diff --git a/src/common/slurm_accounting_storage.c b/src/common/slurm_accounting_storage.c
index 59879e61592c5cc4fcd341632d95c9f391de6193..31d94d6277282d9f514827ac56e4b77fe4ce61c5 100644
--- a/src/common/slurm_accounting_storage.c
+++ b/src/common/slurm_accounting_storage.c
@@ -107,7 +107,8 @@ typedef struct slurm_acct_storage_ops {
void *acct_assoc,
time_t start,
time_t end);
- int (*roll_usage) (void *db_conn);
+ int (*roll_usage) (void *db_conn,
+ time_t sent_start);
int (*node_down) (void *db_conn,
char *cluster,
struct node_record *node_ptr,
@@ -1850,11 +1851,12 @@ extern int acct_storage_g_get_usage(void *db_conn,
(db_conn, acct_assoc, start, end);
}
-extern int acct_storage_g_roll_usage(void *db_conn)
+extern int acct_storage_g_roll_usage(void *db_conn,
+ time_t sent_start)
{
if (slurm_acct_storage_init(NULL) < 0)
return SLURM_ERROR;
- return (*(g_acct_storage_context->ops.roll_usage))(db_conn);
+ return (*(g_acct_storage_context->ops.roll_usage))(db_conn, sent_start);
}
extern int clusteracct_storage_g_node_down(void *db_conn,
diff --git a/src/common/slurm_accounting_storage.h b/src/common/slurm_accounting_storage.h
index 1aec1b9f61913c650aa3f0365846d17675ed07b7..710d0457a50cc10387de02ea407d8400a8cb55dc 100644
--- a/src/common/slurm_accounting_storage.h
+++ b/src/common/slurm_accounting_storage.h
@@ -458,9 +458,11 @@ extern int acct_storage_g_get_usage(
void *db_conn, void *assoc, time_t start, time_t end);
/*
* roll up data in the storage
+ * IN: sent_start (option time to do a re-roll or start from this point)
* RET: SLURM_SUCCESS on success SLURM_ERROR else
*/
-extern int acct_storage_g_roll_usage(void *db_conn);
+extern int acct_storage_g_roll_usage(void *db_conn,
+ time_t sent_start);
/*
* record shares used information for backup in case slurmctld restarts
* IN: account_list List of shares_used_object_t *
diff --git a/src/common/slurmdbd_defs.c b/src/common/slurmdbd_defs.c
index 0b71e38c2bc81f798a16d99fae4465eccde6f47f..7cec1b4967ea658bf67d026cc4f17743812363e9 100644
--- a/src/common/slurmdbd_defs.c
+++ b/src/common/slurmdbd_defs.c
@@ -1288,6 +1288,7 @@ void inline slurmdbd_free_job_complete_msg(dbd_job_comp_msg_t *msg)
void inline slurmdbd_free_job_start_msg(dbd_job_start_msg_t *msg)
{
if (msg) {
+ xfree(msg->account);
xfree(msg->block_id);
xfree(msg->name);
xfree(msg->nodes);
@@ -1752,6 +1753,7 @@ unpack_error:
void inline
slurmdbd_pack_job_start_msg(dbd_job_start_msg_t *msg, Buf buffer)
{
+ packstr(msg->account, buffer);
pack32(msg->alloc_cpus, buffer);
pack32(msg->assoc_id, buffer);
packstr(msg->block_id, buffer);
@@ -1776,6 +1778,7 @@ slurmdbd_unpack_job_start_msg(dbd_job_start_msg_t **msg, Buf buffer)
uint32_t uint32_tmp;
dbd_job_start_msg_t *msg_ptr = xmalloc(sizeof(dbd_job_start_msg_t));
*msg = msg_ptr;
+ safe_unpackstr_xmalloc(&msg_ptr->account, &uint32_tmp, buffer);
safe_unpack32(&msg_ptr->alloc_cpus, buffer);
safe_unpack32(&msg_ptr->assoc_id, buffer);
safe_unpackstr_xmalloc(&msg_ptr->block_id, &uint32_tmp, buffer);
@@ -2147,6 +2150,7 @@ slurmdbd_unpack_roll_usage_msg(dbd_roll_usage_msg_t **msg, Buf buffer)
{
dbd_roll_usage_msg_t *msg_ptr = xmalloc(sizeof(dbd_roll_usage_msg_t));
+ *msg = msg_ptr;
safe_unpack_time(&msg_ptr->start, buffer);
return SLURM_SUCCESS;
diff --git a/src/common/slurmdbd_defs.h b/src/common/slurmdbd_defs.h
index f11647fd2a26943bbb82b2b70e908e7853825f5f..9e3ff3f7aad3cba3ffed536bfa8e80a941cacff0 100644
--- a/src/common/slurmdbd_defs.h
+++ b/src/common/slurmdbd_defs.h
@@ -185,6 +185,8 @@ typedef struct dbd_job_comp_msg {
} dbd_job_comp_msg_t;
typedef struct dbd_job_start_msg {
+ char * account; /* Account name for those not running
+ * with associations */
