(itr))) {
+ cluster_accounting_rec_t *accting = NULL;
+
+ /* check to see if this cluster is around during the
+ time we are looking at */
+ if(!cluster->accounting_list
+ || !list_count(cluster->accounting_list))
+ continue;
+
+ sreport_cluster = xmalloc(sizeof(sreport_cluster_rec_t));
+
+ list_append(cluster_list, sreport_cluster);
+
+ sreport_cluster->name = xstrdup(cluster->name);
+ sreport_cluster->user_list =
+ list_create(destroy_sreport_user_rec);
+
+ /* get the amount of time and the average cpu count
+ during the time we are looking at */
+ cluster_itr = list_iterator_create(cluster->accounting_list);
+ while((accting = list_next(cluster_itr))) {
+ sreport_cluster->cpu_secs += accting->alloc_secs
+ + accting->down_secs + accting->idle_secs
+ + accting->resv_secs;
+ sreport_cluster->cpu_count += accting->cpu_count;
+ }
+ list_iterator_destroy(cluster_itr);
+
+ sreport_cluster->cpu_count /=
+ list_count(cluster->accounting_list);
+ }
+ list_iterator_destroy(itr);
+ list_destroy(usage_cluster_list);
if(print_fields_have_header) {
char start_char[20];
@@ -367,6 +427,9 @@ extern int user_top(int argc, char *argv[])
}
}
if(!sreport_cluster) {
+ error("This cluster '%s' hasn't "
+ "registered yet, but we have jobs "
+ "that ran?", assoc->cluster);
sreport_cluster =
xmalloc(sizeof(sreport_cluster_rec_t));
list_append(cluster_list, sreport_cluster);
@@ -399,8 +462,8 @@ extern int user_top(int argc, char *argv[])
while((assoc_acct = list_next(itr3))) {
sreport_user->cpu_secs +=
(uint64_t)assoc_acct->alloc_secs;
- sreport_cluster->cpu_secs +=
- (uint64_t)assoc_acct->alloc_secs;
+/* sreport_cluster->cpu_secs += */
+/* (uint64_t)assoc_acct->alloc_secs; */
}
list_iterator_destroy(itr3);
}
--
GitLab