From 494086b5847e0b5a61db553f1572d079fa5c3df6 Mon Sep 17 00:00:00 2001
From: Danny Auble <da@llnl.gov>
Date: Fri, 23 May 2008 21:12:21 +0000
Subject: [PATCH] added suspend time to the table along with flushing jobs at
cold start. Suspend time isn't calculated yet, but we will figure it out
---
src/common/assoc_mgr.c | 28 +-
src/common/parse_time.c | 3 +-
src/common/slurm_accounting_storage.c | 71 ++-
src/common/slurm_accounting_storage.h | 11 +
src/common/slurmdbd_defs.c | 2 +
src/common/slurmdbd_defs.h | 2 +
src/database/mysql_common.c | 7 +
src/database/mysql_common.h | 1 +
.../filetxt/accounting_storage_filetxt.c | 7 +
.../mysql/accounting_storage_mysql.c | 477 ++++++++++++++----
.../accounting_storage/mysql/mysql_rollup.c | 322 ++++++++++++
.../accounting_storage/mysql/mysql_rollup.h | 8 +
.../none/accounting_storage_none.c | 6 +
.../pgsql/accounting_storage_pgsql.c | 11 +
.../slurmdbd/accounting_storage_slurmdbd.c | 24 +-
src/sacctmgr/account_functions.c | 34 +-
src/sacctmgr/cluster_functions.c | 169 ++++---
src/sacctmgr/sacctmgr.c | 2 +-
src/sacctmgr/user_functions.c | 34 +-
src/slurmctld/controller.c | 7 +
src/slurmctld/job_mgr.c | 2 +-
src/slurmctld/node_mgr.c | 5 +-
src/slurmdbd/proc_req.c | 42 +-
23 files changed, 1041 insertions(+), 234 deletions(-)
diff --git a/src/common/assoc_mgr.c b/src/common/assoc_mgr.c
index a4af3636e24..1a35ca43221 100644
--- a/src/common/assoc_mgr.c
+++ b/src/common/assoc_mgr.c
@@ -113,11 +113,13 @@ static int _get_local_association_list(void *db_conn, int enforce)
}
list_iterator_reset(itr2);
}
- if(!assoc->user)
+ if(!assoc->user) {
continue;
+ }
passwd_ptr = getpwnam(assoc->user);
if(passwd_ptr)
assoc->uid = passwd_ptr->pw_uid;
+ //log_assoc_rec(assoc);
}
list_iterator_destroy(itr2);
list_iterator_destroy(itr);
@@ -232,8 +234,10 @@ extern int assoc_mgr_fill_in_assoc(void *db_conn, acct_association_rec_t *assoc,
if(!assoc->cluster)
assoc->cluster = local_cluster_name;
}
-/* info("looking for assoc of user=%u, acct=%s, cluster=%s, partition=%s", */
-/* assoc->uid, assoc->acct, assoc->cluster, assoc->partition); */
+/* info("looking for assoc of user=%s(%u), acct=%s, " */
+/* "cluster=%s, partition=%s", */
+/* assoc->user, assoc->uid, assoc->acct, */
+/* assoc->cluster, assoc->partition); */
slurm_mutex_lock(&local_association_lock);
itr = list_iterator_create(local_association_list);
while((found_assoc = list_next(itr))) {
@@ -244,12 +248,13 @@ extern int assoc_mgr_fill_in_assoc(void *db_conn, acct_association_rec_t *assoc,
}
continue;
} else {
- if(!assoc->user && found_assoc->user) {
+ if(!assoc->uid && found_assoc->uid) {
debug3("we are looking for a "
"nonuser association");
continue;
} else if(assoc->uid != found_assoc->uid) {
- debug3("not the right user");
+ debug3("not the right user %u != %u",
+ assoc->uid, found_assoc->uid);
continue;
}
@@ -300,11 +305,17 @@ extern int assoc_mgr_fill_in_assoc(void *db_conn, acct_association_rec_t *assoc,
assoc->cluster = ret_assoc->cluster;
if(!assoc->partition)
assoc->partition = ret_assoc->partition;
+ assoc->fairshare = ret_assoc->fairshare;
assoc->max_cpu_secs_per_job = ret_assoc->max_cpu_secs_per_job;
+ assoc->max_jobs = ret_assoc->max_jobs;
assoc->max_nodes_per_job = ret_assoc->max_nodes_per_job;
assoc->max_wall_duration_per_job = ret_assoc->max_wall_duration_per_job;
- /* The other fields are not relevant to the specific job,
- * for example max_jobs */
+ assoc->parent_acct_ptr = ret_assoc->parent_acct_ptr;
+ if(assoc->parent_acct) {
+ xfree(assoc->parent_acct);
+ assoc->parent_acct = xstrdup(ret_assoc->parent_acct);
+ } else
+ assoc->parent_acct = ret_assoc->parent_acct;
slurm_mutex_unlock(&local_association_lock);
return SLURM_SUCCESS;
@@ -481,7 +492,7 @@ extern int assoc_mgr_update_local_assocs(acct_update_object_t *update)
rc = SLURM_ERROR;
break;
}
- debug("updating the assocs here on %u", rec->id);
+ debug("updating assoc %u", rec->id);
if(object->fairshare != (uint32_t)NO_VAL) {
rec->fairshare = object->fairshare;
}
@@ -517,6 +528,7 @@ extern int assoc_mgr_update_local_assocs(acct_update_object_t *update)
parents_changed = 1;
}
+ log_assoc_rec(rec);
/* FIX ME: do more updates here */
break;
case ACCT_ADD_ASSOC:
diff --git a/src/common/parse_time.c b/src/common/parse_time.c
index 1a7f06ee32e..e72f2c93e45 100644
--- a/src/common/parse_time.c
+++ b/src/common/parse_time.c
@@ -471,7 +471,8 @@ extern int time_str2mins(char *string)
if ((string == NULL) || (string[0] == '\0'))
return -1; /* invalid input */
- if ((!strcasecmp(string, "INFINITE")) ||
+ if ((!strcasecmp(string, "-1")) ||
+ (!strcasecmp(string, "INFINITE")) ||
(!strcasecmp(string, "UNLIMITED"))) {
return INFINITE;
}
diff --git a/src/common/slurm_accounting_storage.c b/src/common/slurm_accounting_storage.c
index 9d0bbe8858d..59879e61592 100644
--- a/src/common/slurm_accounting_storage.c
+++ b/src/common/slurm_accounting_storage.c
@@ -142,6 +142,9 @@ typedef struct slurm_acct_storage_ops {
List selected_parts, void *params);
int (*update_shares_used) (void *db_conn,
List shares_used);
+ int (*flush_jobs) (void *db_conn,
+ char *cluster,
+ time_t event_time);
} slurm_acct_storage_ops_t;
typedef struct slurm_acct_storage_context {
@@ -211,7 +214,8 @@ static slurm_acct_storage_ops_t * _acct_storage_get_ops(
"jobacct_storage_p_suspend",
"jobacct_storage_p_get_jobs",
"jobacct_storage_p_archive",
- "acct_storage_p_update_shares_used"
+ "acct_storage_p_update_shares_used",
+ "acct_storage_p_flush_jobs_on_cluster"
};
int n_syms = sizeof( syms ) / sizeof( char * );
@@ -1561,20 +1565,39 @@ extern acct_admin_level_t str_2_acct_admin_level(char *level)
extern void log_assoc_rec(acct_association_rec_t *assoc_ptr)
{
- info("association rec id: %u", assoc_ptr->id);
- info(" acct : %s", assoc_ptr->acct);
- info(" cluster : %s", assoc_ptr->cluster);
- info(" fairshare : %u", assoc_ptr->fairshare);
- info(" max_cpu_secs_per_job : %u", assoc_ptr->max_cpu_secs_per_job);
- info(" max_jobs : %u", assoc_ptr->max_jobs);
- info(" max_nodes_per_job : %u", assoc_ptr->max_nodes_per_job);
- info(" max_wall_duration_per_job : %u",
- assoc_ptr->max_wall_duration_per_job);
- info(" parent_acct : %s", assoc_ptr->parent_acct);
- info(" partition : %s", assoc_ptr->partition);
- info(" user : %s(%u)", assoc_ptr->user, assoc_ptr->uid);
- info(" used_jobs : %u", assoc_ptr->used_jobs);
- info(" used_share : %u", assoc_ptr->used_share);
+ debug("association rec id : %u", assoc_ptr->id);
+ debug(" acct : %s", assoc_ptr->acct);
+ debug(" cluster : %s", assoc_ptr->cluster);
+ if(assoc_ptr->fairshare == INFINITE)
+ debug(" fairshare : NONE");
+ else
+ debug(" fairshare : %u",
+ assoc_ptr->fairshare);
+ if(assoc_ptr->max_cpu_secs_per_job == INFINITE)
+ debug(" max_cpu_secs_per_job : NONE");
+ else
+ debug(" max_cpu_secs_per_job : %d",
+ assoc_ptr->max_cpu_secs_per_job);
+ if(assoc_ptr->max_jobs == INFINITE)
+ debug(" max_jobs : NONE");
+ else
+ debug(" max_jobs : %u", assoc_ptr->max_jobs);
+ if(assoc_ptr->max_nodes_per_job == INFINITE)
+ debug(" max_nodes_per_job : NONE");
+ else
+ debug(" max_nodes_per_job : %d",
+ assoc_ptr->max_nodes_per_job);
+ if(assoc_ptr->max_wall_duration_per_job == INFINITE)
+ debug(" max_wall_duration_per_job : NONE");
+ else
+ debug(" max_wall_duration_per_job : %d",
+ assoc_ptr->max_wall_duration_per_job);
+ debug(" parent_acct : %s", assoc_ptr->parent_acct);
+ debug(" partition : %s", assoc_ptr->partition);
+ debug(" user : %s(%u)",
+ assoc_ptr->user, assoc_ptr->uid);
+ debug(" used_jobs : %u", assoc_ptr->used_jobs);
+ debug(" used_share : %u", assoc_ptr->used_share);
}
/*
@@ -1984,3 +2007,21 @@ extern int acct_storage_g_update_shares_used(void *db_conn, List acct_list)
acct_list);
}
+/*
+ * This should be called when a cluster does a cold start to flush out
+ * any jobs that were running during the restart so we don't have any
+ * jobs in the database "running" forever since no endtime will be
+ * placed in there other wise.
+ * IN: char * = cluster name
+ * RET: SLURM_SUCCESS on success SLURM_ERROR else
+ */
+extern int acct_storage_g_flush_jobs_on_cluster(
+ void *db_conn, char *cluster, time_t event_time)
+{
+ if (slurm_acct_storage_init(NULL) < 0)
+ return SLURM_ERROR;
+ return (*(g_acct_storage_context->ops.flush_jobs))
+ (db_conn, cluster, event_time);
+
+}
+
diff --git a/src/common/slurm_accounting_storage.h b/src/common/slurm_accounting_storage.h
index e737bd35ae3..1aec1b9f619 100644
--- a/src/common/slurm_accounting_storage.h
+++ b/src/common/slurm_accounting_storage.h
@@ -468,6 +468,17 @@ extern int acct_storage_g_roll_usage(void *db_conn);
*/
extern int acct_storage_g_update_shares_used(void *db_conn, List acct_list);
+/*
+ * This should be called when a cluster does a cold start to flush out
+ * any jobs that were running during the restart so we don't have any
+ * jobs in the database "running" forever since no endtime will be
+ * placed in there other wise.