uint32_t alloc_cpus; /* count of allocated processors */
uint32_t assoc_id; /* accounting association id */
char * block_id; /* Bluegene block id */
diff --git a/src/plugins/accounting_storage/filetxt/accounting_storage_filetxt.c b/src/plugins/accounting_storage/filetxt/accounting_storage_filetxt.c
index 458305e457d552f3f03e8db0e6b78cf4d13c09f1..1e95e1baea91e962b6b6e912bafa40f76fe1f0ae 100644
--- a/src/plugins/accounting_storage/filetxt/accounting_storage_filetxt.c
+++ b/src/plugins/accounting_storage/filetxt/accounting_storage_filetxt.c
@@ -365,7 +365,8 @@ extern int acct_storage_p_get_usage(void *db_conn,
return rc;
}
-extern int acct_storage_p_roll_usage(void *db_conn)
+extern int acct_storage_p_roll_usage(void *db_conn,
+ time_t sent_start)
{
int rc = SLURM_SUCCESS;
diff --git a/src/plugins/accounting_storage/gold/accounting_storage_gold.c b/src/plugins/accounting_storage/gold/accounting_storage_gold.c
index df6fe00da8411fecdb0ea6fc7663725878d50cd0..1697520cdedaa08710828d1737ec7664b98a9372 100644
--- a/src/plugins/accounting_storage/gold/accounting_storage_gold.c
+++ b/src/plugins/accounting_storage/gold/accounting_storage_gold.c
@@ -2511,7 +2511,8 @@ extern int acct_storage_p_get_usage(void *db_conn,
return rc;
}
-extern int acct_storage_p_roll_usage(void *db_conn)
+extern int acct_storage_p_roll_usage(void *db_conn,
+ time_t sent_start)
{
int rc = SLURM_ERROR;
/* FIX ME: This doesn't do anything now */
diff --git a/src/plugins/accounting_storage/mysql/accounting_storage_mysql.c b/src/plugins/accounting_storage/mysql/accounting_storage_mysql.c
index 52bed03a963c03945577c18c87a2dcae6c0851ca..0b309827bf0f590ebd2ff225d85dba9856dbedf3 100644
--- a/src/plugins/accounting_storage/mysql/accounting_storage_mysql.c
+++ b/src/plugins/accounting_storage/mysql/accounting_storage_mysql.c
@@ -901,6 +901,7 @@ static int _mysql_acct_check_tables(MYSQL *acct_mysql_db)
{ "down_cpu_secs", "bigint default 0" },
{ "idle_cpu_secs", "bigint default 0" },
{ "resv_cpu_secs", "bigint default 0" },
+ { "over_cpu_secs", "bigint default 0" },
{ NULL, NULL}
};
@@ -4440,7 +4441,8 @@ extern int acct_storage_p_get_usage(mysql_conn_t *mysql_conn,
#endif
}
-extern int acct_storage_p_roll_usage(mysql_conn_t *mysql_conn)
+extern int acct_storage_p_roll_usage(mysql_conn_t *mysql_conn,
+ time_t sent_start)
{
#ifdef HAVE_MYSQL
int rc = SLURM_SUCCESS;
@@ -4452,11 +4454,12 @@ extern int acct_storage_p_roll_usage(mysql_conn_t *mysql_conn)
MYSQL_ROW row;
char *query = NULL;
char *tmp = NULL;
- time_t last_hour = 0;
- time_t last_day = 0;
- time_t last_month = 0;
+ time_t last_hour = sent_start;
+ time_t last_day = sent_start;
+ time_t last_month = sent_start;
time_t start_time = 0;
time_t end_time = 0;
+ DEF_TIMERS;
char *update_req_inx[] = {
"hourly_rollup",
@@ -4484,46 +4487,60 @@ extern int acct_storage_p_roll_usage(mysql_conn_t *mysql_conn)
}
}
- i=0;
- xstrfmtcat(tmp, "%s", update_req_inx[i]);
- for(i=1; i<UPDATE_COUNT; i++) {
- xstrfmtcat(tmp, ", %s", update_req_inx[i]);
- }
-
- query = xstrdup_printf("select %s from %s", tmp, last_ran_table);
- xfree(tmp);
-
- 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);
- row = mysql_fetch_row(result);
- if(row) {
- /* the last times were one second before the next
- * period so increment here 1
- */
- last_hour = atoi(row[UPDATE_HOUR]);
- last_day = atoi(row[UPDATE_DAY]);
- last_month = atoi(row[UPDATE_MONTH]);
- } else {
- query = xstrdup_printf(
- "insert into %s "
- "(hourly_rollup, daily_rollup, monthly_rollup) "
- "values (0, 0, 0)",
- last_ran_table);
+ if(!sent_start) {
+ i=0;
+ xstrfmtcat(tmp, "%s", update_req_inx[i]);
+ for(i=1; i<UPDATE_COUNT; i++) {
+ xstrfmtcat(tmp, ", %s", update_req_inx[i]);
