diff --git a/NEWS b/NEWS
index de24533f551fa791e3d571c3763c0d69117dd303..686a8550b3d5f1c090b63a921019fc5e9705a168 100644
--- a/NEWS
+++ b/NEWS
@@ -6,11 +6,11 @@ documents those changes that are of interest to users and admins.
-- Removed sched/gang plugin and moved the logic directly into the slurmctld
daemon so that job preemption and gang scheduling can be used with the
sched/backfill plugin. Added configuration parameter:
- PreemptMode=off|suspend|kill|requeue|checkpoint
+ PreemptMode=off|suspend|cancel|requeue|checkpoint
to enable/disable the job preemption logic (disabled by default).
- (NOTE: There are some problems with memory management which could prevent a
+ (NOTE: There are some problems with memory management which could prevent a
job from starting when memory would be freed by a job being requeued or
- otherwise removed)
+ otherwise removed, these are being worked on)
-- If the --partition option is used with the sinfo or squeue command then
print information about even hidden partitions.
-- Replaced misc cpu allocation members in job_info_t with select_job_res_t
diff --git a/doc/html/preempt.shtml b/doc/html/preempt.shtml
index a181809966204926891adfa462b0c36b0952e74c..5b2d4843ce221ba108e1b472948d9df3e1848336 100644
--- a/doc/html/preempt.shtml
+++ b/doc/html/preempt.shtml
@@ -76,15 +76,19 @@ limits, it will be canceled in order to prevent it from adversely effecting
other jobs sharing the same resources.
</LI>
<LI>
-<B>PreemptMode</B>: Configure to <I>CHECKPOINT</I>, <I>SUSPEND</I> or
-<I>REQUEUE</I> depending on the desired action for low priority jobs.
-A value of <I>CHECKPOINT</I> will checkpoint (if possible) or kill low
-priority jobs.
-Checkpointed jobs are not automatically restarted.
-A value of <I>REQUEUE</I> will requeue (if possible) or kill low priority jobs.
-Requeued jobs are permitted to be restarted on different resources.
-A value of <I>SUSPEND</I> will suspend and automatically resume the low
+<B>PreemptMode</B>: Configure to <I>CANCEL</I>, <I>CHECKPOINT</I>,
+<I>SUSPEND</I> or <I>REQUEUE</I> depending on the desired action for low
priority jobs.
+<UL>
+<LI>A value of <I>CANCEL</I> will always cancel the job.</LI>
+<LI>A value of <I>CHECKPOINT</I> will checkpoint (if possible) or kill low
+priority jobs.</LI>
+Checkpointed jobs are not automatically restarted.
+<LI>A value of <I>REQUEUE</I> will requeue (if possible) or kill low priority
+jobs. Requeued jobs are permitted to be restarted on different resources.</LI>
+<LI>A value of <I>SUSPEND</I> will suspend and automatically resume the low
+priority jobs. </LI>
+</UL>
</LI>
<LI>
<B>Priority</B>: Configure the partition's <I>Priority</I> setting relative to
@@ -309,6 +313,6 @@ It would also necessitate major changes to the shadow bitmap data structures
which are currently used.
</P>
-<p style="text-align:center;">Last modified 14 July 2009</p>
+<p style="text-align:center;">Last modified 20 July 2009</p>
<!--#include virtual="footer.txt"-->
diff --git a/doc/man/man5/slurm.conf.5 b/doc/man/man5/slurm.conf.5
index e84dcf0405034de18e1b51943ca940098dfc59ef..9e468ace43bf524cdac8e029ebeec335169ba20d 100644
--- a/doc/man/man5/slurm.conf.5
+++ b/doc/man/man5/slurm.conf.5
@@ -720,8 +720,8 @@ on SPANK plugins, see the \fBspank\fR(8) manual.
Controls the mechanism used to preempt lower priority jobs in order to
execute higher priority jobs.
For gang scheduling of jobs having the same priority, job suspend will
-always be used and \fBPreemptMode\fR must be set to \fIKILL\fR or \fISUSPEND\fR
-(anything other than OFF).
+always be used and \fBPreemptMode\fR must be set to \fICANCEL\fR,
+\fIREQUEUE\fR or \fISUSPEND\fR (anything other than \fIOFF\fR).
