diff --git a/NEWS b/NEWS
index 85d5c5fbfcf6aff4c7ed586c6b86c664e8e036f9..5268a3926d6c15f95543f8d9e07927e72c617b92 100644
--- a/NEWS
+++ b/NEWS
@@ -96,6 +96,8 @@ documents those changes that are of interest to users and admins.
termination time).
- fix for step allocation to be able to specify only a few nodes in a
step and ask for more that specified.
+ - patch from Hongjia Cao for forwarding logic
+ - BLUEGENE - able to allocate specific nodes without locking up.
* Changes in SLURM 1.1.13
=========================
diff --git a/src/common/forward.c b/src/common/forward.c
index b9d006018c409902543cce811f6270f54c86888d..7a41738a0c4dc19fde2451aa9bac6da00a08f629 100644
--- a/src/common/forward.c
+++ b/src/common/forward.c
@@ -182,7 +182,7 @@ void *_forward_thread(void *arg)
}
break;
}
-
+
slurm_mutex_lock(fwd_msg->forward_mutex);
if(ret_list) {
while((ret_data_info = list_pop(ret_list)) != NULL) {
diff --git a/src/common/plugstack.c b/src/common/plugstack.c
index 536df262ac52d43c58e8be4f42f0aca23f1ed827..8f5c87f6b99bcdd07f4c3e54a4a62df1cbb03d49 100644
--- a/src/common/plugstack.c
+++ b/src/common/plugstack.c
@@ -349,11 +349,12 @@ _spank_stack_process_line(const char *file, int line, char *buf,
}
if (!(p = _spank_plugin_create(path, ac, argv, required))) {
- error ("spank: %s:%d: Failed to load %s plugin from %s. %s",
- file, line,
- required ? "required" : "optional",
- path,
- required ? "Aborting." : "Ignoring.");
+ if (required)
+ error ("spank: %s:%d: Failed to load plugin %s. Aborting.",
+ file, line, path);
+ else
+ verbose ("spank: %s:%d: Failed to load optional plugin %s. Ignored.",
+ file, line, path);
return (required ? -1 : 0);
}
@@ -554,12 +555,15 @@ int spank_init(slurmd_job_t * job)
return (-1);
}
+ if (_do_call_stack(SPANK_INIT, job, -1) < 0)
+ return (-1);
+
if (job && spank_get_remote_options(job->options) < 0) {
error("spank: Unable to get remote options");
return (-1);
}
- return (_do_call_stack(SPANK_INIT, job, -1));
+ return (0);
}
int spank_user(slurmd_job_t * job)
diff --git a/src/plugins/select/bluegene/plugin/bg_job_place.c b/src/plugins/select/bluegene/plugin/bg_job_place.c
index 3982bbb57a4afa21c4784b1323a3303ed9aa411d..a5ee2f79e129e62524412906f08b9b7bc5d718dd 100644
--- a/src/plugins/select/bluegene/plugin/bg_job_place.c
+++ b/src/plugins/select/bluegene/plugin/bg_job_place.c
@@ -17,7 +17,7 @@
* any later version.
*
* In addition, as a special exception, the copyright holders give permission
- * to link the code of portions of this program with the OpenSSL library under
+ * to link the code of portions of this program with the OpenSSL library under
* certain conditions as described in each individual source file, and
* distribute linked combinations including the two. You must obey the GNU
* General Public License in all respects for all of the code used other than
@@ -147,6 +147,9 @@ static int _find_best_block_match(struct job_record* job_ptr,
select_g_get_jobinfo(job_ptr->select_jobinfo,
SELECT_DATA_MAX_PROCS, &max_procs);
+ if(start[X] != (uint16_t)NO_VAL)
+ start_req = 1;
+
if(req_geometry[0] != 0 && req_geometry[0] != (uint16_t)NO_VAL) {
target_size = 1;
for (i=0; i<BA_SYSTEM_DIMENSIONS; i++)
@@ -165,7 +168,7 @@ static int _find_best_block_match(struct job_record* job_ptr,
}
if (target_size == 0) { /* no geometry specified */
if(job_ptr->details->req_nodes
- && start[0] == (uint16_t)NO_VAL) {
+ && !start_req) {
bg_record_t *tmp_record = NULL;
char *tmp_nodes= job_ptr->details->req_nodes;
int len = strlen(tmp_nodes);
@@ -450,6 +453,27 @@ try_again:
slurm_mutex_lock(&request_list_mutex);
itr = list_iterator_create(bg_request_list);
while ((try_request = list_next(itr))) {