+ * IN: char * = cluster name
+ * RET: SLURM_SUCCESS on success SLURM_ERROR else
+ */
+extern int acct_storage_g_flush_jobs_on_cluster(
+ void *db_conn, char *cluster, time_t event_time);
+
/*********************** CLUSTER ACCOUNTING STORAGE **************************/
extern int clusteracct_storage_g_node_down(void *db_conn,
diff --git a/src/common/slurmdbd_defs.c b/src/common/slurmdbd_defs.c
index 0fcd2eec408..9b61450b182 100644
--- a/src/common/slurmdbd_defs.c
+++ b/src/common/slurmdbd_defs.c
@@ -353,6 +353,7 @@ extern Buf pack_slurmdbd_msg(slurmdbd_msg_t *req)
buffer);
break;
case DBD_CLUSTER_PROCS:
+ case DBD_FLUSH_JOBS:
slurmdbd_pack_cluster_procs_msg(
(dbd_cluster_procs_msg_t *)req->data, buffer);
break;
@@ -467,6 +468,7 @@ extern int unpack_slurmdbd_msg(slurmdbd_msg_t *resp, Buf buffer)
(dbd_acct_coord_msg_t **)&resp->data, buffer);
break;
case DBD_CLUSTER_PROCS:
+ case DBD_FLUSH_JOBS:
rc = slurmdbd_unpack_cluster_procs_msg(
(dbd_cluster_procs_msg_t **)&resp->data, buffer);
break;
diff --git a/src/common/slurmdbd_defs.h b/src/common/slurmdbd_defs.h
index 0acd5435c43..f11647fd2a2 100644
--- a/src/common/slurmdbd_defs.h
+++ b/src/common/slurmdbd_defs.h
@@ -70,6 +70,8 @@ typedef enum {
DBD_ADD_CLUSTERS, /* Add new cluster to the mix */
DBD_ADD_USERS, /* Add new user to the mix */
DBD_CLUSTER_PROCS, /* Record total processors on cluster */
+ DBD_FLUSH_JOBS, /* End jobs that are still running
+ * when a controller is restarted. */
DBD_GET_ACCOUNTS, /* Get account information */
DBD_GET_ASSOCS, /* Get assocation information */
DBD_GET_ASSOC_USAGE, /* Get assoc usage information */
diff --git a/src/database/mysql_common.c b/src/database/mysql_common.c
index 2a159c980c4..24e4de9ec8f 100644
--- a/src/database/mysql_common.c
+++ b/src/database/mysql_common.c
@@ -304,6 +304,13 @@ extern int mysql_db_query(MYSQL *mysql_db, char *query)
return SLURM_SUCCESS;
}
+extern int mysql_db_ping(MYSQL *mysql_db)
+{
+ /* clear out the old results so we don't get a 2014 error */
+ _clear_results(mysql_db);
+ return mysql_ping(mysql_db);
+}
+
extern int mysql_db_commit(MYSQL *mysql_db)
{
//slurm_mutex_lock(&mysql_lock);
diff --git a/src/database/mysql_common.h b/src/database/mysql_common.h
index 0f2ca83d9c8..69acb54c499 100644
--- a/src/database/mysql_common.h
+++ b/src/database/mysql_common.h
@@ -81,6 +81,7 @@ extern int mysql_get_db_connection(MYSQL **mysql_db, char *db_name,
extern int mysql_close_db_connection(MYSQL **mysql_db);
extern int mysql_cleanup();
extern int mysql_db_query(MYSQL *mysql_db, char *query);
+extern int mysql_db_ping(MYSQL *mysql_db);
extern int mysql_db_commit(MYSQL *mysql_db);
extern int mysql_db_rollback(MYSQL *mysql_db);
diff --git a/src/plugins/accounting_storage/filetxt/accounting_storage_filetxt.c b/src/plugins/accounting_storage/filetxt/accounting_storage_filetxt.c
index 1be52a9415e..458305e457d 100644
--- a/src/plugins/accounting_storage/filetxt/accounting_storage_filetxt.c
+++ b/src/plugins/accounting_storage/filetxt/accounting_storage_filetxt.c
@@ -820,3 +820,10 @@ extern int acct_storage_p_update_shares_used(void *db_conn,
{
return SLURM_SUCCESS;
}
+
+extern int acct_storage_p_flush_jobs_on_cluster(
+ void *db_conn, char *cluster, time_t event_time)
+{
+ /* put end times for a clean start */
+ return SLURM_SUCCESS;
+}
diff --git a/src/plugins/accounting_storage/mysql/accounting_storage_mysql.c b/src/plugins/accounting_storage/mysql/accounting_storage_mysql.c
index 490216118f9..b367d1df1f8 100644
--- a/src/plugins/accounting_storage/mysql/accounting_storage_mysql.c
+++ b/src/plugins/accounting_storage/mysql/accounting_storage_mysql.c
@@ -110,6 +110,7 @@ char *step_table = "step_table";
char *txn_table = "txn_table";
char *user_table = "user_table";
char *last_ran_table = "last_ran_table";
+char *suspend_table = "suspend_table";
extern int acct_storage_p_commit(mysql_conn_t *mysql_conn, bool commit);
@@ -875,7 +876,6 @@ static int _mysql_acct_check_tables(MYSQL *acct_mysql_db)
{ "deleted", "tinyint default 0" },
{ "id", "int not null" },
{ "period_start", "int unsigned not null" },
- { "cpu_count", "int unsigned default 0" },
{ "alloc_cpu_secs", "bigint default 0" },
{ NULL, NULL}
};
@@ -983,6 +983,14 @@ static int _mysql_acct_check_tables(MYSQL *acct_mysql_db)
{ NULL, NULL}
};
+ storage_field_t suspend_table_fields[] = {
+ { "id", "int not null" },
+ { "associd", "mediumint not null" },
+ { "start", "int unsigned default 0 not null" },
+ { "end", "int unsigned default 0 not null" },
+ { NULL, NULL}
+ };
+
storage_field_t txn_table_fields[] = {
{ "id", "int not null auto_increment" },
{ "timestamp", "int unsigned default 0 not null" },
@@ -1124,6 +1132,11 @@ static int _mysql_acct_check_tables(MYSQL *acct_mysql_db)
", primary key (id, stepid))") == SLURM_ERROR)
return SLURM_ERROR;
+ if(mysql_db_create_table(acct_mysql_db, suspend_table,
+ suspend_table_fields,
+ ")") == SLURM_ERROR)
+ return SLURM_ERROR;
+
if(mysql_db_create_table(acct_mysql_db, txn_table, txn_table_fields,
", primary key (id))") == SLURM_ERROR)
return SLURM_ERROR;
@@ -1259,12 +1272,21 @@ extern int acct_storage_p_close_connection(mysql_conn_t **mysql_conn)
extern int acct_storage_p_commit(mysql_conn_t *mysql_conn, bool commit)
{
#ifdef HAVE_MYSQL
-
- if(!mysql_conn)
+
+ if(!mysql_conn) {
+ error("We need a connection to run this");
return SLURM_ERROR;
+ } else if(!mysql_conn->acct_mysql_db
+ || mysql_db_ping(mysql_conn->acct_mysql_db) != 0) {
+ if(!mysql_get_db_connection(&mysql_conn->acct_mysql_db,
+ mysql_db_name, mysql_db_info)) {
+ error("unable to re-connect to mysql database");
+ return SLURM_ERROR;
+ }
+ }
debug4("got %d commits", list_count(mysql_conn->update_list));
-
+
if(mysql_conn->rollback) {
if(!commit) {
if(mysql_db_rollback(mysql_conn->acct_mysql_db))
@@ -1302,7 +1324,7 @@ extern int acct_storage_p_commit(mysql_conn_t *mysql_conn, bool commit)
}
xfree(query);
while((row = mysql_fetch_row(result))) {
- //info("sending to %s(%s)", row[0], row[1]);
+ info("sending to %s(%s)", row[0], row[1]);
slurm_set_addr_char(&req.address, atoi(row[1]), row[0]);
req.msg_type = ACCOUNTING_UPDATE_MSG;
req.flags = SLURM_GLOBAL_AUTH_KEY;
@@ -1390,6 +1412,18 @@ extern int acct_storage_p_add_users(mysql_conn_t *mysql_conn, uint32_t uid,
int affect_rows = 0;
List assoc_list = list_create(destroy_acct_association_rec);
+ if(!mysql_conn) {
+ error("We need a connection to run this");
+ return SLURM_ERROR;
+ } else if(!mysql_conn->acct_mysql_db
+ || mysql_db_ping(mysql_conn->acct_mysql_db) != 0) {
+ if(!mysql_get_db_connection(&mysql_conn->acct_mysql_db,
+ mysql_db_name, mysql_db_info)) {
+ error("unable to re-connect to mysql database");
+ return SLURM_ERROR;
+ }
+ }
+
if((pw=getpwuid(uid))) {
user = pw->pw_name;
}
@@ -1521,6 +1555,18 @@ extern int acct_storage_p_add_accts(mysql_conn_t *mysql_conn, uint32_t uid,
int affect_rows = 0;
List assoc_list = list_create(destroy_acct_association_rec);
+ if(!mysql_conn) {
+ error("We need a connection to run this");
+ return SLURM_ERROR;
+ } else if(!mysql_conn->acct_mysql_db
+ || mysql_db_ping(mysql_conn->acct_mysql_db) != 0) {
+ if(!mysql_get_db_connection(&mysql_conn->acct_mysql_db,
+ mysql_db_name, mysql_db_info)) {
+ error("unable to re-connect to mysql database");
+ return SLURM_ERROR;
+ }
+ }
+
if((pw=getpwuid(uid))) {
user = pw->pw_name;
}
@@ -1638,6 +1684,18 @@ extern int acct_storage_p_add_clusters(mysql_conn_t *mysql_conn, uint32_t uid,
char *user = NULL;
int affect_rows = 0;
+ if(!mysql_conn) {
+ error("We need a connection to run this");
+ return SLURM_ERROR;
+ } else if(!mysql_conn->acct_mysql_db
+ || mysql_db_ping(mysql_conn->acct_mysql_db) != 0) {
+ if(!mysql_get_db_connection(&mysql_conn->acct_mysql_db,
+ mysql_db_name, mysql_db_info)) {
+ error("unable to re-connect to mysql database");
+ return SLURM_ERROR;
+ }
+ }
+
if((pw=getpwuid(uid))) {
user = pw->pw_name;
}
@@ -1833,6 +1891,18 @@ extern int acct_storage_p_add_associations(mysql_conn_t *mysql_conn,
return SLURM_ERROR;
}
+ if(!mysql_conn) {
+ error("We need a connection to run this");
+ return SLURM_ERROR;
+ } else if(!mysql_conn->acct_mysql_db
+ || mysql_db_ping(mysql_conn->acct_mysql_db) != 0) {
+ if(!mysql_get_db_connection(&mysql_conn->acct_mysql_db,
+ mysql_db_name, mysql_db_info)) {
+ error("unable to re-connect to mysql database");
+ return SLURM_ERROR;
+ }
+ }
+
if((pw=getpwuid(uid))) {
user = pw->pw_name;
}
@@ -2199,6 +2269,16 @@ extern List acct_storage_p_modify_users(mysql_conn_t *mysql_conn, uint32_t uid,
return NULL;
}
+ if(!mysql_conn) {
+ error("We need a connection to run this");
+ return NULL;
+ } else if(!mysql_conn->acct_mysql_db
+ || mysql_db_ping(mysql_conn->acct_mysql_db) != 0) {
+ if(!mysql_get_db_connection(&mysql_conn->acct_mysql_db,
+ mysql_db_name, mysql_db_info))
+ return NULL;
+ }
+
if((pw=getpwuid(uid))) {
user_name = pw->pw_name;
}
@@ -2322,6 +2402,18 @@ extern List acct_storage_p_modify_accts(mysql_conn_t *mysql_conn, uint32_t uid,
return NULL;
}
+ if(!mysql_conn) {
+ error("We need a connection to run this");
+ return NULL;
+ } else if(!mysql_conn->acct_mysql_db
+ || mysql_db_ping(mysql_conn->acct_mysql_db) != 0) {
+ if(!mysql_get_db_connection(&mysql_conn->acct_mysql_db,
+ mysql_db_name, mysql_db_info)) {
