From 176f0e6494ada4627dcb8f54e896210ac9db7645 Mon Sep 17 00:00:00 2001
From: Moe Jette <jette1@llnl.gov>
Date: Mon, 5 Aug 2002 18:11:57 +0000
Subject: [PATCH] Fixed support for job steps, added dump of incoming RPC data
for job step.
---
src/slurmctld/controller.c | 4 ++-
src/slurmctld/job_mgr.c | 2 +-
src/slurmctld/pack.c | 6 ++--
src/slurmctld/slurmctld.h | 5 ++-
src/slurmctld/step_mgr.c | 62 +++++++++++++++++++++++++++-----------
5 files changed, 55 insertions(+), 24 deletions(-)
diff --git a/src/slurmctld/controller.c b/src/slurmctld/controller.c
index df02f485b98..155e5ed4e65 100644
--- a/src/slurmctld/controller.c
+++ b/src/slurmctld/controller.c
@@ -1,5 +1,5 @@
/*****************************************************************************\
- * controller.c - main control machine daemon for slurm
+ * controller.c - main control machine daemon for slurm
*****************************************************************************
* Copyright (C) 2002 The Regents of the University of California.
* Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER).
@@ -1184,6 +1184,8 @@ slurm_rpc_job_step_create( slurm_msg_t* msg )
start_time = clock ();
+ /* issue the RPC */
+ dump_step_desc ( req_step_msg );
error_code = step_create ( req_step_msg, &step_rec );
/* return result */
diff --git a/src/slurmctld/job_mgr.c b/src/slurmctld/job_mgr.c
index 59edbaaac14..3312c82b2a4 100644
--- a/src/slurmctld/job_mgr.c
+++ b/src/slurmctld/job_mgr.c
@@ -300,7 +300,7 @@ dump_job_desc(job_desc_msg_t * job_specs)
return;
job_id = (job_specs->job_id != NO_VAL) ? job_specs->job_id : -1 ;
- debug3("JobDesc: user_id=%u job_id=%ld partition=%s, name=%s\n",
+ debug3("JobDesc: user_id=%u job_id=%ld partition=%s name=%s\n",
job_specs->user_id, job_id,
job_specs->partition, job_specs->name);
diff --git a/src/slurmctld/pack.c b/src/slurmctld/pack.c
index 928cb13b2e3..32ae18e67fd 100644
--- a/src/slurmctld/pack.c
+++ b/src/slurmctld/pack.c
@@ -1,5 +1,5 @@
/*****************************************************************************\
- * pack.c - pack slurmctld structures into buffers understood by the
+ * pack.c - pack slurmctld structures into buffers understood by the
* slurm_protocol
*****************************************************************************
* Copyright (C) 2002 The Regents of the University of California.
@@ -100,10 +100,10 @@ pack_ctld_job_step_info_reponse_msg( List steps, void** buffer_base, int* buffer
uint32_t list_size = list_count(steps);
current = *buffer_base = xmalloc( buffer_size );
-
pack32( current_time, ¤t, ¤t_size ); /* FIXME What am I really suppose to put as the time?*/
- debug("job_step_count = %d\n");
+ debug("job_step_count = %u\n", list_size);
pack32( list_size , ¤t, ¤t_size );
+
/* Pack the Steps */
while( ( current_step = (struct step_record*)list_next( iterator ) ) != NULL )
{
diff --git a/src/slurmctld/slurmctld.h b/src/slurmctld/slurmctld.h
index cfb4ded2deb..9ccb2cb678b 100644
--- a/src/slurmctld/slurmctld.h
+++ b/src/slurmctld/slurmctld.h
@@ -262,7 +262,10 @@ extern int delete_part_record (char *name);
extern int delete_step_record (struct job_record *job_ptr, uint32_t step_id);
/* dump_job_desc - dump the incoming job submit request message */
-void dump_job_desc(job_desc_msg_t * job_specs);
+extern void dump_job_desc(job_desc_msg_t * job_specs);
+
+/* dump_step_desc - dump the incoming step initiate request message */
+extern void dump_step_desc(step_specs *step_spec);
/* find_job_record - return a pointer to the job record with the given job_id */