+ }
+ query = xstrdup_printf("select %s from %s",
+ tmp, last_ran_table);
+ xfree(tmp);
+
+ if(!(result = mysql_db_query_ret(
+ mysql_conn->acct_mysql_db, query, 0))) {
+ xfree(query);
+ return SLURM_ERROR;
+ }
- rc = mysql_db_query(mysql_conn->acct_mysql_db, query);
xfree(query);
- if(rc == SLURM_ERROR)
- return rc;
+ row = mysql_fetch_row(result);
+ if(row) {
+ last_hour = atoi(row[UPDATE_HOUR]);
+ last_day = atoi(row[UPDATE_DAY]);
+ last_month = atoi(row[UPDATE_MONTH]);
+ mysql_free_result(result);
+ } else {
+ query = xstrdup_printf(
+ "select @PS := period_start from %s limit 1;"
+ "insert into %s "
+ "(hourly_rollup, daily_rollup, monthly_rollup) "
+ "values (@PS, @PS, @PS);",
+ event_table, last_ran_table);
+
+ mysql_free_result(result);
+ if(!(result = mysql_db_query_ret(
+ mysql_conn->acct_mysql_db, query, 0))) {
+ xfree(query);
+ return SLURM_ERROR;
+ }
+ xfree(query);
+ row = mysql_fetch_row(result);
+ if(!row) {
+ debug("No clusters have been added "
+ "not doing rollup");
+ mysql_free_result(result);
+ return SLURM_SUCCESS;
+ }
+
+ last_hour = last_day = last_month = atoi(row[0]);
+ mysql_free_result(result);
+ }
}
- last_hour = 1211475599;
- last_day = 1211475599;
- last_month = 1211475599;
+/* last_hour = 1211475599; */
+/* last_day = 1211475599; */
+/* last_month = 1211475599; */
+
// last_hour = 1211403599;
// last_hour = 1206946800;
// last_day = 1207033199;
@@ -4558,10 +4575,12 @@ extern int acct_storage_p_roll_usage(mysql_conn_t *mysql_conn)
end_tm.tm_isdst = -1;
end_time = mktime(&end_tm);
if(end_time-start_time > 0) {
+ START_TIMER;
if((rc = mysql_hourly_rollup(mysql_conn, start_time, end_time))
!= SLURM_SUCCESS)
return rc;
- query = xstrdup_printf("update %s set hour_rollup=%d",
+ END_TIMER2("hourly_rollup");
+ query = xstrdup_printf("update %s set hourly_rollup=%d",
last_ran_table, end_time);
} else {
debug2("no need to run this hour %d < %d",
@@ -4583,9 +4602,11 @@ extern int acct_storage_p_roll_usage(mysql_conn_t *mysql_conn)
end_tm.tm_isdst = -1;
end_time = mktime(&end_tm);
if(end_time-start_time > 0) {
+ START_TIMER;
if((rc = mysql_daily_rollup(mysql_conn, start_time, end_time))
!= SLURM_SUCCESS)
return rc;
+ END_TIMER2("daily_rollup");
if(query)
xstrfmtcat(query, ", daily_rollup=%d", end_time);
else
@@ -4614,9 +4635,12 @@ extern int acct_storage_p_roll_usage(mysql_conn_t *mysql_conn)
end_tm.tm_isdst = -1;
end_time = mktime(&end_tm);
if(end_time-start_time > 0) {
+ START_TIMER;
if((rc = mysql_monthly_rollup(
mysql_conn, start_time, end_time)) != SLURM_SUCCESS)
return rc;
+ END_TIMER2("monthly_rollup");
+
if(query)
xstrfmtcat(query, ", montly_rollup=%d", end_time);
else
@@ -4629,8 +4653,8 @@ extern int acct_storage_p_roll_usage(mysql_conn_t *mysql_conn)
}
if(query) {
- info("%s", query);
- //rc = mysql_db_query(mysql_conn->acct_mysql_db, query);
+ debug3("%s", query);
+ 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 58ee6a5afa9d3aac6380ddf4dcdc3db6c1c0f7da..4fe1d41c120c6a67709742dde47fc5e385d65144 100644
--- a/src/plugins/accounting_storage/mysql/mysql_rollup.c
+++ b/src/plugins/accounting_storage/mysql/mysql_rollup.c
@@ -54,6 +54,7 @@ typedef struct {
int cpu_count;
int d_cpu;
int i_cpu;
+ int o_cpu;
int r_cpu;
time_t start;
time_t end;
@@ -166,13 +167,16 @@ extern int mysql_hourly_rollup(mysql_conn_t *mysql_conn,
/* info("begin start %s", ctime(&curr_start)); */
/* info("begin end %s", ctime(&curr_end)); */
- a_itr = list_iterator_create(cluster_usage_list);
+ a_itr = list_iterator_create(assoc_usage_list);
c_itr = list_iterator_create(cluster_usage_list);
while(curr_start < end) {
int last_id = 0;
int seconds = 0;
local_cluster_usage_t *c_usage = NULL;
