diff --git a/NEWS b/NEWS
index c0798f9f81f3b4a32a05ef62c645ec09dc579e88..ff88b93016b19ca93d1eb8cd180d8b180521ae43 100644
--- a/NEWS
+++ b/NEWS
@@ -86,6 +86,8 @@ documents those changes that are of interest to users and admins.
- Fix bug when changing the time limit of a running job that has previously
been suspended (formerly failed to account for suspend time in setting
termination time).
+ - fix for step allocation to be able to specify only a few nodes in a
+ step and ask for more that specified.
* Changes in SLURM 1.1.13
=========================
diff --git a/src/common/env.c b/src/common/env.c
index ded9a70f4eb3de9964e0fbd6ceff0b3d0296451d..99b88c822d0300f2cc0eea5de94f42e7019895aa 100644
--- a/src/common/env.c
+++ b/src/common/env.c
@@ -16,7 +16,7 @@
* 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
diff --git a/src/slurmctld/proc_req.c b/src/slurmctld/proc_req.c
index fd82fb1bce1c64c6b95d532cf0ba18e1a469ceee..c50f652e741c6c30d6614f678c4cea8fd12059e4 100644
--- a/src/slurmctld/proc_req.c
+++ b/src/slurmctld/proc_req.c
@@ -5,7 +5,8 @@
*****************************************************************************
* Copyright (C) 2002-2006 The Regents of the University of California.
* Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER).
- * Written by Morris Jette <jette@llnl.gov>, Kevin Tew <tew1@llnl.gov>,et. al.
+ * Written by Morris Jette <jette@llnl.gov>, Kevin Tew
+ * <tew1@llnl.gov>, et. al.
* UCRL-CODE-217948.
*
* This file is part of SLURM, a resource management program.
diff --git a/src/slurmctld/step_mgr.c b/src/slurmctld/step_mgr.c
index 509547081c0822cdb2152224d05209f5bc46fccf..d5b013ba0c8cbdac744e3b69db41fe92c82e98ff 100644
--- a/src/slurmctld/step_mgr.c
+++ b/src/slurmctld/step_mgr.c
@@ -16,7 +16,7 @@
* 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
@@ -452,17 +452,21 @@ _pick_step_nodes (struct job_record *job_ptr,
xfree(step_spec->node_list);
step_spec->task_dist = SLURM_DIST_BLOCK;
FREE_NULL_BITMAP(selected_nodes);
- } else {
- /* use selected nodes to run the job */
- FREE_NULL_BITMAP(nodes_avail);
- return selected_nodes;
}
- } else {
- /* set the nodes_avail to be the new set */
- FREE_NULL_BITMAP(nodes_avail);
- nodes_avail = selected_nodes;
step_spec->node_count = bit_set_count(nodes_avail);
}
+ if (selected_nodes) {
+ /* use selected nodes to run the job and
+ * make them unavailable for future use */
+ nodes_picked = bit_copy(selected_nodes);
+ bit_not(selected_nodes);
+ bit_and(nodes_avail, selected_nodes);
+ bit_free(selected_nodes);
+ }
+ } else {
+ nodes_picked = bit_alloc(bit_size(nodes_avail));
+ if (nodes_picked == NULL)
+ fatal("bit_alloc malloc failure");
}
if (step_spec->relative != (uint16_t)NO_VAL) {
@@ -480,13 +484,9 @@ _pick_step_nodes (struct job_record *job_ptr,
bit_not (relative_nodes);
bit_and (nodes_avail, relative_nodes);
bit_free (relative_nodes);
- nodes_picked = bit_alloc(bit_size(nodes_avail));
- if ((nodes_picked == NULL))
- fatal("bit_alloc malloc failure");
} else {
- nodes_picked = bit_alloc(bit_size(nodes_avail));
- nodes_idle = bit_alloc(bit_size(nodes_avail));
- if ((nodes_picked == NULL) || (nodes_idle == NULL))
+ nodes_idle = bit_alloc (bit_size (nodes_avail) );
+ if (nodes_idle == NULL)
fatal("bit_alloc malloc failure");
step_iterator =
list_iterator_create(job_ptr->step_list);
@@ -710,6 +710,7 @@ step_create(job_step_create_request_msg_t *step_specs,
fatal ("create_step_record failed with no memory");
/* set the step_record values */
+
/* Here is where the node list is set for the step */
if(step_specs->node_list