+ error("unable to re-connect to mysql database");
+ return NULL;
+ }
+ }
+
if((pw=getpwuid(uid))) {
user = pw->pw_name;
}
@@ -2391,6 +2483,7 @@ extern List acct_storage_p_modify_accts(mysql_conn_t *mysql_conn, uint32_t uid,
if(!(result = mysql_db_query_ret(
mysql_conn->acct_mysql_db, query, 0))) {
xfree(query);
+ xfree(vals);
return NULL;
}
xfree(query);
@@ -2444,7 +2537,7 @@ extern List acct_storage_p_modify_clusters(mysql_conn_t *mysql_conn,
List ret_list = NULL;
int rc = SLURM_SUCCESS;
char *object = NULL;
- char *vals = NULL, *assoc_vals = NULL, *extra = NULL, *query = NULL,
+ char *vals = NULL, *extra = NULL, *query = NULL,
*name_char = NULL, *assoc_char= NULL, *send_char = NULL;
time_t now = time(NULL);
struct passwd *pw = NULL;
@@ -2453,11 +2546,28 @@ extern List acct_storage_p_modify_clusters(mysql_conn_t *mysql_conn,
MYSQL_RES *result = NULL;
MYSQL_ROW row;
+ /* If you need to alter the default values of the cluster use
+ * modify_associations since this is used only for registering
+ * the controller when it loads
+ */
+
if(!cluster_q) {
error("we need something to change");
return NULL;
}
+ if(!mysql_conn) {
+ error("We need a connection to run this");
+ return NULL;
+ } else if(!mysql_conn->acct_mysql_db
+ || mysql_db_ping(mysql_conn->acct_mysql_db) != 0) {
+ if(!mysql_get_db_connection(&mysql_conn->acct_mysql_db,
+ mysql_db_name, mysql_db_info)) {
+ error("unable to re-connect to mysql database");
+ return NULL;
+ }
+ }
+
if((pw=getpwuid(uid))) {
user = pw->pw_name;
}
@@ -2485,40 +2595,7 @@ extern List acct_storage_p_modify_clusters(mysql_conn_t *mysql_conn,
xstrfmtcat(vals, ", control_port=%u", cluster->control_port);
}
- if((int)cluster->default_fairshare >= 0) {
- xstrfmtcat(assoc_vals, ", fairshare=%u",
- cluster->default_fairshare);
- } else if((int)cluster->default_fairshare == -1)
- xstrfmtcat(assoc_vals, ", fairshare=1");
-
- if((int)cluster->default_max_cpu_secs_per_job >= 0) {
- xstrfmtcat(assoc_vals, ", max_cpu_secs_per_job=%u",
- cluster->default_max_cpu_secs_per_job);
- } else if((int)cluster->default_max_cpu_secs_per_job == -1)
- xstrfmtcat(assoc_vals, ", max_cpu_secs_per_job=NULL");
-
- if((int)cluster->default_max_jobs >= 0) {
- xstrfmtcat(assoc_vals, ", max_jobs=%u",
- cluster->default_max_jobs);
- } else if((int)cluster->default_max_jobs == -1)
- xstrfmtcat(assoc_vals, ", max_jobs=NULL");
-
-
- if((int)cluster->default_max_nodes_per_job >= 0) {
- xstrfmtcat(assoc_vals, ", max_nodes_per_job=%u",
- cluster->default_max_nodes_per_job);
- } else if((int)cluster->default_max_nodes_per_job == -1)
- xstrfmtcat(assoc_vals, ", max_nodes_per_job=NULL");
-
-
- if((int)cluster->default_max_wall_duration_per_job >= 0) {
- xstrfmtcat(assoc_vals, ", max_wall_duration_per_job=%u",
- cluster->default_max_wall_duration_per_job);
- } else if((int)cluster->default_max_wall_duration_per_job == -1)
- xstrfmtcat(assoc_vals, ", max_wall_duration_per_job=NULL");
-
-
- if(!vals && !assoc_vals) {
+ if(!vals) {
error("Nothing to change");
return NULL;
}
@@ -2530,7 +2607,6 @@ extern List acct_storage_p_modify_clusters(mysql_conn_t *mysql_conn,
mysql_conn->acct_mysql_db, query, 0))) {
xfree(query);
xfree(vals);
- xfree(assoc_vals);
error("no result given for %s", extra);
return NULL;
}
@@ -2539,42 +2615,19 @@ extern List acct_storage_p_modify_clusters(mysql_conn_t *mysql_conn,
rc = 0;
ret_list = list_create(slurm_destroy_char);
while((row = mysql_fetch_row(result))) {
- acct_association_rec_t *assoc = NULL;
-
object = xstrdup(row[0]);
list_append(ret_list, object);
if(!rc) {
xstrfmtcat(name_char, "name='%s'", object);
- xstrfmtcat(assoc_char, "cluster='%s'", object);
rc = 1;
} else {
xstrfmtcat(name_char, " || name='%s'", object);
- xstrfmtcat(assoc_char, " || cluster='%s'", object);
- }
- if(assoc_vals) {
- assoc = xmalloc(sizeof(acct_association_rec_t));
- assoc->cluster = xstrdup(object);
- assoc->acct = xstrdup("root");
- assoc->fairshare = cluster->default_fairshare;
- assoc->max_jobs = cluster->default_max_jobs;
- assoc->max_nodes_per_job =
- cluster->default_max_nodes_per_job;
- assoc->max_wall_duration_per_job =
- cluster->default_max_wall_duration_per_job;
- assoc->max_cpu_secs_per_job =
- cluster->default_max_cpu_secs_per_job;
-
- if(_addto_update_list(mysql_conn->update_list,
- ACCT_MODIFY_ASSOC,
- assoc) != SLURM_SUCCESS)
- error("couldn't add to the update list");
}
}
mysql_free_result(result);
if(!list_count(ret_list)) {
debug3("didn't effect anything");
- xfree(assoc_vals);
xfree(vals);
return ret_list;
}
@@ -2591,23 +2644,10 @@ extern List acct_storage_p_modify_clusters(mysql_conn_t *mysql_conn,
}
}
- if(assoc_vals) {
- send_char = xstrdup_printf("acct='root' && (%s)",
- assoc_char);
- if(_modify_common(mysql_conn, DBD_MODIFY_CLUSTERS, now,
- user, assoc_table, send_char, assoc_vals)
- == SLURM_ERROR) {
- error("Couldn't modify cluster");
- list_destroy(ret_list);
- ret_list = NULL;
- goto end_it;
- }
- }
end_it:
xfree(name_char);
xfree(assoc_char);
xfree(vals);
- xfree(assoc_vals);
xfree(send_char);
return ret_list;
@@ -2661,6 +2701,18 @@ extern List acct_storage_p_modify_associations(mysql_conn_t *mysql_conn,
return NULL;
}
+ if(!mysql_conn) {
+ error("We need a connection to run this");
+ return NULL;
+ } else if(!mysql_conn->acct_mysql_db
+ || mysql_db_ping(mysql_conn->acct_mysql_db) != 0) {
+ if(!mysql_get_db_connection(&mysql_conn->acct_mysql_db,
+ mysql_db_name, mysql_db_info)) {
+ error("unable to re-connect to mysql database");
+ return NULL;
+ }
+ }
+
if((pw=getpwuid(uid))) {
user = pw->pw_name;
}
@@ -2934,6 +2986,18 @@ extern List acct_storage_p_remove_users(mysql_conn_t *mysql_conn, uint32_t uid,
return NULL;
}
+ if(!mysql_conn) {
+ error("We need a connection to run this");
+ return NULL;
+ } else if(!mysql_conn->acct_mysql_db
+ || mysql_db_ping(mysql_conn->acct_mysql_db) != 0) {
+ if(!mysql_get_db_connection(&mysql_conn->acct_mysql_db,
+ mysql_db_name, mysql_db_info)) {
+ error("unable to re-connect to mysql database");
+ return NULL;
+ }
+ }
+
if((pw=getpwuid(uid))) {
user_name = pw->pw_name;
}
@@ -3033,7 +3097,7 @@ extern List acct_storage_p_remove_coord(mysql_conn_t *mysql_conn, uint32_t uid,
char *acct, acct_user_cond_t *user_q)
{
#ifdef HAVE_MYSQL
- return SLURM_SUCCESS;
+ return NULL;
#else
return NULL;
#endif
@@ -3065,6 +3129,18 @@ extern List acct_storage_p_remove_accts(mysql_conn_t *mysql_conn, uint32_t uid,
user_name = pw->pw_name;
}
+ if(!mysql_conn) {
+ error("We need a connection to run this");
+ return NULL;
+ } else if(!mysql_conn->acct_mysql_db
+ || mysql_db_ping(mysql_conn->acct_mysql_db) != 0) {
+ if(!mysql_get_db_connection(&mysql_conn->acct_mysql_db,
+ mysql_db_name, mysql_db_info)) {
+ error("unable to re-connect to mysql database");
+ return NULL;
+ }
+ }
+
xstrcat(extra, "where deleted=0");
if(acct_q->acct_list && list_count(acct_q->acct_list)) {
set = 0;
@@ -3188,6 +3264,18 @@ extern List acct_storage_p_remove_clusters(mysql_conn_t *mysql_conn,
return NULL;
}
+ if(!mysql_conn) {
+ error("We need a connection to run this");
+ return NULL;
+ } else if(!mysql_conn->acct_mysql_db
+ || mysql_db_ping(mysql_conn->acct_mysql_db) != 0) {
+ if(!mysql_get_db_connection(&mysql_conn->acct_mysql_db,
+ mysql_db_name, mysql_db_info)) {
+ error("unable to re-connect to mysql database");
+ return NULL;
+ }
+ }
+
if((pw=getpwuid(uid))) {
user_name = pw->pw_name;
}
@@ -3336,6 +3424,18 @@ extern List acct_storage_p_remove_associations(mysql_conn_t *mysql_conn,
return NULL;
}
+ if(!mysql_conn) {
+ error("We need a connection to run this");
+ return NULL;
+ } else if(!mysql_conn->acct_mysql_db
+ || mysql_db_ping(mysql_conn->acct_mysql_db) != 0) {
+ if(!mysql_get_db_connection(&mysql_conn->acct_mysql_db,
+ mysql_db_name, mysql_db_info)) {
+ error("unable to re-connect to mysql database");
+ return NULL;
+ }
+ }
+
xstrcat(extra, "where id>0 && deleted=0");
if((pw=getpwuid(uid))) {
@@ -3547,6 +3647,18 @@ extern List acct_storage_p_get_users(mysql_conn_t *mysql_conn,
USER_REQ_COUNT
};
+ if(!mysql_conn) {
+ error("We need a connection to run this");
+ return NULL;
+ } else if(!mysql_conn->acct_mysql_db
+ || mysql_db_ping(mysql_conn->acct_mysql_db) != 0) {
+ if(!mysql_get_db_connection(&mysql_conn->acct_mysql_db,
+ mysql_db_name, mysql_db_info)) {
+ error("unable to re-connect to mysql database");
+ return NULL;
+ }
+ }
+
xstrcat(extra, "where deleted=0");
if(!user_q)
@@ -3712,6 +3824,18 @@ extern List acct_storage_p_get_accts(mysql_conn_t *mysql_conn,
ACCT_REQ_COUNT
};
+ if(!mysql_conn) {
+ error("We need a connection to run this");
+ return NULL;
+ } else if(!mysql_conn->acct_mysql_db
+ || mysql_db_ping(mysql_conn->acct_mysql_db) != 0) {
+ if(!mysql_get_db_connection(&mysql_conn->acct_mysql_db,
+ mysql_db_name, mysql_db_info)) {
+ error("unable to re-connect to mysql database");
+ return NULL;
+ }
+ }
+
xstrcat(extra, "where deleted=0");
if(!acct_q)
goto empty;
@@ -3885,6 +4009,18 @@ extern List acct_storage_p_get_clusters(mysql_conn_t *mysql_conn,
ASSOC_REQ_COUNT
};
+ if(!mysql_conn) {
+ error("We need a connection to run this");
+ return NULL;
+ } else if(!mysql_conn->acct_mysql_db
+ || mysql_db_ping(mysql_conn->acct_mysql_db) != 0) {
+ if(!mysql_get_db_connection(&mysql_conn->acct_mysql_db,