local_assoc_usage_t *a_usage = NULL;
+ debug3("curr hour is now %d-%d", curr_start, curr_end);
+/* info("start %s", ctime(&curr_start)); */
+/* info("end %s", ctime(&curr_end)); */
// first get the events during this time
query = xstrdup_printf("select %s from %s where "
@@ -259,14 +263,15 @@ extern int mysql_hourly_rollup(mysql_conn_t *mysql_conn,
seconds = (local_end - local_start);
- info("node %s adds (%d)(%d-%d) * %d = %d "
- "to %d",
- row[EVENT_REQ_NAME],
- seconds,
- local_end, local_start,
- row_cpu,
- seconds * row_cpu,
- row_cpu);
+/* info("node %s adds " */
+/* "(%d)(%d-%d) * %d = %d " */
+/* "to %d", */
+/* row[EVENT_REQ_NAME], */
+/* seconds, */
+/* local_end, local_start, */
+/* row_cpu, */
+/* seconds * row_cpu, */
+/* row_cpu); */
c_usage->d_cpu += seconds * row_cpu;
break;
@@ -368,7 +373,7 @@ extern int mysql_hourly_rollup(mysql_conn_t *mysql_conn,
}
if(seconds < 1) {
- debug("This job (%u) was suspended "
+ debug4("This job (%u) was suspended "
"the entire hour", job_id);
continue;
}
@@ -387,43 +392,47 @@ extern int mysql_hourly_rollup(mysql_conn_t *mysql_conn,
if(!row_start || seconds < 1)
goto calc_resv;
- info("%d assoc %d adds (%d)(%d-%d) * %d = %d "
- "to %d",
- job_id,
- assoc_id,
- seconds,
- row_end, row_start,
- row_acpu,
- seconds * row_acpu,
- row_acpu);
+/* info("%d assoc %d adds " */
+/* "(%d)(%d-%d) * %d = %d " */
+/* "to %d", */
+/* job_id, */
+/* a_usage->assoc_id, */
+/* seconds, */
+/* row_end, row_start, */
+/* row_acpu, */
+/* seconds * row_acpu, */
+/* row_acpu); */
c_usage->a_cpu += seconds * row_acpu;
calc_resv:
/* now reserved time */
- if(row_start < c_usage->start)
+ if(row_start &&
+ row_start < c_usage->start)
continue;
- row_start = row_eligible;
+
row_end = row_start;
+ row_start = row_eligible;
if(c_usage->start > row_start)
row_start = c_usage->start;
if(c_usage->end < row_end)
row_end = c_usage->end;
-
+
if((row_end - row_start) < 1)
continue;
seconds = (row_end - row_start);
- info("%d assoc %d reserved (%d)(%d-%d) * %d = %d "
- "to %d",
- job_id,
- assoc_id,
- seconds,
- row_end, row_start,
- row_rcpu,
- seconds * row_rcpu,
- row_rcpu);
+/* info("%d assoc %d reserved " */
+/* "(%d)(%d-%d) * %d = %d " */
+/* "to %d", */
+/* job_id, */
+/* assoc_id, */
+/* seconds, */
+/* row_end, row_start, */
+/* row_rcpu, */
+/* seconds * row_rcpu, */
+/* row_rcpu); */
c_usage->r_cpu += seconds * row_rcpu;
break;
@@ -433,55 +442,76 @@ extern int mysql_hourly_rollup(mysql_conn_t *mysql_conn,
mysql_free_result(result);
list_iterator_reset(c_itr);
-
while((c_usage = list_next(c_itr))) {
c_usage->i_cpu = c_usage->total_time - c_usage->a_cpu -
c_usage->d_cpu - c_usage->r_cpu;
/* sanity check just to make sure we have a
* legitimate time after we calulated
- * idle/reserved time
+ * idle/reserved time put extra in the over
+ * commit field
*/
if(c_usage->i_cpu < 0) {
c_usage->r_cpu += c_usage->i_cpu;
+ c_usage->o_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 = %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,
- c_usage->total_time,
- ctime(&c_usage->start));
- info("to %s", ctime(&c_usage->end));
+/* info("cluster %s(%d) down %d alloc %d " */
+/* "resv %d idle %d over %d " */
+/* "total= %d = %d 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->o_cpu, */
+/* c_usage->d_cpu + c_usage->a_cpu + */
+/* c_usage->r_cpu + c_usage->i_cpu, */
+/* c_usage->total_time, */
+/* ctime(&c_usage->start)); */
+/* info("to %s", ctime(&c_usage->end)); */
if(query) {
xstrfmtcat(query,
- ", (%d, %d, '%s', %d, %u, "
- "%u, %u, %u, %u)",
+ ", (%d, %d, '%s', %d, %d, "
+ "%d, %d, %d, %d, %d) "
+ "on duplicate key update "
+ "mod_time=%d, cpu_count=%d, "
+ "alloc_cpu_secs=%d, "