&& step_specs->task_dist == SLURM_DIST_ARBITRARY) {
@@ -720,6 +721,7 @@ step_create(job_step_create_request_msg_t *step_specs,
step_node_list = bitmap2node_name(nodeset);
step_specs->node_list = xstrdup(step_node_list);
}
+
step_ptr->step_node_bitmap = nodeset;
step_ptr->cyclic_alloc =
(uint16_t) (step_specs->task_dist == SLURM_DIST_CYCLIC);
diff --git a/src/srun/allocate.c b/src/srun/allocate.c
index db22359fdc25278b2f482d772d4dd3424fdea2fc..66fd2217eebaefc667a4213d394b0861b8cf25d8 100644
--- a/src/srun/allocate.c
+++ b/src/srun/allocate.c
@@ -16,7 +16,7 @@
* 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
@@ -102,6 +102,7 @@ allocate_nodes(void)
sigset_t oset;
resource_allocation_response_msg_t *resp = NULL;
job_desc_msg_t *j = job_desc_msg_create_from_opts (NULL);
+
if(!j)
return NULL;
@@ -132,21 +133,11 @@ allocate_nodes(void)
if ((rc == 0) && (resp->node_list == NULL)) {
if (resp->error_code)
- verbose("Warning: %s", slurm_strerror(resp->error_code));
+ verbose("Warning: %s",
+ slurm_strerror(resp->error_code));
_wait_for_resources(&resp);
}
- /* For diagnosing a node problem, administrators need to sometimes
- * run a job on N nodes one of which must be the node believed to
- * have a problem (e.g. "srun -N4 -w bad_node diagnostic"). The
- * below logic prevents this from working and necessiates the
- * admin identify four specific nodes to use for the above test
- * instead of just the one bad node. Otherwise only the one
- * bad node is used in the job's allocation. */
- if(resp->node_list && j->req_nodes) {
- xfree(resp->node_list);
- resp->node_list = xstrdup(j->req_nodes);
- }
-
+
done:
xsignal_set_mask(&oset);
xsignal(SIGINT, ointf);
diff --git a/src/srun/srun_job.c b/src/srun/srun_job.c
index 1f1c72d1634b320b0350b8dcbc298b1e1afe707e..91c09d359734e0649d3f00b6f4e1c9087ca8d7f9 100644
--- a/src/srun/srun_job.c
+++ b/src/srun/srun_job.c
@@ -215,8 +215,8 @@ job_step_create_allocation(uint32_t job_id)
hostlist_destroy(hl);
xfree(opt.nodelist);
opt.nodelist = xstrdup(buf);
- xfree(ai->nodelist);
- ai->nodelist = xstrdup(buf);
+ /* xfree(ai->nodelist); */
+/* ai->nodelist = xstrdup(buf); */
}
if(opt.nodelist) {
@@ -228,11 +228,12 @@ job_step_create_allocation(uint32_t job_id)
hostlist_ranged_string(hl, sizeof(buf), buf);
count = hostlist_count(hl);
hostlist_destroy(hl);
- xfree(ai->nodelist);
- ai->nodelist = xstrdup(buf);
+ /* xfree(ai->nodelist); */
+/* ai->nodelist = xstrdup(buf); */
xfree(opt.nodelist);
opt.nodelist = xstrdup(buf);
}
+
if(opt.distribution == SLURM_DIST_ARBITRARY) {
if(count != opt.nprocs) {
error("You asked for %d tasks but specified %d nodes",
@@ -308,14 +309,16 @@ job_step_create_allocation(uint32_t job_id)
}
/* get the correct number of hosts to run tasks on */
- if(opt.nodelist) {
- hl = hostlist_create(opt.nodelist);
- hostlist_uniq(hl);
- ai->nnodes = hostlist_count(hl);
- hostlist_destroy(hl);
- } else if((opt.max_nodes > 0) && (opt.max_nodes < ai->nnodes))
+ /* if(opt.nodelist) { */
+/* hl = hostlist_create(opt.nodelist); */
+/* hostlist_uniq(hl); */
+/* ai->nnodes = hostlist_count(hl); */
+/* hostlist_destroy(hl); */
+/* } else */
+ if((opt.max_nodes > 0) && (opt.max_nodes < ai->nnodes))
ai->nnodes = opt.max_nodes;
-
+/* info("looking for %d nodes out of %s with a must list of %s", */
+/* ai->nnodes, ai->nodelist, opt.nodelist); */
/*
* Create job
*/
@@ -334,7 +337,7 @@ job_create_allocation(resource_allocation_response_msg_t *resp)
{
srun_job_t *job;
allocation_info_t *i = xmalloc(sizeof(*i));
-
+
i->nodelist = _normalize_hostlist(resp->node_list);
i->nnodes = resp->node_cnt;
i->jobid = resp->job_id;