From 9037cdd6e164a2ac92d33cd2a8e543174527f9f0 Mon Sep 17 00:00:00 2001
From: Moe Jette <jette1@llnl.gov>
Date: Thu, 2 Dec 2004 18:04:56 +0000
Subject: [PATCH] Add new job reason value "JobHeld" for jobs with priority==0
---
NEWS | 1 +
slurm/slurm.h.in | 3 ++-
src/common/slurm_protocol_defs.c | 31 ++++++++++++++++---------------
src/slurmctld/job_mgr.c | 13 ++++++++-----
src/slurmctld/node_scheduler.c | 5 ++++-
5 files changed, 31 insertions(+), 22 deletions(-)
diff --git a/NEWS b/NEWS
index ae263f66b09..f7eafd0b5d2 100644
--- a/NEWS
+++ b/NEWS
@@ -3,6 +3,7 @@ documents those changes that are of interest to users and admins.
* Changes in SLURM 0.4.0-pre6
=============================
+ -- Add new job reason value "JobHeld" for jobs with priority==0
* Changes in SLURM 0.4.0-pre5
=============================
diff --git a/slurm/slurm.h.in b/slurm/slurm.h.in
index daedc351623..f78ceb11730 100644
--- a/slurm/slurm.h.in
+++ b/slurm/slurm.h.in
@@ -151,7 +151,8 @@ enum job_wait_reason {
WAIT_RESOUCES, /* required resources not available */
WAIT_PART_NODE_LIMIT, /* request exceeds partition node limit */
WAIT_PART_TIME_LIMIT, /* request exceeds partition time limit */
- WAIT_PART_STATE /* requested partition is down */
+ WAIT_PART_STATE, /* requested partition is down */
+ WAIT_HELD /* job is held, priority==0 */
};
#ifdef HAVE_BGL
diff --git a/src/common/slurm_protocol_defs.c b/src/common/slurm_protocol_defs.c
index 2f8acf87296..64ed8d88d9d 100644
--- a/src/common/slurm_protocol_defs.c
+++ b/src/common/slurm_protocol_defs.c
@@ -422,13 +422,14 @@ void inline slurm_free_checkpoint_resp_msg(checkpoint_resp_msg_t *msg)
extern char *job_reason_string(enum job_wait_reason inx)
{
static char *job_reason_string[] = {
- "None", /* WAIT_NO_REASON */
- "Priority",
- "Dependency",
- "Resources",
- "PartitionNodeLimit",
- "PartitionTimeLimit",
- "PartitionDown"
+ "None", /* WAIT_NO_REASON */
+ "Priority", /* WAIT_PRIORITY */
+ "Dependency", /* WAIT_DEPENDENCY */
+ "Resources", /* WAIT_RESOUCES */
+ "PartitionNodeLimit", /* WAIT_PART_NODE_LIMIT */
+ "PartitionTimeLimit", /* WAIT_PART_TIME_LIMIT */
+ "PartitionDown", /* WAIT_PART_STATE */
+ "JobHeld" /* WAIT_HELD */
};
return job_reason_string[inx];
}
@@ -436,14 +437,14 @@ extern char *job_reason_string(enum job_wait_reason inx)
char *job_state_string(enum job_states inx)
{
static char *job_state_string[] = {
- "PENDING",
- "RUNNING",
- "COMPLETED",
- "CANCELLED",
- "FAILED",
- "TIMEOUT",
- "NODE_FAIL",
- "END"
+ "PENDING", /* JOB_PENDING */
+ "RUNNING", /* JOB_RUNNING */
+ "COMPLETED", /* JOB_COMPLETE */
+ "CANCELLED", /* JOB_CANCELLED */
+ "FAILED", /* JOB_FAILED */
+ "TIMEOUT", /* JOB_TIMEOUT */
+ "NODE_FAIL", /* JOB_NODE_FAIL */
+ "END" /* JOB_END */
};
if (inx & JOB_COMPLETING)
return "COMPLETING";
diff --git a/src/slurmctld/job_mgr.c b/src/slurmctld/job_mgr.c
index b3a84914ad8..465b3d880cc 100644
--- a/src/slurmctld/job_mgr.c
+++ b/src/slurmctld/job_mgr.c
@@ -1219,7 +1219,8 @@ extern int job_allocate(job_desc_msg_t * job_specs, int immediate, int will_run,
if (independent && (!too_fragmented))
top_prio = _top_priority(job_ptr);
else
- top_prio = true; /* don't bother testing */
+ top_prio = true; /* don't bother testing,
+ * it is not runable anyway */
if (immediate && (too_fragmented || (!top_prio) || (!independent))) {
job_ptr->job_state = JOB_FAILED;
job_ptr->start_time = job_ptr->end_time = time(NULL);
@@ -2827,10 +2828,12 @@ static bool _top_priority(struct job_record *job_ptr)
list_iterator_destroy(job_iterator);
}
- if ((!top) && /* not top prio and */
- (job_ptr->priority != 1) && /* not system hold */
- (detail_ptr))
- detail_ptr->wait_reason = WAIT_PRIORITY;
+ if ((!top) && detail_ptr) { /* not top prio */
+ if (job_ptr->priority == 0) /* user/admin hold */
+ detail_ptr->wait_reason = WAIT_HELD;
+ else if (job_ptr->priority != 1) /* not system hold */
+ detail_ptr->wait_reason = WAIT_PRIORITY;
+ }
return top;
}
diff --git a/src/slurmctld/node_scheduler.c b/src/slurmctld/node_scheduler.c
index 4d594ec68a0..49924ace465 100644
--- a/src/slurmctld/node_scheduler.c
+++ b/src/slurmctld/node_scheduler.c
@@ -642,6 +642,8 @@ extern int select_nodes(struct job_record *job_ptr, bool test_only)
fail_reason = WAIT_NO_REASON;
if (part_ptr->state_up == 0)
fail_reason = WAIT_PART_STATE;
+ else if (job_ptr->priority == 0) /* user or administrator hold */
+ fail_reason = WAIT_HELD;
else if (super_user)
; /* ignore any time or node count limits */
else if ((job_ptr->time_limit != NO_VAL) &&
@@ -654,7 +656,8 @@ extern int select_nodes(struct job_record *job_ptr, bool test_only)
if (fail_reason != WAIT_NO_REASON) {
if (detail_ptr)
detail_ptr->wait_reason = fail_reason;
- job_ptr->priority = 1; /* sys hold, move to end of queue */
+ if (job_ptr->priority != 0) /* not user/admin hold */
+ job_ptr->priority = 1; /* sys hold, move to end of queue */
last_job_update = time(NULL);
return ESLURM_REQUESTED_PART_CONFIG_UNAVAILABLE;
}
--
GitLab