+ "down_cpu_secs=%d, "
+ "idle_cpu_secs=%d, "
+ "over_cpu_secs=%d, resv_cpu_secs=%d",
now, now,
c_usage->name, c_usage->start,
c_usage->cpu_count, c_usage->a_cpu,
c_usage->d_cpu, c_usage->i_cpu,
- c_usage->r_cpu);
+ c_usage->o_cpu, c_usage->r_cpu,
+ now,
+ c_usage->cpu_count, c_usage->a_cpu,
+ c_usage->d_cpu, c_usage->i_cpu,
+ c_usage->o_cpu, c_usage->r_cpu);
} else {
xstrfmtcat(query,
"insert into %s (creation_time, "
"mod_time, cluster, period_start, "
"cpu_count, alloc_cpu_secs, "
"down_cpu_secs, idle_cpu_secs, "
- "resv_cpu_secs) values "
- "(%d, %d, '%s', %d, %u, "
- "%u, %u, %u, %u)",
+ "over_cpu_secs, resv_cpu_secs) "
+ "values (%d, %d, '%s', %d, %d, "
+ "%d, %d, %d, %d, %d) "
+ "on duplicate key update "
+ "mod_time=%d, cpu_count=%d, "
+ "alloc_cpu_secs=%d, "
+ "down_cpu_secs=%d, "
+ "idle_cpu_secs=%d, "
+ "over_cpu_secs=%d, resv_cpu_secs=%d",
cluster_hour_table, now, now,
c_usage->name, c_usage->start,
c_usage->cpu_count, c_usage->a_cpu,
c_usage->d_cpu, c_usage->i_cpu,
- c_usage->r_cpu);
+ c_usage->o_cpu, c_usage->r_cpu,
+ now,
+ c_usage->cpu_count, c_usage->a_cpu,
+ c_usage->d_cpu, c_usage->i_cpu,
+ c_usage->o_cpu, c_usage->r_cpu);
}
}
if(query) {
@@ -495,28 +525,36 @@ extern int mysql_hourly_rollup(mysql_conn_t *mysql_conn,
list_iterator_reset(a_itr);
while((a_usage = list_next(a_itr))) {
- info("association (%u) alloc %u",
- a_usage->assoc_id,
- a_usage->a_cpu);
+/* info("association (%d) %d alloc %d", */
+/* a_usage->assoc_id, last_id, */
+/* a_usage->a_cpu); */
if(query) {
xstrfmtcat(query,
- ", (%d, %d, '%u', %d, %u, "
- "%u, %u, %u, %u)",
+ ", (%d, %d, %d, %d, %d, "
+ "%d, %d, %d, %d) "
+ "on duplicate key update "
+ "mod_time=%d, alloc_cpu_secs=%d",
now, now,
a_usage->assoc_id, curr_start,
- a_usage->a_cpu);
+ a_usage->a_cpu,
+ now, a_usage->a_cpu);
} else {
xstrfmtcat(query,
"insert into %s (creation_time, "
"mod_time, id, period_start, "
"alloc_cpu_secs) values "
- "(%d, %d, %u, %d, %u)",
+ "(%d, %d, %d, %d, %d) "
+ "on duplicate key update "
+ "mod_time=%d, alloc_cpu_secs=%d",
assoc_hour_table, now, now,
a_usage->assoc_id, curr_start,
- a_usage->a_cpu);
+ a_usage->a_cpu,
+ now, a_usage->a_cpu);
}
}
+
if(query) {
+ debug3("%d query\n%s", mysql_conn->conn, query);
rc = mysql_db_query(mysql_conn->acct_mysql_db, query);
xfree(query);
if(rc != SLURM_SUCCESS) {
@@ -528,7 +566,6 @@ extern int mysql_hourly_rollup(mysql_conn_t *mysql_conn,
list_flush(cluster_usage_list);
curr_start = curr_end;
curr_end = curr_start + add_sec;
- debug3("curr hour is now %d-%d", curr_start, curr_end);
}
end_it:
xfree(suspend_str);
@@ -567,32 +604,43 @@ extern int mysql_daily_rollup(mysql_conn_t *mysql_conn,
start_tm.tm_isdst = -1;
curr_end = mktime(&start_tm);
- info("begin start %s", ctime(&curr_start));
- info("begin end %s", ctime(&curr_end));
while(curr_start < end) {
+ debug3("curr day is now %d-%d", curr_start, curr_end);
+/* info("start %s", ctime(&curr_start)); */
+/* info("end %s", ctime(&curr_end)); */
query = xstrdup_printf(
"insert into %s (creation_time, mod_time, id, "
"period_start, alloc_cpu_secs) select %d, %d, id, "
- "%d, SUM(alloc_cpu_secs) from %s where "
+ "%d, @ASUM:=SUM(alloc_cpu_secs) from %s where "
"(period_start < %d && period_start >= %d) "
- "group by id;",
+ "group by id on duplicate key update mod_time=%d, "
+ "alloc_cpu_secs=@ASUM;",
assoc_day_table, now, now, curr_start,
assoc_hour_table,
- curr_end, curr_start);
+ curr_end, curr_start, now);
xstrfmtcat(query,
"insert into %s (creation_time, "
"mod_time, cluster, period_start, cpu_count, "
"alloc_cpu_secs, down_cpu_secs, idle_cpu_secs, "
- "resv_cpu_secs) select %d, %d, cluster, "
- "%d, MAX(cpu_count), SUM(alloc_cpu_secs), "