Two different mechanisms can be used to determine when preeption can
take place.
SLURM partitions can have different \fBPriority\fR values and contain
@@ -740,11 +740,14 @@ This is the only option compatible with \fBSchedulerType=sched/wiki\fR
or \fBSchedulerType=sched/wiki2\fR (used by Maui and Moab respecitvely,
which provide their own job preemption functionality).
.TP
+\fBCANCEL\fR
+always cancel the job.
+.TP
\fBCHECKPOINT\fR
-preempts jobs by checkpointing them (if possible) or killing them.
+preempts jobs by checkpointing them (if possible) or cancelling them.
.TP
\fBREQUEUE\fR
-preempts jobs by requeuing them (if possible) or killing them.
+preempts jobs by requeuing them (if possible) or cancelling them.
.TP
\fBSUSPEND\fR
preempts jobs by suspending them.
diff --git a/slurm/slurm.h.in b/slurm/slurm.h.in
index a7d731e3d5bd792fdb7017c2c20939791e380fca..1a6795fbf705962a98160153bf381d25d68d931a 100644
--- a/slurm/slurm.h.in
+++ b/slurm/slurm.h.in
@@ -1109,6 +1109,7 @@ typedef struct reservation_name_msg {
#define PREEMPT_MODE_REQUEUE 0x0002 /* requeue or kill jobs to preempt */
#define PREEMPT_MODE_CHECKPOINT 0x0003 /* checkpoint job to preempt,
* no automatic restart */
+#define PREEMPT_MODE_CANCEL 0x0004 /* always cancel the job */
typedef struct slurm_ctl_conf {
time_t last_update; /* last update time of the build parameters */
diff --git a/src/api/config_info.c b/src/api/config_info.c
index 8245f212951125fe046184dca76f0fd497bdba66..1da7e62df9f6358ab5dee7c7603bdbaad6476274 100644
--- a/src/api/config_info.c
+++ b/src/api/config_info.c
@@ -301,12 +301,14 @@ void slurm_print_ctl_conf ( FILE* out,
if (slurm_ctl_conf_ptr->preempt_mode == PREEMPT_MODE_OFF)
fprintf(out, "PreemptMode = OFF\n");
- else if (slurm_ctl_conf_ptr->preempt_mode == PREEMPT_MODE_SUSPEND)
- fprintf(out, "PreemptMode = SUSPEND\n");
+ else if (slurm_ctl_conf_ptr->preempt_mode == PREEMPT_MODE_CANCEL)
+ fprintf(out, "PreemptMode = CANCEL\n");
else if (slurm_ctl_conf_ptr->preempt_mode == PREEMPT_MODE_CHECKPOINT)
fprintf(out, "PreemptMode = CHECKPOINT\n");
else if (slurm_ctl_conf_ptr->preempt_mode == PREEMPT_MODE_REQUEUE)
fprintf(out, "PreemptMode = REQUEUE\n");
+ else if (slurm_ctl_conf_ptr->preempt_mode == PREEMPT_MODE_SUSPEND)
+ fprintf(out, "PreemptMode = SUSPEND\n");
if (strcmp(slurm_ctl_conf_ptr->priority_type, "priority/basic") == 0) {
fprintf(out, "PriorityType = %s\n",
diff --git a/src/common/read_config.c b/src/common/read_config.c
index d8bc858d00b50b6cc820a7d14783398cb69ca933..f56180da69da6a613de790a6b3c7389449cdc9ab 100644
--- a/src/common/read_config.c
+++ b/src/common/read_config.c
@@ -2121,14 +2121,16 @@ _validate_and_set_defaults(slurm_ctl_conf_t *conf, s_p_hashtbl_t *hashtbl)
if (s_p_get_string(&temp_str, "PreemptMode", hashtbl)) {
if (strcasecmp(temp_str, "off") == 0)
conf->preempt_mode = PREEMPT_MODE_OFF;
+ else if (strcasecmp(temp_str, "cancel") == 0)
+ conf->preempt_mode = PREEMPT_MODE_CANCEL;
+ else if (strcasecmp(temp_str, "checkpoint") == 0)
+ conf->preempt_mode = PREEMPT_MODE_CHECKPOINT;
