diff --git a/NEWS b/NEWS
index a4ac5fe60e28a948c046465cc53d5b5f30666836..afcee9e1599fa4adc772069f2a88d43216f5a8cd 100644
--- a/NEWS
+++ b/NEWS
@@ -171,6 +171,10 @@ documents those changes that are of interest to users and admins.
-- Fix slurmstepd core dump.
-- Modify the description of -E and -S option of sacct command as point in time
'before' or 'after' the database records are returned.
+ -- Correct support for partition with Shared=YES configuration.
+ -- If job requests --exclusive then do not use nodes which have any cores in an
+ advanced reservation. Also prevents case where nodes can be shared by other
+ jobs.
* Changes in Slurm 14.03.3-2
============================
diff --git a/doc/html/faq.shtml b/doc/html/faq.shtml
index 1b854f08e9209a90a254d15f9c0749de7b5f6a72..6bd15524a1d3b77d81afc8ea405db7e12fe5196a 100644
--- a/doc/html/faq.shtml
+++ b/doc/html/faq.shtml
@@ -403,8 +403,8 @@ as part of the process launch.</p>
<p><a name="prompt"><b>12. How can I get shell prompts in interactive
mode?</b></a><br>
-<i>srun -u bash -i</i><br>
-Srun's <i>-u</i> option turns off buffering of stdout.
+<i>srun --pty bash -i</i><br>
+Srun's <i>--pty</i> option runs task zero in pseudo terminal mode.
Bash's <i>-i</i> option tells it to run in interactive mode (with prompts).
<p><a name="batch_out"><b>13. How can I get the task ID in the output
@@ -1931,6 +1931,6 @@ sacctmgr delete user name=adam cluster=tux account=chemistry
<p class="footer"><a href="#top">top</a></p>
-<p style="text-align:center;">Last modified 30 April 2014</p>
+<p style="text-align:center;">Last modified 12 June 2014</p>
<!--#include virtual="footer.txt"-->
diff --git a/src/plugins/sched/backfill/backfill.c b/src/plugins/sched/backfill/backfill.c
index d9a96a8c86b99992cc35e51501bff5fb50cd5c76..68dfedf021899f1188e84c17eb34fbefb948affe 100644
--- a/src/plugins/sched/backfill/backfill.c
+++ b/src/plugins/sched/backfill/backfill.c
@@ -1189,8 +1189,14 @@ static int _attempt_backfill(void)
job_ptr->time_limit = orig_time_limit;
}
- if (later_start &&
- (job_ptr->start_time > (later_start+backfill_resolution))) {
+ start_time = job_ptr->start_time;
+ end_reserve = job_ptr->start_time + (time_limit * 60);
+ start_time = (start_time / backfill_resolution) *
+ backfill_resolution;
+ end_reserve = (end_reserve / backfill_resolution) *
+ backfill_resolution;
+
+ if (later_start && (start_time > later_start)) {
/* Try later when some nodes currently reserved for
* pending jobs are free */
job_ptr->start_time = 0;
@@ -1210,12 +1216,6 @@ static int _attempt_backfill(void)
break;
}
- start_time = job_ptr->start_time;
- end_reserve = job_ptr->start_time + (time_limit * 60);
- start_time = (start_time / backfill_resolution) *
- backfill_resolution;
- end_reserve = (end_reserve / backfill_resolution) *
- backfill_resolution;
if ((job_ptr->start_time > now) &&
_test_resv_overlap(node_space, avail_bitmap,
start_time, end_reserve)) {
diff --git a/src/plugins/select/cons_res/job_test.c b/src/plugins/select/cons_res/job_test.c
index c5d829bc51f18aa64284cea88737a3671c7ccd40..7eddf74dc680f6a7c5454864a9a15b7621cfcf23 100644
--- a/src/plugins/select/cons_res/job_test.c
+++ b/src/plugins/select/cons_res/job_test.c
@@ -634,10 +634,11 @@ static int _verify_node_state(struct part_res_record *cr_part_ptr,
bitstr_t *node_bitmap,
uint16_t cr_type,
struct node_use_record *node_usage,
- enum node_cr_state job_node_req)
+ enum node_cr_state job_node_req,
+ bitstr_t *exc_core_bitmap)
{
struct node_record *node_ptr;
- uint32_t i, free_mem, gres_cpus, gres_cores, min_mem;
+ uint32_t i, j, free_mem, gres_cpus, gres_cores, min_mem;
int core_start_bit, core_end_bit, cpus_per_core;
List gres_list;
int i_first, i_last;
@@ -683,6 +684,17 @@ static int _verify_node_state(struct part_res_record *cr_part_ptr,
}
}
+ /* Exclude nodes with reserved cores */
+ if (job_ptr->details->whole_node && exc_core_bitmap) {
+ for (j = core_start_bit; j <= core_end_bit; j++) {
+ if (bit_test(exc_core_bitmap, j))
+ continue;
+ debug3("cons_res: _vns: node %s exc",
+ select_node_record[i].node_ptr->name);
+ goto clear_bit;
+ }
+ }
+
/* node-level gres check */
if (node_usage[i].gres_list)
gres_list = node_usage[i].gres_list;
@@ -2024,6 +2036,35 @@ static uint16_t *_select_nodes(struct job_record *job_ptr, uint32_t min_nodes,
return cpus;
}
+/* When any cores on a node are removed from being available for a job,
+ * then remove the entire node from being available. */