+ mysql_db_name, mysql_db_info)) {
+ error("unable to re-connect to mysql database");
+ return NULL;
+ }
+ }
+
xstrcat(extra, "where deleted=0");
if(!cluster_q)
@@ -4053,6 +4189,18 @@ extern List acct_storage_p_get_associations(mysql_conn_t *mysql_conn,
ASSOC2_REQ_MCPJ
};
+ if(!mysql_conn) {
+ error("We need a connection to run this");
+ return NULL;
+ } else if(!mysql_conn->acct_mysql_db
+ || mysql_db_ping(mysql_conn->acct_mysql_db) != 0) {
+ if(!mysql_get_db_connection(&mysql_conn->acct_mysql_db,
+ mysql_db_name, mysql_db_info)) {
+ error("unable to re-connect to mysql database");
+ return NULL;
+ }
+ }
+
xstrcat(extra, "where deleted=0");
if(!assoc_q)
goto empty;
@@ -4307,6 +4455,18 @@ extern int acct_storage_p_roll_usage(mysql_conn_t *mysql_conn)
UPDATE_COUNT
};
+ if(!mysql_conn) {
+ error("We need a connection to run this");
+ return SLURM_ERROR;
+ } else if(!mysql_conn->acct_mysql_db
+ || mysql_db_ping(mysql_conn->acct_mysql_db) != 0) {
+ if(!mysql_get_db_connection(&mysql_conn->acct_mysql_db,
+ mysql_db_name, mysql_db_info)) {
+ error("unable to re-connect to mysql database");
+ return SLURM_ERROR;
+ }
+ }
+
i=0;
xstrfmtcat(tmp, "%s", update_req_inx[i]);
for(i=1; i<UPDATE_COUNT; i++) {
@@ -4344,10 +4504,13 @@ extern int acct_storage_p_roll_usage(mysql_conn_t *mysql_conn)
if(rc == SLURM_ERROR)
return rc;
}
- last_hour = 1211403599;
+ last_hour = 1211475599;
+ last_day = 1211475599;
+ last_month = 1211475599;
+// last_hour = 1211403599;
// last_hour = 1206946800;
- last_day = 1207033199;
- last_month = 1204358399;
+// last_day = 1207033199;
+// last_month = 1204358399;
if(!localtime_r(&last_hour, &start_tm)) {
error("Couldn't get localtime from hour start %d", last_hour);
@@ -4469,6 +4632,18 @@ extern int clusteracct_storage_p_node_down(mysql_conn_t *mysql_conn,
char *query = NULL;
char *my_reason;
+ if(!mysql_conn) {
+ error("We need a connection to run this");
+ return SLURM_ERROR;
+ } else if(!mysql_conn->acct_mysql_db
+ || mysql_db_ping(mysql_conn->acct_mysql_db) != 0) {
+ if(!mysql_get_db_connection(&mysql_conn->acct_mysql_db,
+ mysql_db_name, mysql_db_info)) {
+ error("unable to re-connect to mysql database");
+ return SLURM_ERROR;
+ }
+ }
+
if (slurmctld_conf.fast_schedule && !slurmdbd_conf)
cpus = node_ptr->config_ptr->cpus;
else
@@ -4508,6 +4683,18 @@ extern int clusteracct_storage_p_node_up(mysql_conn_t *mysql_conn,
char* query;
int rc = SLURM_SUCCESS;
+ if(!mysql_conn) {
+ error("We need a connection to run this");
+ return SLURM_ERROR;
+ } else if(!mysql_conn->acct_mysql_db
+ || mysql_db_ping(mysql_conn->acct_mysql_db) != 0) {
+ if(!mysql_get_db_connection(&mysql_conn->acct_mysql_db,
+ mysql_db_name, mysql_db_info)) {
+ error("unable to re-connect to mysql database");
+ return SLURM_ERROR;
+ }
+ }
+
query = xstrdup_printf(
"update %s set period_end=%d where cluster='%s' "
"and period_end=0 and node_name='%s';",
@@ -4545,6 +4732,18 @@ extern int clusteracct_storage_p_cluster_procs(mysql_conn_t *mysql_conn,
}
last_procs = procs;
+ if(!mysql_conn) {
+ error("We need a connection to run this");
+ return SLURM_ERROR;
+ } else if(!mysql_conn->acct_mysql_db
+ || mysql_db_ping(mysql_conn->acct_mysql_db) != 0) {
+ if(!mysql_get_db_connection(&mysql_conn->acct_mysql_db,
+ mysql_db_name, mysql_db_info)) {
+ error("unable to re-connect to mysql database");
+ return SLURM_ERROR;
+ }
+ }
+
/* Record the processor count */
query = xstrdup_printf(
"select cpu_count from %s where cluster='%s' "
@@ -4631,13 +4830,14 @@ extern int jobacct_storage_p_job_start(mysql_conn_t *mysql_conn,
error("We need a connection to run this");
return SLURM_ERROR;
} else if(!mysql_conn->acct_mysql_db
- || mysql_ping(mysql_conn->acct_mysql_db) != 0) {
+ || mysql_db_ping(mysql_conn->acct_mysql_db) != 0) {
if(!mysql_get_db_connection(&mysql_conn->acct_mysql_db,
- mysql_db_name, mysql_db_info))
+ mysql_db_name, mysql_db_info)) {
+ error("unable to re-connect to mysql database");
return SLURM_ERROR;
+ }
}
-
debug2("mysql_jobacct_job_start() called");
priority = (job_ptr->priority == NO_VAL) ?
-1L : (long) job_ptr->priority;
@@ -4758,10 +4958,12 @@ extern int jobacct_storage_p_job_complete(mysql_conn_t *mysql_conn,
error("We need a connection to run this");
return SLURM_ERROR;
} else if(!mysql_conn->acct_mysql_db
- || mysql_ping(mysql_conn->acct_mysql_db) != 0) {
+ || mysql_db_ping(mysql_conn->acct_mysql_db) != 0) {
if(!mysql_get_db_connection(&mysql_conn->acct_mysql_db,
- mysql_db_name, mysql_db_info))
+ mysql_db_name, mysql_db_info)) {
+ error("unable to re-connect to mysql database");
return SLURM_ERROR;
+ }
}
debug2("mysql_jobacct_job_complete() called");
if (job_ptr->end_time == 0) {
@@ -4826,10 +5028,12 @@ extern int jobacct_storage_p_step_start(mysql_conn_t *mysql_conn,
error("We need a connection to run this");
return SLURM_ERROR;
} else if(!mysql_conn->acct_mysql_db
- || mysql_ping(mysql_conn->acct_mysql_db) != 0) {
+ || mysql_db_ping(mysql_conn->acct_mysql_db) != 0) {
if(!mysql_get_db_connection(&mysql_conn->acct_mysql_db,
- mysql_db_name, mysql_db_info))
+ mysql_db_name, mysql_db_info)) {
+ error("unable to re-connect to mysql database");
return SLURM_ERROR;
+ }
}
if(slurmdbd_conf) {
cpus = step_ptr->job_ptr->total_procs;
@@ -4931,10 +5135,12 @@ extern int jobacct_storage_p_step_complete(mysql_conn_t *mysql_conn,
error("We need a connection to run this");
return SLURM_ERROR;
} else if(!mysql_conn->acct_mysql_db
- || mysql_ping(mysql_conn->acct_mysql_db) != 0) {
+ || mysql_db_ping(mysql_conn->acct_mysql_db) != 0) {
if(!mysql_get_db_connection(&mysql_conn->acct_mysql_db,
- mysql_db_name, mysql_db_info))
+ mysql_db_name, mysql_db_info)) {
+ error("unable to re-connect to mysql database");
return SLURM_ERROR;
+ }
}
if(slurmdbd_conf) {
@@ -5048,17 +5254,20 @@ extern int jobacct_storage_p_suspend(mysql_conn_t *mysql_conn,
struct job_record *job_ptr)
{
#ifdef HAVE_MYSQL
- char query[1024];
+ char *query = NULL;
int rc = SLURM_SUCCESS;
-
+ bool suspended = false;
+
if(!mysql_conn) {
error("We need a connection to run this");
return SLURM_ERROR;
} else if(!mysql_conn->acct_mysql_db
- || mysql_ping(mysql_conn->acct_mysql_db) != 0) {
+ || mysql_db_ping(mysql_conn->acct_mysql_db) != 0) {
if(!mysql_get_db_connection(&mysql_conn->acct_mysql_db,
- mysql_db_name, mysql_db_info))
+ mysql_db_name, mysql_db_info)) {
+ error("unable to re-connect to mysql database");
return SLURM_ERROR;
+ }
}
if(!job_ptr->db_index) {
job_ptr->db_index = _get_db_index(mysql_conn->acct_mysql_db,
@@ -5069,20 +5278,38 @@ extern int jobacct_storage_p_suspend(mysql_conn_t *mysql_conn,
return SLURM_ERROR;
}
- snprintf(query, sizeof(query),
- "update %s set suspended=%u-suspended, state=%d "
- "where id=%u",
- job_table, (int)job_ptr->suspend_time,
- job_ptr->job_state & (~JOB_COMPLETING),
- job_ptr->db_index);
+ if (job_ptr->job_state == JOB_SUSPENDED)
+ suspended = true;
+
+ xstrfmtcat(query,
+ "update %s set suspended=%d-suspended, state=%d "
+ "where id=%u;",
+ job_table, (int)job_ptr->suspend_time,
+ job_ptr->job_state & (~JOB_COMPLETING),
+ job_ptr->db_index);
+ if(suspended)
+ xstrfmtcat(query,
+ "insert into %s (id, associd, start, end) "
+ "values (%u, %u, %d, 0);",
+ suspend_table, job_ptr->assoc_id, job_ptr->db_index,
+ (int)job_ptr->suspend_time);
+ else
+ xstrfmtcat(query,
+ "update %s set end=%d where id=%u && end=0;",
+ suspend_table, (int)job_ptr->suspend_time,
+ job_ptr->db_index);
+
rc = mysql_db_query(mysql_conn->acct_mysql_db, query);
+
+ xfree(query);
if(rc != SLURM_ERROR) {
- snprintf(query, sizeof(query),
- "update %s set suspended=%u-suspended, "
- "state=%d where id=%u and end=0",
- step_table, (int)job_ptr->suspend_time,
- job_ptr->job_state, job_ptr->db_index);
+ xstrfmtcat(query,
+ "update %s set suspended=%u-suspended, "
+ "state=%d where id=%u and end=0",
+ step_table, (int)job_ptr->suspend_time,
+ job_ptr->job_state, job_ptr->db_index);
rc = mysql_db_query(mysql_conn->acct_mysql_db, query);
+ xfree(query);
}
return rc;
@@ -5107,10 +5334,12 @@ extern List jobacct_storage_p_get_jobs(mysql_conn_t *mysql_conn,
error("We need a connection to run this");
return NULL;
} else if(!mysql_conn->acct_mysql_db
- || mysql_ping(mysql_conn->acct_mysql_db) != 0) {
+ || mysql_db_ping(mysql_conn->acct_mysql_db) != 0) {
if(!mysql_get_db_connection(&mysql_conn->acct_mysql_db,
- mysql_db_name, mysql_db_info))
+ mysql_db_name, mysql_db_info)) {
+ error("unable to re-connect to mysql database");
return NULL;
+ }
}
job_list = mysql_jobacct_process_get_jobs(mysql_conn,
selected_steps,
@@ -5132,7 +5361,7 @@ extern void jobacct_storage_p_archive(mysql_conn_t *mysql_conn,
error("We need a connection to run this");
return;
} else if(!mysql_conn->acct_mysql_db
- || mysql_ping(mysql_conn->acct_mysql_db) != 0) {
+ || mysql_db_ping(mysql_conn->acct_mysql_db) != 0) {
if(!mysql_get_db_connection(&mysql_conn->acct_mysql_db,
mysql_db_name, mysql_db_info))
return;
@@ -5143,10 +5372,40 @@ extern void jobacct_storage_p_archive(mysql_conn_t *mysql_conn,
return;
}
-extern int acct_storage_p_update_shares_used(void *db_conn,
+extern int acct_storage_p_update_shares_used(mysql_conn_t *mysql_conn,
List shares_used)
{
/* This definitely needs to be fleshed out.