+ else if (strcasecmp(temp_str, "requeue") == 0)
+ conf->preempt_mode = PREEMPT_MODE_REQUEUE;
else if ((strcasecmp(temp_str, "on") == 0) ||
(strcasecmp(temp_str, "suspend") == 0))
conf->preempt_mode = PREEMPT_MODE_SUSPEND;
- else if ((strcasecmp(temp_str, "kill") == 0) ||
- (strcasecmp(temp_str, "requeue") == 0))
- conf->preempt_mode = PREEMPT_MODE_REQUEUE;
- else if (strcasecmp(temp_str, "checkpoint") == 0)
- conf->preempt_mode = PREEMPT_MODE_CHECKPOINT;
+
else
fatal("Invalid PreemptMode: %s", temp_str);
}
diff --git a/src/plugins/select/cons_res/select_cons_res.c b/src/plugins/select/cons_res/select_cons_res.c
index d929bb08b2072e1332a009386a2421d01de573e6..f955843e47cbd2dd29d3b15449f57f4b73177856 100644
--- a/src/plugins/select/cons_res/select_cons_res.c
+++ b/src/plugins/select/cons_res/select_cons_res.c
@@ -255,7 +255,8 @@ extern bool cr_preemption_enabled(void)
uint16_t mode = slurm_get_preempt_mode();
if (mode == PREEMPT_MODE_SUSPEND)
job_preemption_enabled = true;
- else if ((mode == PREEMPT_MODE_CHECKPOINT) ||
+ else if ((mode == PREEMPT_MODE_CANCEL) ||
+ (mode == PREEMPT_MODE_CHECKPOINT) ||
(mode == PREEMPT_MODE_REQUEUE)) {
job_preemption_enabled = true;
job_preemption_killing = true;
diff --git a/src/plugins/select/linear/select_linear.c b/src/plugins/select/linear/select_linear.c
index 5f745b325b4ca60da601de32651d08cc755309ba..83ab35d3714bb0e5c6502b376ae3738d20177746 100644
--- a/src/plugins/select/linear/select_linear.c
+++ b/src/plugins/select/linear/select_linear.c
@@ -279,7 +279,8 @@ static inline bool _job_preemption_enabled(void)
uint16_t mode = slurm_get_preempt_mode();
if (mode == PREEMPT_MODE_SUSPEND)
job_preemption_enabled = true;
- else if ((mode == PREEMPT_MODE_CHECKPOINT) ||
+ else if ((mode == PREEMPT_MODE_CANCEL) ||
+ (mode == PREEMPT_MODE_CHECKPOINT) ||
(mode == PREEMPT_MODE_REQUEUE)) {
job_preemption_enabled = true;
job_preemption_killing = true;
diff --git a/src/slurmctld/gang.c b/src/slurmctld/gang.c
index b1fc03959c6d165f7602776f15cf700428deddb5..d85f47a2cbe2605bfb5ec0c876ca2842098e3b51 100644
--- a/src/slurmctld/gang.c
+++ b/src/slurmctld/gang.c
@@ -672,6 +672,16 @@ static void _resume_job(uint32_t job_id)
}
}
+static int _cancel_job(uint32_t job_id)
+{
+ int rc;
+
+ rc = job_signal(job_id, SIGKILL, 0, 0);
+ if (rc == SLURM_SUCCESS)
+ info("gang: preempted job %u has been killed", job_id);
+
+ return rc;
+}
static int _checkpoint_job(uint32_t job_id)
{
int rc;
@@ -731,6 +741,8 @@ static void _preempt_job_dequeue(void)
(void) _suspend_job(job_id);
else if (preempt_mode == PREEMPT_MODE_REQUEUE)
rc = _requeue_job(job_id);
+ else if (preempt_mode == PREEMPT_MODE_CANCEL)
+ rc = _cancel_job(job_id);
else if (preempt_mode == PREEMPT_MODE_CHECKPOINT)
rc = _checkpoint_job(job_id);
else
@@ -739,7 +751,7 @@ static void _preempt_job_dequeue(void)
if (rc != SLURM_SUCCESS) {
rc = job_signal(job_id, SIGKILL, 0, 0);
if (rc == SLURM_SUCCESS)
- info("gang: preempted job %u had to be killed",
+ info("gang: preempted job %u had to be killed",
job_id);
else {
info("gang: preempted job %u kill failure %s",