+static void _block_whole_nodes(bitstr_t *node_bitmap,
+ bitstr_t *orig_core_bitmap,
+ bitstr_t *new_core_bitmap)
+{
+ int first_node, last_node, i_node;
+ int first_core, last_core, i_core;
+
+ first_node = bit_ffs(node_bitmap);
+ if (first_node >= 0)
+ last_node = bit_fls(node_bitmap);
+ else
+ last_node = first_node - 1;
+
+ for (i_node = first_node; i_node <= last_node; i_node++) {
+ if (!bit_test(node_bitmap, i_node))
+ continue;
+ first_core = cr_get_coremap_offset(i_node);
+ last_core = cr_get_coremap_offset(i_node + 1) - 1;
+ for (i_core = first_core; i_core <= last_core; i_core++) {
+ if ( bit_test(orig_core_bitmap, i_core) &&
+ !bit_test(new_core_bitmap, i_core)) {
+ bit_clear(node_bitmap, i_node);
+ break;
+ }
+ }
+ }
+}
/* cr_job_test - does most of the real work for select_p_job_test(), which
* includes contiguous selection, load-leveling and max_share logic
@@ -2084,7 +2125,8 @@ extern int cr_job_test(struct job_record *job_ptr, bitstr_t *node_bitmap,
if (!test_only) {
error_code = _verify_node_state(cr_part_ptr, job_ptr,
node_bitmap, cr_type,
- node_usage, job_node_req);
+ node_usage, job_node_req,
+ exc_core_bitmap);
if (error_code != SLURM_SUCCESS) {
return error_code;
}
@@ -2214,10 +2256,11 @@ extern int cr_job_test(struct job_record *job_ptr, bitstr_t *node_bitmap,
}
}
if (exc_core_bitmap) {
+#if _DEBUG
char str[100];
bit_fmt(str, (sizeof(str) - 1), exc_core_bitmap);
debug2("excluding cores reserved: %s", str);
-
+#endif
bit_not(exc_core_bitmap);
bit_and(free_cores, exc_core_bitmap);
bit_not(exc_core_bitmap);
@@ -2234,7 +2277,6 @@ extern int cr_job_test(struct job_record *job_ptr, bitstr_t *node_bitmap,
bit_copybits(tmpcore, p_ptr->row[i].row_bitmap);
bit_not(tmpcore); /* set bits now "free" resources */
bit_and(free_cores, tmpcore);
-
if (p_ptr->part_ptr != job_ptr->part_ptr)
continue;
if (part_core_map) {
@@ -2245,6 +2287,9 @@ extern int cr_job_test(struct job_record *job_ptr, bitstr_t *node_bitmap,
}
}
}
+ if (job_ptr->details->whole_node)
+ _block_whole_nodes(node_bitmap, avail_cores, free_cores);
+
cpu_count = _select_nodes(job_ptr, min_nodes, max_nodes, req_nodes,
node_bitmap, cr_node_cnt, free_cores,
node_usage, cr_type, test_only,
@@ -2307,7 +2352,7 @@ extern int cr_job_test(struct job_record *job_ptr, bitstr_t *node_bitmap,
(p_ptr->part_ptr->preempt_mode != PREEMPT_MODE_OFF)) {
if (select_debug_flags & DEBUG_FLAG_SELECT_TYPE) {
info("cons_res: cr_job_test: continuing on "
- "part: %s ", p_ptr->part_ptr->name);
+ "part: %s", p_ptr->part_ptr->name);
}
continue;
}
@@ -2321,6 +2366,8 @@ extern int cr_job_test(struct job_record *job_ptr, bitstr_t *node_bitmap,
bit_and(free_cores, tmpcore);
}
}
+ if (job_ptr->details->whole_node)
+ _block_whole_nodes(node_bitmap, avail_cores, free_cores);
/* make these changes permanent */
bit_copybits(avail_cores, free_cores);
cpu_count = _select_nodes(job_ptr, min_nodes, max_nodes, req_nodes,
diff --git a/src/plugins/select/cons_res/select_cons_res.h b/src/plugins/select/cons_res/select_cons_res.h
index 27ad9088eb98b2e35abb857102b78b06d88b33f8..caf75ce3bbd727b7b9e4115cf57fe029a70d0dab 100644
--- a/src/plugins/select/cons_res/select_cons_res.h
+++ b/src/plugins/select/cons_res/select_cons_res.h
@@ -67,7 +67,8 @@
/* a partition's per-row CPU allocation data */
struct part_row_data {
- bitstr_t *row_bitmap; /* contains all jobs for this row */
+ bitstr_t *row_bitmap; /* contains core bitmap for all jobs in
+ * this row */
uint32_t num_jobs; /* Number of jobs in this row */
struct job_resources **job_list;/* List of jobs in this row */
uint32_t job_list_size; /* Size of job_list array */
diff --git a/src/slurmctld/node_scheduler.c b/src/slurmctld/node_scheduler.c
index e46deb4338ee31d3c7404efb0ca4a45b8d7aad1c..462e9de7b12267df49da52a66f08ed50d1e666db 100644
--- a/src/slurmctld/node_scheduler.c
+++ b/src/slurmctld/node_scheduler.c
@@ -700,14 +700,16 @@ _resolve_shared_status(struct job_record *job_ptr, uint16_t part_max_share,
return 1;
if (cons_res_flag) {
+ if (part_max_share == 1) /* partition configured Shared=NO */
+ return 0;
if ((job_ptr->details->share_res == 0) ||
+ (job_ptr->details->share_res == (uint8_t) NO_VAL) ||
(job_ptr->details->whole_node == 1))
return 0;
return 1;
} else {
job_ptr->details->whole_node = 1;
- /* no sharing if partition Shared=NO */
- if (part_max_share == 1)
+ if (part_max_share == 1) /* partition configured Shared=NO */
return 0;
/* share if the user requested it */
if (job_ptr->details->share_res == 1)