extern struct job_record *find_job_record (uint32_t job_id);
diff --git a/src/slurmctld/step_mgr.c b/src/slurmctld/step_mgr.c
index 835fa65ea98..da33410cb64 100644
--- a/src/slurmctld/step_mgr.c
+++ b/src/slurmctld/step_mgr.c
@@ -59,7 +59,7 @@ create_step_record (struct job_record *job_ptr)
step_record_point->job_ptr = job_ptr;
step_record_point->step_id = (job_ptr->next_step_id)++;
- step_record_point->start_time = time( NULL ) ;
+ step_record_point->start_time = time ( NULL ) ;
if (list_append (job_ptr->step_list, step_record_point) == NULL)
fatal ("create_step_record: unable to allocate memory");
@@ -105,6 +105,20 @@ delete_step_record (struct job_record *job_ptr, uint32_t step_id)
}
+/* dump_step_desc - dump the incoming step initiate request message */
+void
+dump_step_desc(step_specs *step_spec)
+{
+ if (step_spec == NULL)
+ return;
+
+ debug3("StepDesc: user_id=%u job_id=%u node_count=%u, cpu_count=%u\n",
+ step_spec->user_id, step_spec->job_id, step_spec->node_count, step_spec->cpu_count);
+ debug3(" relative=%u node_list=%s\n",
+ step_spec->relative, step_spec->node_list);
+}
+
+
/*
* find_step_record - return a pointer to the step record with the given job_id and step_id
* input: job_ptr - pointer to job table entry to have step record added
@@ -266,30 +280,36 @@ pick_step_nodes (struct job_record *job_ptr, step_specs *step_spec ) {
if (job_ptr->node_bitmap == NULL)
return NULL;
- nodes_avail = bit_copy(job_ptr->node_bitmap);
+ nodes_avail = bit_copy (job_ptr->node_bitmap);
if ( step_spec->node_count == INFINITE) /* return all available nodes */
return nodes_avail;
if (step_spec->node_list) {
- if ( step_spec->relative ) {
- /* FIXME need to resolve format of relative_node_list */
- info ("pick_step_nodes: relative_node_list not yet supported");
-
- }
- else {
- error_code = node_name2bitmap (step_spec->node_list, &nodes_picked);
- if (error_code) {
- info ("pick_step_nodes: invalid node list %s", step_spec->node_list);
- goto cleanup;
- }
- if (bit_super_set (nodes_picked, job_ptr->node_bitmap) == 0) {
- info ("pick_step_nodes: requested nodes %s not part of job %u",
- step_spec->node_list, job_ptr->job_id);
- goto cleanup;
- }
+ error_code = node_name2bitmap (step_spec->node_list, &nodes_picked);
+ if (error_code) {
+ info ("pick_step_nodes: invalid node list %s", step_spec->node_list);
+ goto cleanup;
+ }
+ if (bit_super_set (nodes_picked, job_ptr->node_bitmap) == 0) {
+ info ("pick_step_nodes: requested nodes %s not part of job %u",
+ step_spec->node_list, job_ptr->job_id);
+ goto cleanup;
}
}
+ else if (step_spec->relative) {
+ /* Remove first (step_spec->relative) nodes from available list */
+ bitstr_t *relative_nodes = NULL;
+ relative_nodes = bit_pick_cnt (nodes_avail, step_spec->relative);
+ if (relative_nodes == NULL) {
+ info ("pick_step_nodes: Invalid relative value (%u) for job %u",
+ step_spec->relative, job_ptr->job_id);
+ goto cleanup;
+ }
+ bit_not (relative_nodes);
+ bit_and (nodes_avail, relative_nodes);
+ bit_free (relative_nodes);
+ }
else
nodes_picked = bit_alloc (bit_size (nodes_avail) );
@@ -377,6 +397,12 @@ step_create ( step_specs *step_specs, struct step_record** new_step_record )
step_specs->user_id != 0)
return ESLURM_ACCESS_DENIED ;
+ if ((job_ptr->job_state == JOB_COMPLETE) ||
+ (job_ptr->job_state == JOB_FAILED) ||
+ (job_ptr->job_state == JOB_TIMEOUT) ||
+ (job_ptr->job_state == JOB_STAGE_OUT))
+ return ESLURM_ALREADY_DONE;
+
nodeset = pick_step_nodes (job_ptr, step_specs );
if (nodeset == NULL)
--
GitLab