* Go through the list of shares_used_object_t objects and store them */
return SLURM_SUCCESS;
}
+
+extern int acct_storage_p_flush_jobs_on_cluster(
+ mysql_conn_t *mysql_conn, char *cluster, time_t event_time)
+{
+ /* put end times for a clean start */
+ char *query = NULL;
+ int rc = SLURM_SUCCESS;
+
+ if(!mysql_conn) {
+ error("We need a connection to run this");
+ return SLURM_ERROR;
+ } else if(!mysql_conn->acct_mysql_db
+ || mysql_db_ping(mysql_conn->acct_mysql_db) != 0) {
+ if(!mysql_get_db_connection(&mysql_conn->acct_mysql_db,
+ mysql_db_name, mysql_db_info)) {
+ error("unable to re-connect to mysql database");
+ return SLURM_ERROR;
+ }
+ }
+
+ query = xstrdup_printf("update %s as t1, %s as t2 set t1.end=%u where "
+ "t2.id=t1.associd and t2.cluster='%s' "
+ "&& t1.end=0;",
+ job_table, assoc_table, event_time, cluster);
+
+ rc = mysql_db_query(mysql_conn->acct_mysql_db, query);
+ xfree(query);
+
+ return rc;
+}
diff --git a/src/plugins/accounting_storage/mysql/mysql_rollup.c b/src/plugins/accounting_storage/mysql/mysql_rollup.c
index aeda3153dd9..452d1bb9aa2 100644
--- a/src/plugins/accounting_storage/mysql/mysql_rollup.c
+++ b/src/plugins/accounting_storage/mysql/mysql_rollup.c
@@ -42,21 +42,343 @@
#ifdef HAVE_MYSQL
+typedef struct {
+ int assoc_id;
+ int a_cpu;
+} local_assoc_usage_t;
+
+typedef struct {
+ char *name;
+ int a_cpu;
+ int cpu_count;
+ int d_cpu;
+ int i_cpu;
+ int r_cpu;
+ time_t start;
+ time_t end;
+} local_cluster_usage_t;
+
+
+extern void _destroy_local_assoc_usage(void *object)
+{
+ local_assoc_usage_t *a_usage = (local_assoc_usage_t *)object;
+ if(a_usage) {
+ xfree(a_usage);
+ }
+}
+
+extern void _destroy_local_cluster_usage(void *object)
+{
+ local_cluster_usage_t *c_usage = (local_cluster_usage_t *)object;
+ if(c_usage) {
+ xfree(c_usage->name);
+ xfree(c_usage);
+ }
+}
+
extern int mysql_hourly_rollup(mysql_conn_t *mysql_conn,
time_t start, time_t end)
{
int add_sec = 3599;
+ int i=0;
time_t curr_start = start;
time_t curr_end = curr_start + add_sec;
+ char *query = NULL;
+ MYSQL_RES *result = NULL;
+ MYSQL_ROW row;
+ ListIterator a_itr = NULL;
+ ListIterator c_itr = NULL;
+ List assoc_usage_list = list_create(_destroy_local_assoc_usage);
+ List cluster_usage_list = list_create(_destroy_local_cluster_usage);
+ char *event_req_inx[] = {
+ "node_name",
+ "cluster",
+ "cpu_count",
+ "period_start",
+ "period_end"
+ };
+ char *event_str = NULL;
+ enum {
+ EVENT_REQ_NAME,
+ EVENT_REQ_CLUSTER,
+ EVENT_REQ_CPU,
+ EVENT_REQ_START,
+ EVENT_REQ_END,
+ EVENT_REQ_COUNT
+ };
+ char *job_req_inx[] = {
+ "jobid",
+ "associd",
+ "cluster",
+ "eligible",
+ "start",
+ "end",
+ "suspended",
+ "alloc_cpus",
+ "req_cpus"
+ };
+ char *job_str = NULL;
+ enum {
+ JOB_REQ_JOBID,
+ JOB_REQ_ASSOCID,
+ JOB_REQ_CLUSTER,
+ JOB_REQ_ELG,
+ JOB_REQ_START,
+ JOB_REQ_END,
+ JOB_REQ_SUSPENDED,
+ JOB_REQ_ACPU,
+ JOB_REQ_RCPU,
+ JOB_REQ_COUNT
+ };
+
+ i=0;
+ xstrfmtcat(event_str, "%s", event_req_inx[i]);
+ for(i=1; i<EVENT_REQ_COUNT; i++) {
+ xstrfmtcat(event_str, ", %s", event_req_inx[i]);
+ }
+
+ i=0;
+ xstrfmtcat(job_str, "%s", job_req_inx[i]);
+ for(i=1; i<JOB_REQ_COUNT; i++) {
+ xstrfmtcat(job_str, ", %s", job_req_inx[i]);
+ }
/* info("begin start %s", ctime(&curr_start)); */
/* info("begin end %s", ctime(&curr_end)); */
+ a_itr = list_iterator_create(cluster_usage_list);
+ c_itr = list_iterator_create(cluster_usage_list);
while(curr_start < end) {
+ int last_id = 0;
+ local_cluster_usage_t *c_usage = NULL;
+ local_assoc_usage_t *a_usage = NULL;
+
+ // first get the events during this time
+ query = xstrdup_printf("select %s from %s where "
+ "(period_start <= %d "
+ "&& period_end >= %d) "
+ "|| period_end = 0 "
+ "order by node_name, period_start",
+ event_str, event_table,
+ curr_end, curr_start);
+
+ debug3("%d query\n%s", mysql_conn->conn, query);
+ if(!(result = mysql_db_query_ret(
+ mysql_conn->acct_mysql_db, query, 0))) {
+ xfree(query);
+ return SLURM_ERROR;
+ }
+ xfree(query);
+
+ while((row = mysql_fetch_row(result))) {
+ int row_start = atoi(row[EVENT_REQ_START]);
+ int row_end = atoi(row[EVENT_REQ_END]);
+ int row_cpu = atoi(row[EVENT_REQ_CPU]);
+
+ if(row_start < curr_start)
+ row_start = curr_start;
+
+ if(!row_end)
+ row_end = curr_end;
+ if(!row[EVENT_REQ_NAME][0]) {
+ c_usage =
+ xmalloc(sizeof(local_cluster_usage_t));
+ c_usage->name = xstrdup(row[EVENT_REQ_CLUSTER]);
+ c_usage->cpu_count = row_cpu;
+ c_usage->start = row_start;
+ c_usage->end = row_end;
+ list_append(cluster_usage_list, c_usage);
+ continue;
+ }
+
+ list_iterator_reset(c_itr);
+ while((c_usage = list_next(c_itr))) {
+ if(!strcmp(c_usage->name,
+ row[EVENT_REQ_CLUSTER])) {
+ int local_start = row_start;
+ int local_end = row_end;
+ if(c_usage->start > local_start)
+ local_start = c_usage->start;
+ if(c_usage->end < local_end)
+ local_end = c_usage->end;
+
+ if((local_end - local_start) < 1)
+ continue;
+
+ info("node %s adds (%d)(%d-%d) * %d = %d "
+ "to %d",
+ row[EVENT_REQ_NAME],
+ (local_end - local_start)+1,
+ local_end, local_start,
+ row_cpu,
+ ((local_end - local_start)+1)
+ * row_cpu,
+ row_cpu);
+ /* need to add 1 sec to the
+ subtraction to get the
+ total time */
+ c_usage->d_cpu +=
+ ((local_end - local_start) + 1)
+ * row_cpu;
+
+ /* don't break here just
+ incase the cpu count changed during
+ this time period.
+ */
+ }
+ }
+ }
+ mysql_free_result(result);
+
+ query = xstrdup_printf("select %s from %s, %s as t2 where "
+ "((eligible <= %d && end >= %d) "
+ "|| end = 0 || start = 0) "
+ "&& associd=t2.id "
+ "order by associd, eligible",
+ job_str, job_table, assoc_table,
+ curr_end, curr_start);
+
+ debug3("%d query\n%s", mysql_conn->conn, query);
+ if(!(result = mysql_db_query_ret(
+ mysql_conn->acct_mysql_db, query, 0))) {
+ xfree(query);
+ return SLURM_ERROR;
+ }
+ xfree(query);
+
+
+ while((row = mysql_fetch_row(result))) {
+ int job_id = atoi(row[JOB_REQ_ASSOCID]);
+ int assoc_id = atoi(row[JOB_REQ_ASSOCID]);
+ int row_eligible = atoi(row[JOB_REQ_ELG]);
+ int row_start = atoi(row[JOB_REQ_START]);
+ int row_end = atoi(row[JOB_REQ_END]);
+ int row_acpu = atoi(row[JOB_REQ_ACPU]);
+ int row_rcpu = atoi(row[JOB_REQ_RCPU]);
+
+ if(row_start && (row_start < curr_start))
+ row_start = curr_start;
+ if(!row_start && row_end)
+ row_start = row_end;
+ if(!row_end)
+ row_end = curr_end;
+
+ if(last_id != assoc_id) {
+ a_usage =
+ xmalloc(sizeof(local_cluster_usage_t));
+ a_usage->assoc_id = assoc_id;
+ list_append(assoc_usage_list, a_usage);
+ last_id = assoc_id;
+ }
+
+ if(row_start) {
+ a_usage->a_cpu +=
+ (row_end - row_start)
+ * row_acpu;
+ }
+ list_iterator_reset(c_itr);
+ while((c_usage = list_next(c_itr))) {
+ if(!strcmp(c_usage->name,
+ row[JOB_REQ_CLUSTER])) {
+ int local_start = row_start;
+ int local_end = row_end;
+ if(!local_start)
+ goto calc_resv;
+
+ if(c_usage->start > local_start)
+ local_start = c_usage->start;
+ if(c_usage->end < local_end)
+ local_end = c_usage->end;
+
+ if((local_end - local_start) < 1)
+ goto calc_resv;
+ info("%d assoc %d adds (%d)(%d-%d) * %d = %d "
+ "to %d",
+ job_id,
+ assoc_id,
+ local_end - local_start,
+ local_end, local_start,
+ row_acpu,
+ (local_end - local_start)
+ * row_acpu,
+ row_acpu);
+ c_usage->a_cpu +=
+ (local_end - local_start)
+ * row_acpu;
+ calc_resv:
+ /* now reserved time */
+ if(row_start < c_usage->start)
+ continue;
+ local_start = row_eligible;
+ local_end = row_start;
+ if(c_usage->start > local_start)
+ local_start = c_usage->start;
+ if(c_usage->end < local_end)
+ local_end = c_usage->end;
+
+ if((local_end - local_start) < 1)
+ continue;
+
+ info("%d assoc %d reserved (%d)(%d-%d) * %d = %d "
+ "to %d",
+ job_id,
+ assoc_id,
+ (local_end - local_start),
+ local_end, local_start,
+ row_rcpu,
+ (local_end - local_start)
+ * row_rcpu,
+ row_rcpu);
+ c_usage->r_cpu +=
+ (local_end - local_start)
+ * row_rcpu;
+
+
+ /* don't break here just
+ incase the cpu count changed during
+ this time period.
