diff --git a/NEWS b/NEWS
index b614c39103518aec7ebffaab8a16d306591664ea..e035f4c92032395710ba4935d024ce4e73ac6236 100644
--- a/NEWS
+++ b/NEWS
@@ -157,6 +157,8 @@ documents those changes that are of interest to users and admins.
is already running.
-- Email messages for job array events print now use the job ID using the
format "#_# (#)" rather than just the internal job ID.
+ -- Set the number of free licenses to be 0 if the global license count decreases
+ and total is less than in use.
* Changes in Slurm 14.03.3-2
============================
diff --git a/doc/html/faq.shtml b/doc/html/faq.shtml
index 00a7a3cce4317fdeadf70c598a8268f823ddce38..1b854f08e9209a90a254d15f9c0749de7b5f6a72 100644
--- a/doc/html/faq.shtml
+++ b/doc/html/faq.shtml
@@ -1809,7 +1809,7 @@ case_sensitive = False
<p><a name="ha_db"><b>54. How critical is configuring high availability for my
database?</b></a></br>
<ul>
-<li>Consider if you really need mysql failover. Short outage of slurdbd is not
+<li>Consider if you really need mysql failover. Short outage of slurmdbd is not
a problem, because slurmctld will store all data in memory and send it to
slurmdbd when it's back operating. The slurmctld daemon will also cache all
user limits and fair share information.</li>
diff --git a/src/common/slurm_protocol_pack.c b/src/common/slurm_protocol_pack.c
index 33c29295a6069322b8af258bef6a74eb89381844..28f118fcbca35860cc3e42441b317bd11f398cd5 100644
--- a/src/common/slurm_protocol_pack.c
+++ b/src/common/slurm_protocol_pack.c
@@ -11376,10 +11376,15 @@ _unpack_license_info_msg(license_info_msg_t **msg,
&zz, buffer);
safe_unpack32(&((*msg)->lic_array[i]).total, buffer);
safe_unpack32(&((*msg)->lic_array[i]).in_use, buffer);
- (*msg)->lic_array[i].available =
- (*msg)->lic_array[i].total -
- (*msg)->lic_array[i].in_use;
- xassert((*msg)->lic_array[i].available >= 0);
+ /* The total number of licenses can decrease
+ * at runtime.
+ */
+ if ((*msg)->lic_array[i].total < (*msg)->lic_array[i].in_use)
+ (*msg)->lic_array[i].available = 0;
+ else
+ (*msg)->lic_array[i].available =
+ (*msg)->lic_array[i].total -
+ (*msg)->lic_array[i].in_use;
safe_unpack8(&((*msg)->lic_array[i]).remote, buffer);
}
diff --git a/src/plugins/sched/backfill/backfill.c b/src/plugins/sched/backfill/backfill.c
index eca08576597afda0b2c22efb64d69cbdba410f72..8ddb1c59defa65c432d30115360badc026d665eb 100644
--- a/src/plugins/sched/backfill/backfill.c
+++ b/src/plugins/sched/backfill/backfill.c
@@ -174,16 +174,15 @@ static void _dump_job_sched(struct job_record *job_ptr, time_t end_time,
xfree(node_list);
}
-static void _dump_job_test(struct job_record *job_ptr, bitstr_t *avail_bitmap)
+static void _dump_job_test(struct job_record *job_ptr, bitstr_t *avail_bitmap,
+ time_t start_time)
{
char begin_buf[32], *node_list;
- if (job_ptr->start_time == 0) {
+ if (start_time == 0)
strcpy(begin_buf, "NOW");
- } else {
- slurm_make_time_str(&job_ptr->start_time, begin_buf,
- sizeof(begin_buf));
- }
+ else
+ slurm_make_time_str(&start_time, begin_buf, sizeof(begin_buf));
node_list = bitmap2node_name(avail_bitmap);
info("Test job %u at %s on %s", job_ptr->job_id, begin_buf, node_list);
xfree(node_list);
@@ -654,6 +653,7 @@ static int _attempt_backfill(void)
uint32_t min_nodes, max_nodes, req_nodes;
bitstr_t *avail_bitmap = NULL, *resv_bitmap = NULL;
bitstr_t *exc_core_bitmap = NULL, *non_cg_bitmap = NULL;
+ bitstr_t *previous_bitmap = NULL;
time_t now, sched_start, later_start, start_res, resv_end;
node_space_map_t *node_space;
struct timeval bf_time1, bf_time2;
@@ -934,6 +934,7 @@ static int _attempt_backfill(void)
/* Determine impact of any resource reservations */
later_start = now;
+ FREE_NULL_BITMAP(previous_bitmap);
TRY_LATER:
if (slurmctld_config.shutdown_time)
break;
@@ -945,7 +946,7 @@ static int _attempt_backfill(void)
job_ptr->time_limit = orig_time_limit;
if (debug_flags & DEBUG_FLAG_BACKFILL) {
END_TIMER;
- info("backfill: completed yielding locks 2"
+ info("backfill: completed yielding locks "
"after testing %d jobs, %s",
job_test_count, TIME_STR);
}
@@ -1030,12 +1031,16 @@ static int _attempt_backfill(void)
/* Test if insufficient nodes remain OR
* required nodes missing OR
- * nodes lack features */
+ * nodes lack features OR
+ * no change since previously tested nodes (only changes
+ * in other partition nodes) */
if ((bit_set_count(avail_bitmap) < min_nodes) ||
((job_ptr->details->req_node_bitmap) &&
(!bit_super_set(job_ptr->details->req_node_bitmap,
avail_bitmap))) ||
- (job_req_node_filter(job_ptr, avail_bitmap))) {
+ (job_req_node_filter(job_ptr, avail_bitmap)) ||
+ (previous_bitmap &&
+ bit_equal(previous_bitmap, avail_bitmap))) {
if (later_start) {
job_ptr->start_time = 0;
goto TRY_LATER;
@@ -1046,6 +1051,9 @@ static int _attempt_backfill(void)
continue;
}
+ FREE_NULL_BITMAP(previous_bitmap);
+ previous_bitmap = bit_copy(avail_bitmap);
+
/* Identify nodes which are definitely off limits */
FREE_NULL_BITMAP(resv_bitmap);
resv_bitmap = bit_copy(avail_bitmap);
@@ -1061,7 +1069,7 @@ static int _attempt_backfill(void)
}
if (debug_flags & DEBUG_FLAG_BACKFILL)
- _dump_job_test(job_ptr, avail_bitmap);
+ _dump_job_test(job_ptr, avail_bitmap, start_res);
j = _try_sched(job_ptr, &avail_bitmap, min_nodes, max_nodes,
req_nodes, exc_core_bitmap);
@@ -1117,7 +1125,7 @@ static int _attempt_backfill(void)
continue;
} else if (rc != SLURM_SUCCESS) {
/* Planned to start job, but something bad
- * happended. */
+ * happened. */
job_ptr->start_time = 0;
break;
} else {
@@ -1191,6 +1199,7 @@ static int _attempt_backfill(void)
FREE_NULL_BITMAP(exc_core_bitmap);
FREE_NULL_BITMAP(resv_bitmap);
FREE_NULL_BITMAP(non_cg_bitmap);
+ FREE_NULL_BITMAP(previous_bitmap);
for (i=0; ; ) {
FREE_NULL_BITMAP(node_space[i].avail_bitmap);