- "SUM(down_cpu_secs), SUM(idle_cpu_secs), "
- "SUM(resv_cpu_secs) from %s where "
+ "over_cpu_secs, resv_cpu_secs) "
+ "select %d, %d, cluster, "
+ "%d, @CPU:=MAX(cpu_count), "
+ "@ASUM:=SUM(alloc_cpu_secs), "
+ "@DSUM:=SUM(down_cpu_secs), "
+ "@ISUM:=SUM(idle_cpu_secs), "
+ "@OSUM:=SUM(over_cpu_secs), "
+ "@RSUM:=SUM(resv_cpu_secs) from %s where "
"(period_start < %d && period_start >= %d) "
- "group by cluster;",
+ "group by cluster on duplicate key update "
+ "mod_time=%d, cpu_count=@CPU, "
+ "alloc_cpu_secs=@ASUM, down_cpu_secs=@DSUM, "
+ "idle_cpu_secs=@ISUM, over_cpu_secs=@OSUM, "
+ "resv_cpu_secs=@RSUM;",
cluster_day_table, now, now, curr_start,
cluster_hour_table,
- curr_end, curr_start);
- //rc = mysql_db_query(mysql_conn->acct_mysql_db, query);
+ curr_end, curr_start, now);
+ debug3("%d query\n%s", mysql_conn->conn, query);
+ rc = mysql_db_query(mysql_conn->acct_mysql_db, query);
xfree(query);
if(rc != SLURM_SUCCESS) {
error("Couldn't add day rollup");
@@ -611,10 +659,20 @@ extern int mysql_daily_rollup(mysql_conn_t *mysql_conn,
start_tm.tm_mday++;
start_tm.tm_isdst = -1;
curr_end = mktime(&start_tm);
- debug3("curr day is now %d-%d", curr_start, curr_end);
- info("stop start %s", ctime(&curr_start));
- info("stop end %s", ctime(&curr_end));
}
+ /* remove all data from suspend table that was older than
+ * start.
+ */
+ query = xstrdup_printf("delete from %s where end < %d && end != 0",
+ suspend_table, start);
+ rc = mysql_db_query(mysql_conn->acct_mysql_db, query);
+ xfree(query);
+ if(rc != SLURM_SUCCESS) {
+ error("Couldn't remove old suspend data");
+ return SLURM_ERROR;
+ }
+
+
/* info("stop start %s", ctime(&curr_start)); */
/* info("stop end %s", ctime(&curr_end)); */
@@ -623,12 +681,13 @@ extern int mysql_daily_rollup(mysql_conn_t *mysql_conn,
extern int mysql_monthly_rollup(mysql_conn_t *mysql_conn,
time_t start, time_t end)
{
+ int rc = SLURM_SUCCESS;
struct tm start_tm;
time_t curr_start = start;
time_t curr_end;
+ time_t now = time(NULL);
+ char *query = NULL;
-/* info("begin month start %s", ctime(&start)); */
-/* info("begin month end %s", ctime(&end)); */
if(!localtime_r(&curr_start, &start_tm)) {
error("Couldn't get localtime from month start %d", curr_start);
return SLURM_ERROR;
@@ -640,11 +699,49 @@ extern int mysql_monthly_rollup(mysql_conn_t *mysql_conn,
start_tm.tm_mon++;
start_tm.tm_isdst = -1;
curr_end = mktime(&start_tm);
-/* info("begin start %s", ctime(&curr_start)); */
-/* info("begin end %s", ctime(&curr_end)); */
- while(curr_start < end) {
- /* FIX ME: DO CALCULATIONS HERE */
+ while(curr_start < end) {
+ debug3("curr month is now %d-%d", curr_start, curr_end);
+/* info("start %s", ctime(&curr_start)); */
+/* info("end %s", ctime(&curr_end)); */
+ query = xstrdup_printf(
+ "insert into %s (creation_time, mod_time, id, "
+ "period_start, alloc_cpu_secs) select %d, %d, id, "
+ "%d, @ASUM:=SUM(alloc_cpu_secs) from %s where "
+ "(period_start < %d && period_start >= %d) "
+ "group by id on duplicate key update mod_time=%d, "
+ "alloc_cpu_secs=@ASUM;",
+ assoc_month_table, now, now, curr_start,
+ assoc_day_table,
+ curr_end, curr_start, now);
+ xstrfmtcat(query,
+ "insert into %s (creation_time, "
+ "mod_time, cluster, period_start, cpu_count, "
+ "alloc_cpu_secs, down_cpu_secs, idle_cpu_secs, "
+ "over_cpu_secs, resv_cpu_secs) "
+ "select %d, %d, cluster, "
+ "%d, @CPU:=MAX(cpu_count), "
+ "@ASUM:=SUM(alloc_cpu_secs), "
+ "@DSUM:=SUM(down_cpu_secs), "
+ "@ISUM:=SUM(idle_cpu_secs), "
+ "@OSUM:=SUM(over_cpu_secs), "
+ "@RSUM:=SUM(resv_cpu_secs) from %s where "
+ "(period_start < %d && period_start >= %d) "
+ "group by cluster on duplicate key update "
+ "mod_time=%d, cpu_count=@CPU, "