+ if(start_req) {
+ if ((try_request->start[X] != start[X])
+ || (try_request->start[Y] != start[Y])
+ || (try_request->start[Z] != start[Z])) {
+ debug4("got %d%d%d looking for %d%d%d",
+ try_request->start[X],
+ try_request->start[Y],
+ try_request->start[Z],
+ start[X],
+ start[Y],
+ start[Z]);
+ continue;
+ }
+ debug3("found %d%d%d looking for %d%d%d",
+ try_request->start[X],
+ try_request->start[Y],
+ try_request->start[Z],
+ start[X],
+ start[Y],
+ start[Z]);
+ }
if(try_request->procs >= req_procs) {
debug("already tried to create but "
"can't right now.");
@@ -465,7 +489,7 @@ try_again:
slurm_mutex_unlock(&request_list_mutex);
}
- if(!found && test_only) {
+ if(test_only) {
for(i=0; i<BA_SYSTEM_DIMENSIONS; i++)
request.start[i] = start[i];
@@ -504,6 +528,9 @@ try_again:
try_request->procs = req_procs;
try_request->save_name = NULL;
try_request->elongate_geos = NULL;
+ try_request->start_req = request.start_req;
+ for(i=0; i<BA_SYSTEM_DIMENSIONS; i++)
+ try_request->start[i] = start[i];
slurm_mutex_lock(&request_list_mutex);
list_push(bg_request_list, try_request);
slurm_mutex_unlock(&request_list_mutex);
diff --git a/src/plugins/select/bluegene/plugin/bluegene.c b/src/plugins/select/bluegene/plugin/bluegene.c
index f28d2398695f873575c503771fccd9c8e8465644..02316495be48cbac38ec0124cabdb1703a06832d 100644
--- a/src/plugins/select/bluegene/plugin/bluegene.c
+++ b/src/plugins/select/bluegene/plugin/bluegene.c
@@ -1098,6 +1098,7 @@ extern int create_dynamic_block(ba_request_t *request, List my_block_list)
} else {
num_quarter=4;
}
+
if(_breakup_blocks(request, my_block_list, &block_inx)
!= SLURM_SUCCESS) {
debug2("small block not able to be placed");
@@ -1125,41 +1126,46 @@ extern int create_dynamic_block(ba_request_t *request, List my_block_list)
}
/*Try to put block starting in the smallest of the exisiting blocks*/
- itr = list_iterator_create(bg_list);
- while ((bg_record = (bg_record_t *) list_next(itr)) != NULL) {
- request->rotate_count = 0;
- request->elongate_count = 1;
+ if(!request->start_req) {
+ itr = list_iterator_create(bg_list);
+ while ((bg_record = (bg_record_t *) list_next(itr)) != NULL) {
+ request->rotate_count = 0;
+ request->elongate_count = 1;
- if(bg_record->job_running == -1
- && (bg_record->quarter == (uint16_t) NO_VAL
- || (bg_record->quarter == 0
- && (bg_record->nodecard == (uint16_t) NO_VAL
- || bg_record->nodecard == 0)))) {
-
- for(i=0; i<BA_SYSTEM_DIMENSIONS; i++)
- request->start[i] = bg_record->start[i];
- debug2("allocating %s %d%d%d %d",
- bg_record->nodes,
- request->start[X],
- request->start[Y],
- request->start[Z],
- request->size);
- request->start_req = 1;
- rc = SLURM_SUCCESS;
- if (!allocate_block(request, NULL)){
- debug2("allocate failure for size %d "
- "base partitions",
+ if(bg_record->job_running == -1
+ && (bg_record->quarter == (uint16_t) NO_VAL
+ || (bg_record->quarter == 0
+ && (bg_record->nodecard == (uint16_t) NO_VAL
+ || bg_record->nodecard == 0)))) {
+
+ for(i=0; i<BA_SYSTEM_DIMENSIONS; i++)
+ request->start[i] =
+ bg_record->start[i];
+ debug2("allocating %s %d%d%d %d",
+ bg_record->nodes,
+ request->start[X],
+ request->start[Y],
+ request->start[Z],
request->size);
- rc = SLURM_ERROR;
- } else
- break;
+ request->start_req = 1;
+ rc = SLURM_SUCCESS;
+ if (!allocate_block(request, NULL)){
+ debug2("allocate failure for size %d "
+ "base partitions",
+ request->size);
+ rc = SLURM_ERROR;
+ } else
+ break;
+ }
}
+ list_iterator_destroy(itr);
+
+ request->start_req = 0;
+ for(i=0; i<BA_SYSTEM_DIMENSIONS; i++)
+ request->start[i] = (uint16_t) NO_VAL;
}
- list_iterator_destroy(itr);
-
no_list:
- if(!bg_record) {