+ */
+ }
+ }
+ }
+ mysql_free_result(result);
+
+ list_iterator_reset(c_itr);
+ while((c_usage = list_next(c_itr))) {
+ int total_time = ((curr_end - curr_start) + 1)
+ * c_usage->cpu_count;
+
+ c_usage->i_cpu = total_time - c_usage->a_cpu -
+ c_usage->d_cpu - c_usage->r_cpu;
+ if(c_usage->i_cpu < 0) {
+ c_usage->r_cpu += c_usage->i_cpu;
+ c_usage->i_cpu = 0;
+ if(c_usage->r_cpu < 0)
+ c_usage->r_cpu = 0;
+ }
+
+ info("cluster %s(%u) down %u alloc %u "
+ "resv %u idle %u total= %u from %s", c_usage->name,
+ c_usage->cpu_count, c_usage->d_cpu, c_usage->a_cpu,
+ c_usage->r_cpu, c_usage->i_cpu,
+ c_usage->d_cpu + c_usage->a_cpu +
+ c_usage->r_cpu + c_usage->i_cpu,
+ ctime(&c_usage->start));
+ info("to %s", ctime(&c_usage->end));
+ }
+ list_flush(assoc_usage_list);
+ list_flush(cluster_usage_list);
curr_start = curr_end+1;
curr_end = curr_start + add_sec;
debug3("curr hour is now %d-%d", curr_start, curr_end);
}
+ xfree(event_str);
+ xfree(job_str);
+ list_iterator_destroy(a_itr);
+ list_iterator_destroy(c_itr);
+
+ list_destroy(assoc_usage_list);
+ list_destroy(cluster_usage_list);
/* info("stop start %s", ctime(&curr_start)); */
/* info("stop end %s", ctime(&curr_end)); */
return SLURM_SUCCESS;
diff --git a/src/plugins/accounting_storage/mysql/mysql_rollup.h b/src/plugins/accounting_storage/mysql/mysql_rollup.h
index d04db2387b1..199e30d802b 100644
--- a/src/plugins/accounting_storage/mysql/mysql_rollup.h
+++ b/src/plugins/accounting_storage/mysql/mysql_rollup.h
@@ -44,6 +44,14 @@
#include "mysql_jobacct_process.h"
#ifdef HAVE_MYSQL
+extern char *assoc_table;
+extern char *assoc_day_table;
+extern char *assoc_hour_table;
+extern char *assoc_month_table;
+extern char *cluster_day_table;
+extern char *cluster_hour_table;
+extern char *cluster_month_table;
+extern char *event_table;
extern int mysql_hourly_rollup(mysql_conn_t *mysql_conn,
time_t start, time_t end);
diff --git a/src/plugins/accounting_storage/none/accounting_storage_none.c b/src/plugins/accounting_storage/none/accounting_storage_none.c
index 29deeafbe2e..24de0dd251d 100644
--- a/src/plugins/accounting_storage/none/accounting_storage_none.c
+++ b/src/plugins/accounting_storage/none/accounting_storage_none.c
@@ -340,3 +340,9 @@ extern int acct_storage_p_update_shares_used(void *db_conn,
{
return SLURM_SUCCESS;
}
+
+extern int acct_storage_p_flush_jobs_on_cluster(
+ void *db_conn, char *cluster, time_t event_time)
+{
+ return SLURM_SUCCESS;
+}
diff --git a/src/plugins/accounting_storage/pgsql/accounting_storage_pgsql.c b/src/plugins/accounting_storage/pgsql/accounting_storage_pgsql.c
index 6b732283a28..0c7c390ab16 100644
--- a/src/plugins/accounting_storage/pgsql/accounting_storage_pgsql.c
+++ b/src/plugins/accounting_storage/pgsql/accounting_storage_pgsql.c
@@ -1498,5 +1498,16 @@ extern int acct_storage_p_update_shares_used(void *db_conn,
{
/* This definitely needs to be fleshed out.
* Go through the list of shares_used_object_t objects and store them */
+ return SLURM_SUCCESS;
+}
+
+extern int acct_storage_p_flush_jobs_on_cluster(
+ void *db_conn, char *cluster, time_t event_time)
+{
+ /* put end times for a clean start */
+
+
+
+
return SLURM_SUCCESS;
}
diff --git a/src/plugins/accounting_storage/slurmdbd/accounting_storage_slurmdbd.c b/src/plugins/accounting_storage/slurmdbd/accounting_storage_slurmdbd.c
index b8dce12a15a..2ba67e2f42d 100644
--- a/src/plugins/accounting_storage/slurmdbd/accounting_storage_slurmdbd.c
+++ b/src/plugins/accounting_storage/slurmdbd/accounting_storage_slurmdbd.c
@@ -357,8 +357,8 @@ extern List acct_storage_p_modify_clusters(void *db_conn, uint32_t uid,
}
extern List acct_storage_p_modify_associations(void *db_conn, uint32_t uid,
- acct_association_cond_t *assoc_q,
- acct_association_rec_t *assoc)
+ acct_association_cond_t *assoc_q,
+ acct_association_rec_t *assoc)
{
slurmdbd_msg_t req;
dbd_modify_msg_t get_msg;
@@ -1203,3 +1203,23 @@ extern int acct_storage_p_update_shares_used(void *db_conn,
return rc;
}
+
+extern int acct_storage_p_flush_jobs_on_cluster(void *db_conn, char *cluster,
+ time_t event_time)
+{
+ slurmdbd_msg_t msg;
+ dbd_cluster_procs_msg_t req;
+
+ info("Ending any jobs in accounting that were running when controller "
+ "went down on cluster %s", cluster);
+ req.cluster_name = cluster;
+ req.proc_count = 0;
+ req.event_time = event_time;
+ msg.msg_type = DBD_FLUSH_JOBS;
+ msg.data = &req;
+
+ if (slurm_send_slurmdbd_msg(&msg) < 0)
+ return SLURM_ERROR;
+
+ return SLURM_SUCCESS;
+}
diff --git a/src/sacctmgr/account_functions.c b/src/sacctmgr/account_functions.c
index a09912fe1f8..ed0e3d916b9 100644
--- a/src/sacctmgr/account_functions.c
+++ b/src/sacctmgr/account_functions.c
@@ -142,13 +142,10 @@ static int _set_rec(int *start, int argc, char *argv[],
a_set = 1;
} else if (strncasecmp (argv[i], "MaxWall", 4) == 0) {
mins = time_str2mins(argv[i]+end);
- if (mins >= 0) {
+ if (mins != NO_VAL) {
assoc->max_wall_duration_per_job
= (uint32_t) mins;
a_set = 1;
- } else if (strcmp(argv[i]+end, "-1") == 0) {
- assoc->max_wall_duration_per_job = -1;
- a_set = 1;
} else {
printf(" Bad MaxWall time format: %s\n",
argv[i]);
@@ -297,12 +294,9 @@ extern int sacctmgr_add_account(int argc, char *argv[])
limit_set = 1;
} else if (strncasecmp (argv[i], "MaxWall", 4) == 0) {
mins = time_str2mins(argv[i]+end);
- if (mins >= 0) {
+ if (mins != NO_VAL) {
max_wall_duration_per_job = (uint32_t) mins;
limit_set = 1;
- } else if (strcmp(argv[i]+end, "-1") == 0) {
- max_wall_duration_per_job = -1;
- limit_set = 1;
} else {
printf(" Bad MaxWall time format: %s\n",
argv[i]);
@@ -561,16 +555,30 @@ extern int sacctmgr_add_account(int argc, char *argv[])
if(limit_set) {
printf(" Settings\n");
- if(fairshare != NO_VAL)
+ if(fairshare == INFINITE)
+ printf(" Fairshare = NONE\n");
+ else if(fairshare != NO_VAL)
printf(" Fairshare = %u\n", fairshare);
- if(max_cpu_secs_per_job != NO_VAL)
+
+ if(max_cpu_secs_per_job == INFINITE)
+ printf(" MaxCPUSecs = NONE\n");
+ else if(max_cpu_secs_per_job != NO_VAL)
printf(" MaxCPUSecs = %u\n",
max_cpu_secs_per_job);
- if(max_jobs != NO_VAL)
+
+ if(max_jobs == INFINITE)
+ printf(" MaxJobs = NONE\n");
+ else if(max_jobs != NO_VAL)
printf(" MaxJobs = %u\n", max_jobs);
- if(max_nodes_per_job != NO_VAL)
+
+ if(max_nodes_per_job == INFINITE)
+ printf(" MaxNodes = NONE\n");
+ else if(max_nodes_per_job != NO_VAL)
printf(" MaxNodes = %u\n", max_nodes_per_job);
- if(max_wall_duration_per_job != NO_VAL) {
+
+ if(max_wall_duration_per_job == INFINITE)
+ printf(" MaxWall = NONE\n");
+ else if(max_wall_duration_per_job != NO_VAL) {
char time_buf[32];
mins2time_str((time_t) max_wall_duration_per_job,
time_buf, sizeof(time_buf));
diff --git a/src/sacctmgr/cluster_functions.c b/src/sacctmgr/cluster_functions.c
index c53ed62de49..8a099cbce44 100644
--- a/src/sacctmgr/cluster_functions.c
+++ b/src/sacctmgr/cluster_functions.c
@@ -41,7 +41,7 @@
#include "src/sacctmgr/print.h"
static int _set_cond(int *start, int argc, char *argv[],
- acct_cluster_cond_t *cluster_cond,
+ List cluster_list,
List format_list)
{
int i;
@@ -54,13 +54,13 @@ static int _set_cond(int *start, int argc, char *argv[],
i--;
break;
} else if(!end) {
- addto_char_list(cluster_cond->cluster_list, argv[i]);
+ addto_char_list(cluster_list, argv[i]);
set = 1;
} else if (strncasecmp (argv[i], "Format", 1) == 0) {
if(format_list)
addto_char_list(format_list, argv[i]+end);
} else if (strncasecmp (argv[i], "Names", 1) == 0) {
- addto_char_list(cluster_cond->cluster_list,
+ addto_char_list(cluster_list,
argv[i]+end);
set = 1;
} else {
@@ -74,7 +74,7 @@ static int _set_cond(int *start, int argc, char *argv[],
}
static int _set_rec(int *start, int argc, char *argv[],
- acct_cluster_rec_t *cluster)
+ acct_association_rec_t *assoc)
{
int i, mins;
int set = 0;
@@ -89,34 +89,31 @@ static int _set_rec(int *start, int argc, char *argv[],
printf(" Bad format on %s: End your option with "
"an '=' sign\n", argv[i]);
} else if (strncasecmp (argv[i], "FairShare", 1) == 0) {
- if (get_uint(argv[i]+end, &cluster->default_fairshare,
+ if (get_uint(argv[i]+end, &assoc->fairshare,
"FairShare") == SLURM_SUCCESS)
set = 1;
} else if (strncasecmp (argv[i], "MaxJobs", 4) == 0) {
- if (get_uint(argv[i]+end, &cluster->default_max_jobs,
+ if (get_uint(argv[i]+end, &assoc->max_jobs,
"MaxJobs") == SLURM_SUCCESS)
set = 1;
} else if (strncasecmp (argv[i], "MaxNodes", 4) == 0) {