+ "alloc_cpu_secs=@ASUM, down_cpu_secs=@DSUM, "
+ "idle_cpu_secs=@ISUM, over_cpu_secs=@OSUM, "
+ "resv_cpu_secs=@RSUM;",
+ cluster_month_table, now, now, curr_start,
+ cluster_day_table,
+ curr_end, curr_start, now);
+ debug3("%d query\n%s", mysql_conn->conn, query);
+ rc = mysql_db_query(mysql_conn->acct_mysql_db, query);
+ xfree(query);
+ if(rc != SLURM_SUCCESS) {
+ error("Couldn't add day rollup");
+ return SLURM_ERROR;
+ }
curr_start = curr_end;
if(!localtime_r(&curr_start, &start_tm)) {
@@ -658,10 +755,7 @@ extern int mysql_monthly_rollup(mysql_conn_t *mysql_conn,
start_tm.tm_mon++;
start_tm.tm_isdst = -1;
curr_end = mktime(&start_tm);
- debug3("curr month is now %d-%d", curr_start, curr_end);
}
-/* info("stop start %s", ctime(&curr_start)); */
-/* info("stop end %s", ctime(&curr_end)); */
return SLURM_SUCCESS;
}
diff --git a/src/plugins/accounting_storage/none/accounting_storage_none.c b/src/plugins/accounting_storage/none/accounting_storage_none.c
index 24de0dd251dc37bb7ca0a09883afa4d65e30caac..0701440e26b77f8f7577cd08981bfe852ce8e520 100644
--- a/src/plugins/accounting_storage/none/accounting_storage_none.c
+++ b/src/plugins/accounting_storage/none/accounting_storage_none.c
@@ -223,7 +223,8 @@ extern int acct_storage_p_get_usage(void *db_conn,
return rc;
}
-extern int acct_storage_p_roll_usage(void *db_conn)
+extern int acct_storage_p_roll_usage(void *db_conn,
+ time_t sent_start)
{
int rc = 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 0c7c390ab1652f992222bd96a9588afca8e6a6ed..9fdbc22682082bf2a6c9a4f7d6e537f8e010f24f 100644
--- a/src/plugins/accounting_storage/pgsql/accounting_storage_pgsql.c
+++ b/src/plugins/accounting_storage/pgsql/accounting_storage_pgsql.c
@@ -827,7 +827,8 @@ extern int acct_storage_p_get_usage(PGconn *acct_pgsql_db,
return rc;
}
-extern int acct_storage_p_roll_usage(PGconn *acct_pgsql_db)
+extern int acct_storage_p_roll_usage(PGconn *acct_pgsql_db,
+ time_t sent_start)
{
int rc = 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 c75a2b8077784ee53a2a7cef2958934546c372b2..3dd41ad223dd35fdee5027640a23c45642f576b5 100644
--- a/src/plugins/accounting_storage/slurmdbd/accounting_storage_slurmdbd.c
+++ b/src/plugins/accounting_storage/slurmdbd/accounting_storage_slurmdbd.c
@@ -710,13 +710,14 @@ extern int acct_storage_p_get_usage(void *db_conn,
return rc;
}
-extern int acct_storage_p_roll_usage(void *db_conn)
+extern int acct_storage_p_roll_usage(void *db_conn,
+ time_t sent_start)
{
slurmdbd_msg_t req;
dbd_roll_usage_msg_t get_msg;
int rc, resp_code;
- get_msg.start = time(NULL);
+ get_msg.start = sent_start;
req.msg_type = DBD_ROLL_USAGE;
@@ -879,6 +880,7 @@ extern int jobacct_storage_p_job_start(void *db_conn,
}
req.alloc_cpus = job_ptr->total_procs;
+ req.account = job_ptr->account;
req.assoc_id = job_ptr->assoc_id;
#ifdef HAVE_BG
select_g_get_jobinfo(job_ptr->select_jobinfo,
@@ -886,8 +888,7 @@ extern int jobacct_storage_p_job_start(void *db_conn,
&block_id);
#endif
req.block_id = block_id;
- xfree(block_id);
- req.db_index = job_ptr->db_index;
+ req.db_index = job_ptr->db_index;
if (job_ptr->details)
req.eligible_time = job_ptr->details->begin_time;
req.gid = job_ptr->group_id;
@@ -910,8 +911,11 @@ extern int jobacct_storage_p_job_start(void *db_conn,
* again just send the message
*/
if(req.db_index) {
- if (slurm_send_slurmdbd_msg(&msg) < 0)
+ if (slurm_send_slurmdbd_msg(&msg) < 0) {
+ xfree(block_id);
return SLURM_ERROR;
+ }
+ xfree(block_id);
return SLURM_SUCCESS;
}
@@ -920,8 +924,10 @@ extern int jobacct_storage_p_job_start(void *db_conn,
*/
rc = slurm_send_recv_slurmdbd_msg(&msg, &msg_rc);
if (rc != SLURM_SUCCESS) {
- if (slurm_send_slurmdbd_msg(&msg) < 0)