- request->start_req = 0;
+ if(!bg_record) {
rc = SLURM_SUCCESS;
if (!allocate_block(request, NULL)){
debug("allocate failure for size %d base partitions",
@@ -2344,6 +2350,27 @@ static int _breakup_blocks(ba_request_t *request, List my_block_list,
continue;
if(bg_record->state != RM_PARTITION_FREE)
continue;
+ if(request->start_req) {
+ if ((request->start[X] != bg_record->start[X])
+ || (request->start[Y] != bg_record->start[Y])
+ || (request->start[Z] != bg_record->start[Z])) {
+ debug4("small got %d%d%d looking for %d%d%d",
+ bg_record->start[X],
+ bg_record->start[Y],
+ bg_record->start[Z],
+ request->start[X],
+ request->start[Y],
+ request->start[Z]);
+ continue;
+ }
+ debug3("small found %d%d%d looking for %d%d%d",
+ bg_record->start[X],
+ bg_record->start[Y],
+ bg_record->start[Z],
+ request->start[X],
+ request->start[Y],
+ request->start[Z]);
+ }
proc_cnt = bg_record->bp_count *
bg_record->cpus_per_bp;
if(proc_cnt == request->procs) {
@@ -2410,6 +2437,28 @@ static int _breakup_blocks(ba_request_t *request, List my_block_list,
!= NULL) {
if(bg_record->job_running != -1)
continue;
+ if(request->start_req) {
+ if ((request->start[X] != bg_record->start[X])
+ || (request->start[Y] != bg_record->start[Y])
+ || (request->start[Z] != bg_record->start[Z])) {
+ debug4("small 2 got %d%d%d looking for %d%d%d",
+ bg_record->start[X],
+ bg_record->start[Y],
+ bg_record->start[Z],
+ request->start[X],
+ request->start[Y],
+ request->start[Z]);
+ continue;
+ }
+ debug3("small 2 found %d%d%d looking for %d%d%d",
+ bg_record->start[X],
+ bg_record->start[Y],
+ bg_record->start[Z],
+ request->start[X],
+ request->start[Y],
+ request->start[Z]);
+ }
+
proc_cnt = bg_record->bp_count * bg_record->cpus_per_bp;
if(proc_cnt == request->procs) {
debug2("found it here %s, %s",
diff --git a/src/sacct/print.c b/src/sacct/print.c
index 8834e4f12fc9ede44228f082cf0ce569e9a73e72..f7f96a642655af7aff328cd9ac1b5865d892eea6 100644
--- a/src/sacct/print.c
+++ b/src/sacct/print.c
@@ -17,7 +17,7 @@
* any later version.
*
* In addition, as a special exception, the copyright holders give permission
- * to link the code of portions of this program with the OpenSSL library under
+ * to link the code of portions of this program with the OpenSSL library under
* certain conditions as described in each individual source file, and
* distribute linked combinations including the two. You must obey the GNU
* General Public License in all respects for all of the code used other than
@@ -217,7 +217,7 @@ void print_idrss(type_t type, void *object)
struct rusage rusage;
char outbuf[FORMAT_STRING_SIZE];
rusage.ru_idrss = 0;
-
+
switch(type) {
case HEADLINE:
printf("%8s", "Idrss");
diff --git a/src/smap/configure_functions.c b/src/smap/configure_functions.c
index 8b5805326d3ebf49105ddf638292a1d0fa821400..5620cc2c123f53e4e92f429882f04267633b7192 100644
--- a/src/smap/configure_functions.c
+++ b/src/smap/configure_functions.c
@@ -17,7 +17,7 @@
* any later version.
*
* In addition, as a special exception, the copyright holders give permission
- * to link the code of portions of this program with the OpenSSL library under
+ * to link the code of portions of this program with the OpenSSL library under
* certain conditions as described in each individual source file, and
* distribute linked combinations including the two. You must obey the GNU
* General Public License in all respects for all of the code used other than
@@ -899,6 +899,10 @@ static int _add_bg_record(blockreq_t *blockreq, List allocated_blocks)
geo[Y] = 0;
geo[Z] = 0;
+ start1[X] = -1;
+ start1[Y] = -1;
+ start1[Z] = -1;
+
end1[X] = -1;
end1[Y] = -1;
end1[Z] = -1;