if (get_uint(argv[i]+end,
- &cluster->default_max_nodes_per_job,
+ &assoc->max_nodes_per_job,
"MaxNodes") == SLURM_SUCCESS)
set = 1;
} else if (strncasecmp (argv[i], "MaxWall", 4) == 0) {
mins = time_str2mins(argv[i]+end);
- if (mins >= 0) {
- cluster->default_max_wall_duration_per_job
+ if (mins != NO_VAL) {
+ assoc->max_wall_duration_per_job
= (uint32_t) mins;
set = 1;
- } else if (strcmp(argv[i]+end, "-1") == 0) {
- cluster->default_max_wall_duration_per_job = -1;
- set = 1;
} else {
printf(" Bad MaxWall time format: %s\n",
argv[i]);
}
} else if (strncasecmp (argv[i], "MaxCPUSecs", 4) == 0) {
if (get_uint(argv[i]+end,
- &cluster->default_max_cpu_secs_per_job,
+ &assoc->max_cpu_secs_per_job,
"MaxCPUSecs") == SLURM_SUCCESS)
set = 1;
} else {
@@ -154,7 +151,7 @@ extern int sacctmgr_add_cluster(int argc, char *argv[])
} else if (strncasecmp (argv[i], "FairShare", 1) == 0) {
fairshare = atoi(argv[i]+end);
limit_set = 1;
- } else if (strncasecmp (argv[i], "MaxCPUSecs4", 4) == 0) {
+ } else if (strncasecmp (argv[i], "MaxCPUSecs", 4) == 0) {
max_cpu_secs_per_job = atoi(argv[i]+end);
limit_set = 1;
} else if (strncasecmp (argv[i], "MaxJobs=", 4) == 0) {
@@ -165,12 +162,9 @@ extern int sacctmgr_add_cluster(int argc, char *argv[])
limit_set = 1;
} else if (strncasecmp (argv[i], "MaxWall", 4) == 0) {
mins = time_str2mins(argv[i]+end);
- if (mins >= 0) {
+ if (mins != NO_VAL) {
max_wall_duration_per_job = (uint32_t) mins;
limit_set = 1;
- } else if (strcmp(argv[i]+end, "-1") == 0) {
- max_wall_duration_per_job = -1;
- limit_set = 1;
} else {
printf(" Bad MaxWall time format: %s\n",
argv[i]);
@@ -248,16 +242,30 @@ extern int sacctmgr_add_cluster(int argc, char *argv[])
if(limit_set) {
printf(" User Defaults\n");
- if(fairshare != NO_VAL)
+ if(fairshare == INFINITE)
+ printf(" Fairshare = NONE\n");
+ else if(fairshare != NO_VAL)
printf(" Fairshare = %u\n", fairshare);
- if(max_cpu_secs_per_job != NO_VAL)
+
+ if(max_cpu_secs_per_job == INFINITE)
+ printf(" MaxCPUSecs = NONE\n");
+ else if(max_cpu_secs_per_job != NO_VAL)
printf(" MaxCPUSecs = %u\n",
max_cpu_secs_per_job);
- if(max_jobs != NO_VAL)
+
+ if(max_jobs == INFINITE)
+ printf(" MaxJobs = NONE\n");
+ else if(max_jobs != NO_VAL)
printf(" MaxJobs = %u\n", max_jobs);
- if(max_nodes_per_job != NO_VAL)
+
+ if(max_nodes_per_job == INFINITE)
+ printf(" MaxNodes = NONE\n");
+ else if(max_nodes_per_job != NO_VAL)
printf(" MaxNodes = %u\n", max_nodes_per_job);
- if(max_wall_duration_per_job != NO_VAL) {
+
+ if(max_wall_duration_per_job == INFINITE)
+ printf(" MaxWall = NONE\n");
+ else if(max_wall_duration_per_job != NO_VAL) {
char time_buf[32];
mins2time_str((time_t) max_wall_duration_per_job,
time_buf, sizeof(time_buf));
@@ -319,7 +327,7 @@ extern int sacctmgr_list_cluster(int argc, char *argv[])
cluster_cond->cluster_list = list_create(slurm_destroy_char);
- _set_cond(&i, argc, argv, cluster_cond, format_list);
+ _set_cond(&i, argc, argv, cluster_cond->cluster_list, format_list);
cluster_list = acct_storage_g_get_clusters(db_conn, cluster_cond);
destroy_acct_cluster_cond(cluster_cond);
@@ -455,47 +463,55 @@ extern int sacctmgr_modify_cluster(int argc, char *argv[])
{
int rc = SLURM_SUCCESS;
int i=0;
- acct_cluster_rec_t *cluster = xmalloc(sizeof(acct_cluster_rec_t));
- acct_cluster_cond_t *cluster_cond =
- xmalloc(sizeof(acct_cluster_cond_t));
- List cluster_list = NULL;
- int cond_set = 0, rec_set = 0;
+ acct_association_rec_t *assoc = xmalloc(sizeof(acct_association_rec_t));
+ acct_association_cond_t *assoc_cond =
+ xmalloc(sizeof(acct_association_cond_t));
+ int cond_set = 0, rec_set = 0, set = 0;
List ret_list = NULL;
- cluster_cond->cluster_list = list_create(slurm_destroy_char);
-
- cluster->default_fairshare = -2;
- cluster->default_max_cpu_secs_per_job = -2;
- cluster->default_max_jobs = -2;
- cluster->default_max_nodes_per_job = -2;
- cluster->default_max_wall_duration_per_job = -2;
+ assoc_cond = xmalloc(sizeof(acct_association_cond_t));
+ assoc_cond->cluster_list = list_create(slurm_destroy_char);
+ assoc_cond->acct_list = list_create(NULL);
+ assoc_cond->fairshare = NO_VAL;
+ assoc_cond->max_cpu_secs_per_job = NO_VAL;
+ assoc_cond->max_jobs = NO_VAL;
+ assoc_cond->max_nodes_per_job = NO_VAL;
+ assoc_cond->max_wall_duration_per_job = NO_VAL;
+
+ assoc->fairshare = NO_VAL;
+ assoc->max_cpu_secs_per_job = NO_VAL;
+ assoc->max_jobs = NO_VAL;
+ assoc->max_nodes_per_job = NO_VAL;
+ assoc->max_wall_duration_per_job = NO_VAL;
for (i=0; i<argc; i++) {
if (strncasecmp (argv[i], "Where", 5) == 0) {
i++;
- if(_set_cond(&i, argc, argv, cluster_cond, NULL))
+ if(_set_cond(&i, argc, argv,
+ assoc_cond->cluster_list, NULL))
cond_set = 1;
} else if (strncasecmp (argv[i], "Set", 3) == 0) {
i++;
- if(_set_rec(&i, argc, argv, cluster))
+ if(_set_rec(&i, argc, argv, assoc))
rec_set = 1;
} else {
- if(_set_cond(&i, argc, argv, cluster_cond, NULL))
+ if(_set_cond(&i, argc, argv,
+ assoc_cond->cluster_list, NULL))
cond_set = 1;
}
}
if(!rec_set) {
printf(" You didn't give me anything to set\n");
- destroy_acct_cluster_rec(cluster);
- destroy_acct_cluster_cond(cluster_cond);
+ destroy_acct_association_rec(assoc);
+ destroy_acct_association_cond(assoc_cond);
return SLURM_ERROR;
} else if(!cond_set) {
if(!commit_check("You didn't set any conditions with 'WHERE'.\n"
"Are you sure you want to continue?")) {
printf("Aborted\n");
- destroy_acct_cluster_rec(cluster);
- destroy_acct_cluster_cond(cluster_cond);
+ destroy_acct_association_rec(assoc);
+ destroy_acct_association_cond(assoc);
return SLURM_SUCCESS;
}
}
@@ -503,45 +519,53 @@ extern int sacctmgr_modify_cluster(int argc, char *argv[])
printf(" Setting\n");
if(rec_set)
printf(" User Defaults =\n");
- if(cluster->default_fairshare != NO_VAL)
- printf(" Fairshare = %u\n", cluster->default_fairshare);
- if(cluster->default_max_cpu_secs_per_job != NO_VAL)
+ if(assoc->fairshare == INFINITE)
+ printf(" Fairshare = NONE\n");
+ else if(assoc->fairshare != NO_VAL)
+ printf(" Fairshare = %u\n", assoc->fairshare);
+
+ if(assoc->max_cpu_secs_per_job == INFINITE)
+ printf(" MaxCPUSecs = NONE\n");
+ else if(assoc->max_cpu_secs_per_job != NO_VAL)
printf(" MaxCPUSecs = %u\n",
- cluster->default_max_cpu_secs_per_job);
- if(cluster->default_max_jobs != NO_VAL)
- printf(" MaxJobs = %u\n", cluster->default_max_jobs);
- if(cluster->default_max_nodes_per_job != NO_VAL)
+ assoc->max_cpu_secs_per_job);
+
+ if(assoc->max_jobs == INFINITE)
+ printf(" MaxJobs = NONE\n");
+ else if(assoc->max_jobs != NO_VAL)
+ printf(" MaxJobs = %u\n", assoc->max_jobs);
+
+ if(assoc->max_nodes_per_job == INFINITE)
+ printf(" MaxNodes = NONE\n");
+ else if(assoc->max_nodes_per_job != NO_VAL)
printf(" MaxNodes = %u\n",
- cluster->default_max_nodes_per_job);
- if(cluster->default_max_wall_duration_per_job != NO_VAL) {
+ assoc->max_nodes_per_job);
+
+ if(assoc->max_wall_duration_per_job == INFINITE)
+ printf(" MaxWall = NONE\n");
+ else if(assoc->max_wall_duration_per_job != NO_VAL) {
char time_buf[32];
mins2time_str((time_t)
- cluster->default_max_wall_duration_per_job,
+ assoc->max_wall_duration_per_job,
time_buf, sizeof(time_buf));
printf(" MaxWall = %s\n", time_buf);
}
- cluster_list = list_create(destroy_acct_cluster_rec);
- list_append(cluster_list, cluster);
+ list_append(assoc_cond->acct_list, "root");
notice_thread_init();
- ret_list = acct_storage_g_modify_clusters(
- db_conn, my_uid, cluster_cond, cluster);
- notice_thread_fini();
+ ret_list = acct_storage_g_modify_associations(
+ db_conn, my_uid, assoc_cond, assoc);
+
if(ret_list && list_count(ret_list)) {
char *object = NULL;
ListIterator itr = list_iterator_create(ret_list);
- printf(" Modifying clusters...\n");
+ printf(" Modified cluster defaults for associations...\n");
while((object = list_next(itr))) {
printf(" %s\n", object);
}
list_iterator_destroy(itr);
- if(commit_check("Would you like to commit changes?")) {
- acct_storage_g_commit(db_conn, 1);
- } else {
- printf(" Changes Discarded\n");
- acct_storage_g_commit(db_conn, 0);
- }
+ set = 1;
} else if(ret_list) {
printf(" Nothing modified\n");
} else {
@@ -551,9 +575,18 @@ extern int sacctmgr_modify_cluster(int argc, char *argv[])
if(ret_list)
list_destroy(ret_list);
+ notice_thread_fini();
- destroy_acct_cluster_cond(cluster_cond);
- destroy_acct_cluster_rec(cluster);
+ if(set) {
+ if(commit_check("Would you like to commit changes?"))