+ if (slurm_send_slurmdbd_msg(&msg) < 0) {
+ xfree(block_id);
return SLURM_ERROR;
+ }
} else if (msg_rc.msg_type != DBD_JOB_START_RC) {
error("slurmdbd: response type not DBD_GOT_JOBS: %u",
msg_rc.msg_type);
@@ -930,6 +936,7 @@ extern int jobacct_storage_p_job_start(void *db_conn,
job_ptr->db_index = resp->db_index;
slurmdbd_free_job_start_rc_msg(resp);
}
+ xfree(block_id);
return rc;
}
diff --git a/src/sacctmgr/sacctmgr.c b/src/sacctmgr/sacctmgr.c
index 536081bbec291c6b18343dd473acea983211cc0e..e451f74d4d1d8f2d5422bf38134112a92056ff8c 100644
--- a/src/sacctmgr/sacctmgr.c
+++ b/src/sacctmgr/sacctmgr.c
@@ -427,15 +427,17 @@ _process_command (int argc, char *argv[])
}
quiet_flag = -1;
} else if (strncasecmp (argv[0], "rollup", 2) == 0) {
- if (argc > 1) {
+ time_t my_time = 0;
+ if (argc > 2) {
exit_code = 1;
fprintf (stderr,
"too many arguments for %s keyword\n",
argv[0]);
}
- printf("ROLLUP DOESN'T WORK YET, "
- "this is for test purposes only.\n");
- if(acct_storage_g_roll_usage(db_conn) == SLURM_SUCCESS) {
+ if(argc > 1)
+ my_time = parse_time(argv[1]);
+ if(acct_storage_g_roll_usage(db_conn, my_time)
+ == SLURM_SUCCESS) {
if(commit_check("Would you like to commit rollup?")) {
acct_storage_g_commit(db_conn, 1);
} else {
diff --git a/src/slurmctld/job_mgr.c b/src/slurmctld/job_mgr.c
index 4174b65678729ca20a7d87f3a21f7637457c92d6..6e3b19b11caf49225dd8d0de7fd1b4fd0dad7168 100644
--- a/src/slurmctld/job_mgr.c
+++ b/src/slurmctld/job_mgr.c
@@ -1011,8 +1011,8 @@ extern int kill_job_by_part_name(char *part_name)
difftime(now, job_ptr->suspend_time);
} else
job_ptr->end_time = now;
- job_completion_logger(job_ptr);
deallocate_nodes(job_ptr, false, suspended);
+ job_completion_logger(job_ptr);
} else if (job_ptr->job_state == JOB_PENDING) {
job_count++;
info("Killing job_id %u on defunct partition %s",
@@ -4769,6 +4769,7 @@ extern void job_completion_logger(struct job_record *job_ptr)
}
g_slurm_jobcomp_write(job_ptr);
+ jobacct_storage_g_job_complete(acct_db_conn, job_ptr);
}
/*
diff --git a/src/slurmctld/node_scheduler.c b/src/slurmctld/node_scheduler.c
index d6ea15e892d16efd6a398af51ff3744d0c19a887..2798ba45ecfed288d005fe274f874c05c87dfa97 100644
--- a/src/slurmctld/node_scheduler.c
+++ b/src/slurmctld/node_scheduler.c
@@ -210,10 +210,6 @@ extern void deallocate_nodes(struct job_record *job_ptr, bool timeout,
return;
}
- /* log this in the accounting plugin since it was allocated
- * something */
- jobacct_storage_g_job_complete(acct_db_conn, job_ptr);
-
agent_args->msg_args = kill_job;
agent_queue_request(agent_args);
return;
diff --git a/src/slurmdbd/proc_req.c b/src/slurmdbd/proc_req.c
index 945732abb7e5a956defeef24afc23f726917f09a..a3065e982a1ee0c46fbb3eee92c2cc7c07de28d8 100644
--- a/src/slurmdbd/proc_req.c
+++ b/src/slurmdbd/proc_req.c
@@ -962,6 +962,7 @@ static int _job_start(void *db_conn,
memset(&job_start_rc_msg, 0, sizeof(dbd_job_start_rc_msg_t));
job.total_procs = job_start_msg->alloc_cpus;
+ job.account = job_start_msg->account;
job.assoc_id = job_start_msg->assoc_id;
job.comment = job_start_msg->block_id;
job.db_index = job_start_msg->db_index;
@@ -1043,7 +1044,7 @@ static int _job_suspend(void *db_conn,
rc = SLURM_SUCCESS;
end_it:
slurmdbd_free_job_suspend_msg(job_suspend_msg);
- *out_buffer = make_dbd_rc_msg(rc, comment, DBD_JOB_START);
+ *out_buffer = make_dbd_rc_msg(rc, comment, DBD_JOB_SUSPEND);
return SLURM_SUCCESS;
}
@@ -1634,8 +1635,7 @@ static int _roll_usage(void *db_conn,
goto end_it;
}
- rc = acct_storage_g_roll_usage(db_conn);
- slurmdbd_free_roll_usage_msg(get_msg);
+ rc = acct_storage_g_roll_usage(db_conn, get_msg->start);
end_it:
slurmdbd_free_roll_usage_msg(get_msg);