+ acct_storage_g_commit(db_conn, 1);
+ else {
+ printf(" Changes Discarded\n");
+ acct_storage_g_commit(db_conn, 0);
+ }
+ }
+ destroy_acct_association_cond(assoc_cond);
+ destroy_acct_association_rec(assoc);
return rc;
}
@@ -568,7 +601,7 @@ extern int sacctmgr_delete_cluster(int argc, char *argv[])
cluster_cond->cluster_list = list_create(slurm_destroy_char);
- if(!_set_cond(&i, argc, argv, cluster_cond, NULL)) {
+ if(!_set_cond(&i, argc, argv, cluster_cond->cluster_list, NULL)) {
printf(" No conditions given to remove, not executing.\n");
destroy_acct_cluster_cond(cluster_cond);
return SLURM_ERROR;
diff --git a/src/sacctmgr/sacctmgr.c b/src/sacctmgr/sacctmgr.c
index 73f6b23dc31..536081bbec2 100644
--- a/src/sacctmgr/sacctmgr.c
+++ b/src/sacctmgr/sacctmgr.c
@@ -426,7 +426,7 @@ _process_command (int argc, char *argv[])
argv[0]);
}
quiet_flag = -1;
- } else if (strncasecmp (argv[0], "rollup", 1) == 0) {
+ } else if (strncasecmp (argv[0], "rollup", 2) == 0) {
if (argc > 1) {
exit_code = 1;
fprintf (stderr,
diff --git a/src/sacctmgr/user_functions.c b/src/sacctmgr/user_functions.c
index 358e53809aa..5891753a8be 100644
--- a/src/sacctmgr/user_functions.c
+++ b/src/sacctmgr/user_functions.c
@@ -151,13 +151,10 @@ static int _set_rec(int *start, int argc, char *argv[],
a_set = 1;
} else if (strncasecmp (argv[i], "MaxWall", 4) == 0) {
mins = time_str2mins(argv[i]+end);
- if (mins >= 0) {
+ if (mins != NO_VAL) {
association->max_wall_duration_per_job
= (uint32_t) mins;
a_set = 1;
- } else if (strcmp(argv[i]+end, "-1") == 0) {
- association->max_wall_duration_per_job = -1;
- a_set = 1;
} else {
printf(" Bad MaxWall time format: %s\n",
argv[i]);
@@ -319,12 +316,9 @@ extern int sacctmgr_add_user(int argc, char *argv[])
limit_set = 1;
} else if (strncasecmp (argv[i], "MaxWall", 4) == 0) {
mins = time_str2mins(argv[i]+end);
- if (mins >= 0) {
+ if (mins != NO_VAL) {
max_wall_duration_per_job = (uint32_t) mins;
limit_set = 1;
- } else if (strcmp(argv[i]+end, "-1") == 0) {
- max_wall_duration_per_job = -1;
- limit_set = 1;
} else {
printf(" Bad MaxWall time format: %s\n",
argv[i]);
@@ -606,16 +600,30 @@ no_default:
if(limit_set) {
printf(" Non Default Settings\n");
- if(fairshare != NO_VAL)
+ if(fairshare == INFINITE)
+ printf(" Fairshare = NONE\n");
+ else if(fairshare != NO_VAL)
printf(" Fairshare = %u\n", fairshare);
- if(max_cpu_secs_per_job != NO_VAL)
+
+ if(max_cpu_secs_per_job == INFINITE)
+ printf(" MaxCPUSecs = NONE\n");
+ else if(max_cpu_secs_per_job != NO_VAL)
printf(" MaxCPUSecs = %u\n",
max_cpu_secs_per_job);
- if(max_jobs != NO_VAL)
+
+ if(max_jobs == INFINITE)
+ printf(" MaxJobs = NONE\n");
+ else if(max_jobs != NO_VAL)
printf(" MaxJobs = %u\n", max_jobs);
- if(max_nodes_per_job != NO_VAL)
+
+ if(max_nodes_per_job == INFINITE)
+ printf(" MaxNodes = NONE\n");
+ else if(max_nodes_per_job != NO_VAL)
printf(" MaxNodes = %u\n", max_nodes_per_job);
- if(max_wall_duration_per_job != NO_VAL) {
+
+ if(max_wall_duration_per_job == INFINITE)
+ printf(" MaxWall = NONE\n");
+ else if(max_wall_duration_per_job != NO_VAL) {
char time_buf[32];
mins2time_str((time_t) max_wall_duration_per_job,
time_buf, sizeof(time_buf));
diff --git a/src/slurmctld/controller.c b/src/slurmctld/controller.c
index cb762c9d836..0e8f00bad41 100644
--- a/src/slurmctld/controller.c
+++ b/src/slurmctld/controller.c
@@ -377,6 +377,7 @@ int main(int argc, char *argv[])
* information to Gold or SlurmDBD, create
* a file called "/tmp/slurm_accounting_first" to
* capture node initialization information */
+
_accounting_mark_all_nodes_down("cold-start");
unlink("/tmp/slurm_accounting_first");
}
@@ -928,6 +929,12 @@ static int _accounting_mark_all_nodes_down(char *reason)
}
xfree(state_file);
+ if((rc = acct_storage_g_flush_jobs_on_cluster(acct_db_conn,
+ slurmctld_cluster_name,
+ event_time))
+ == SLURM_ERROR)
+ return rc;
+
node_ptr = node_record_table_ptr;
for (i = 0; i < node_record_count; i++, node_ptr++) {
if (node_ptr->name == '\0')
diff --git a/src/slurmctld/job_mgr.c b/src/slurmctld/job_mgr.c
index 85cbb7be008..4174b656787 100644
--- a/src/slurmctld/job_mgr.c
+++ b/src/slurmctld/job_mgr.c
@@ -5296,7 +5296,7 @@ static bool _validate_acct_policy(job_desc_msg_t *job_desc,
{
uint32_t time_limit;
- log_assoc_rec(assoc_ptr);
+ //log_assoc_rec(assoc_ptr);
if ((assoc_ptr->max_wall_duration_per_job != NO_VAL) &&
(assoc_ptr->max_wall_duration_per_job != INFINITE)) {
time_limit = assoc_ptr->max_wall_duration_per_job;
diff --git a/src/slurmctld/node_mgr.c b/src/slurmctld/node_mgr.c
index a99a99030a1..834ecc65ad6 100644
--- a/src/slurmctld/node_mgr.c
+++ b/src/slurmctld/node_mgr.c
@@ -1638,7 +1638,10 @@ extern int validate_nodes_via_front_end(
slurm_node_registration_status_msg_t *reg_msg)
{
int error_code = 0, i, jobs_on_node;
- bool updated_job = false, failure_logged = false;
+ bool updated_job = false;
+#ifdef HAVE_BG
+ bool failure_logged = false;
+#endif
struct job_record *job_ptr;
struct config_record *config_ptr;
struct node_record *node_ptr;
diff --git a/src/slurmdbd/proc_req.c b/src/slurmdbd/proc_req.c
index f2b6627b012..945732abb7e 100644
--- a/src/slurmdbd/proc_req.c
+++ b/src/slurmdbd/proc_req.c
@@ -66,6 +66,8 @@ static int _get_jobs(void *db_conn, Buf in_buffer, Buf *out_buffer);
static int _get_usage(uint16_t type, void *db_conn,
Buf in_buffer, Buf *out_buffer);
static int _get_users(void *db_conn, Buf in_buffer, Buf *out_buffer);
+static int _flush_jobs(void *db_conn,
+ Buf in_buffer, Buf *out_buffer, uint32_t *uid);
static void *_init_conn(Buf in_buffer, Buf *out_buffer, uint32_t *uid);
static int _fini_conn(void **db_conn, Buf in_buffer, Buf *out_buffer);
static int _job_complete(void *db_conn,
@@ -177,6 +179,9 @@ proc_req(void **db_conn, slurm_fd orig_fd,
case DBD_GET_USERS:
rc = _get_users(*db_conn, in_buffer, out_buffer);
break;
+ case DBD_FLUSH_JOBS:
+ rc = _flush_jobs(*db_conn, in_buffer, out_buffer, uid);
+ break;
case DBD_INIT:
if (first)
(*db_conn) = _init_conn(
@@ -786,6 +791,39 @@ static int _get_users(void *db_conn, Buf in_buffer, Buf *out_buffer)
return SLURM_SUCCESS;
}
+static int _flush_jobs(void *db_conn,
+ Buf in_buffer, Buf *out_buffer, uint32_t *uid)
+{
+ dbd_cluster_procs_msg_t *cluster_procs_msg = NULL;
+ int rc = SLURM_SUCCESS;
+ char *comment = NULL;
+
+ if (*uid != slurmdbd_conf->slurm_user_id) {
+ comment = "DBD_FLUSH_JOBS message from invalid uid";
+ error("DBD_FLUSH_JOBS message from invalid uid %u", *uid);
+ rc = ESLURM_ACCESS_DENIED;
+ goto end_it;
+ }
+ if (slurmdbd_unpack_cluster_procs_msg(&cluster_procs_msg, in_buffer) !=
+ SLURM_SUCCESS) {
+ comment = "Failed to unpack DBD_FLUSH_JOBS message";
+ error("%s", comment);
+ rc = SLURM_ERROR;
+ goto end_it;
+ }
+ debug2("DBD_FLUSH_JOBS: called for %s",
+ cluster_procs_msg->cluster_name);
+
+ rc = acct_storage_g_flush_jobs_on_cluster(
+ db_conn,
+ cluster_procs_msg->cluster_name,
+ cluster_procs_msg->event_time);
+end_it:
+ slurmdbd_free_cluster_procs_msg(cluster_procs_msg);
+ *out_buffer = make_dbd_rc_msg(rc, comment, DBD_FLUSH_JOBS);
+ return rc;
+}
+
static void *_init_conn(Buf in_buffer, Buf *out_buffer, uint32_t *uid)
{
dbd_init_msg_t *init_msg = NULL;
@@ -1085,7 +1123,7 @@ static int _modify_assocs(void *db_conn,
get_msg->cond, get_msg->rec);
slurmdbd_free_modify_msg(DBD_MODIFY_ASSOCS, get_msg);
- *out_buffer = init_buf(1024);
+ *out_buffer = init_buf(1024);
pack16((uint16_t) DBD_GOT_LIST, *out_buffer);
slurmdbd_pack_list_msg(DBD_GOT_LIST, &list_msg, *out_buffer);
if(list_msg.my_list)
@@ -1128,7 +1166,7 @@ static int _modify_clusters(void *db_conn,
get_msg->cond, get_msg->rec);
slurmdbd_free_modify_msg(DBD_MODIFY_CLUSTERS, get_msg);
- *out_buffer = init_buf(1024);
+ *out_buffer = init_buf(1024);
pack16((uint16_t) DBD_GOT_LIST, *out_buffer);
slurmdbd_pack_list_msg(DBD_GOT_LIST, &list_msg, *out_buffer);
if(list_msg.my_list)
--
GitLab