node_scheduler.c

/*****************************************************************************\
 *  node_scheduler.c - select and allocated nodes to jobs 
 *	Note: there is a global node table (node_record_table_ptr) 
 *
 *  $Id$
 *****************************************************************************
 *  Copyright (C) 2002-2006 The Regents of the University of California.
 *  Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER).
 *  Written by Morris Jette <jette1@llnl.gov>
 *  UCRL-CODE-226842.
 *  
 *  This file is part of SLURM, a resource management program.
 *  For details, see <http://www.llnl.gov/linux/slurm/>.
 *  
 *  SLURM is free software; you can redistribute it and/or modify it under
 *  the terms of the GNU General Public License as published by the Free
 *  Software Foundation; either version 2 of the License, or (at your option)
 *  any later version.
 *
 *  In addition, as a special exception, the copyright holders give permission 
 *  to link the code of portions of this program with the OpenSSL library under
 *  certain conditions as described in each individual source file, and 
 *  distribute linked combinations including the two. You must obey the GNU 
 *  General Public License in all respects for all of the code used other than 
 *  OpenSSL. If you modify file(s) with this exception, you may extend this 
 *  exception to your version of the file(s), but you are not obligated to do 
 *  so. If you do not wish to do so, delete this exception statement from your
 *  version.  If you delete this exception statement from all source files in 
 *  the program, then also delete it here.
 *  
 *  SLURM is distributed in the hope that it will be useful, but WITHOUT ANY
 *  WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
 *  FOR A PARTICULAR PURPOSE.  See the GNU General Public License for more
 *  details.
 *  
 *  You should have received a copy of the GNU General Public License along
 *  with SLURM; if not, write to the Free Software Foundation, Inc.,
 *  51 Franklin Street, Fifth Floor, Boston, MA 02110-1301  USA.
\*****************************************************************************/

#ifdef HAVE_CONFIG_H
#  include "config.h"
#endif

#include <errno.h>
#include <pthread.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <syslog.h>
#include <unistd.h>

#include <slurm/slurm_errno.h>

#include "src/common/hostlist.h"
#include "src/common/node_select.h"
#include "src/common/xassert.h"
#include "src/common/xmalloc.h"
#include "src/common/xstring.h"

#include "src/slurmctld/agent.h"
#include "src/slurmctld/node_scheduler.h"
#include "src/slurmctld/sched_plugin.h"
#include "src/slurmctld/slurmctld.h"

#define MAX_RETRIES 10

struct node_set {		/* set of nodes with same configuration */
	uint32_t cpus_per_node;	/* NOTE: This is the minimum count,
				 * if FastSchedule==0 then individual 
				 * nodes within the same configuration 
				 * line (in slurm.conf) can actually 
				 * have different CPU counts */
	uint32_t real_memory;
	uint32_t nodes;
	uint32_t weight;
	int feature;
	bitstr_t *my_bitmap;
};

static int _add_node_set_info(struct node_set *node_set_ptr, 
			      bitstr_t ** node_bitmap, 
			      int *node_cnt, int *cpu_cnt, 
			      const int mem_cnt, int cr_enabled,
			      struct job_record *job);
static int  _build_node_list(struct job_record *job_ptr, 
			     struct node_set **node_set_pptr,
			     int *node_set_size);
static void _filter_nodes_in_set(struct node_set *node_set_ptr,
				 struct job_details *detail_ptr);
static int _match_feature(char *seek, char *available);
static int _nodes_in_sets(bitstr_t *req_bitmap, 
			  struct node_set * node_set_ptr, 
			  int node_set_size);
static void _node_load_bitmaps(bitstr_t * bitmap, bitstr_t ** no_load_bit, 
			       bitstr_t ** light_load_bit, 
			       bitstr_t ** heavy_load_bit);
static int _pick_best_load(struct job_record *job_ptr, bitstr_t * bitmap, 
			uint32_t min_nodes, uint32_t max_nodes, 
			uint32_t req_nodes, bool test_only);
static int _pick_best_nodes(struct node_set *node_set_ptr,
			    int node_set_size, bitstr_t ** select_bitmap,
			    struct job_record *job_ptr,
			    struct part_record *part_ptr,
			    uint32_t min_nodes, uint32_t max_nodes,
			    uint32_t req_nodes);
static int _valid_features(char *requested, char *available);


/*
 * allocate_nodes - change state of specified nodes to NODE_STATE_ALLOCATED
 * IN job_ptr - job being allocated resources
 * globals: node_record_count - number of nodes in the system
 *	node_record_table_ptr - pointer to global node table
 *	last_node_update - last update time of node table
 */
extern void allocate_nodes(struct job_record *job_ptr)
{
	int i;

	last_node_update = time(NULL);

	for (i = 0; i < node_record_count; i++) {
		if (bit_test(job_ptr->node_bitmap, i))
			make_node_alloc(&node_record_table_ptr[i], job_ptr);
	}
	return;
}


/*
 * count_cpus - report how many cpus are associated with the identified nodes 
 * IN bitmap - map of nodes to tally
 * RET cpu count
 * globals: node_record_count - number of nodes configured
 *	node_record_table_ptr - pointer to global node table
 */
extern int count_cpus(unsigned *bitmap)
{
	int i, sum;

	sum = 0;
	for (i = 0; i < node_record_count; i++) {
		if (bit_test(bitmap, i) != 1)
			continue;
		if (slurmctld_conf.fast_schedule)
			sum += node_record_table_ptr[i].config_ptr->cpus;
		else
			sum += node_record_table_ptr[i].cpus;
	}
	return sum;
}


/*
 * deallocate_nodes - for a given job, deallocate its nodes and make 
 *	their state NODE_STATE_COMPLETING
 * IN job_ptr - pointer to terminating job (already in some COMPLETING state)
 * IN timeout - true if job exhausted time limit, send REQUEST_KILL_TIMELIMIT
 *	RPC instead of REQUEST_TERMINATE_JOB
 * IN suspended - true if job was already suspended (node's job_run_cnt 
 *	already decremented);
 * globals: node_record_count - number of nodes in the system
 *	node_record_table_ptr - pointer to global node table
 */
extern void deallocate_nodes(struct job_record *job_ptr, bool timeout, 
		bool suspended)
{
	int i;
	kill_job_msg_t *kill_job = NULL;
	agent_arg_t *agent_args = NULL;
	int down_node_cnt = 0;
	uint16_t base_state;

	xassert(job_ptr);
	xassert(job_ptr->details);

	if (select_g_job_fini(job_ptr) != SLURM_SUCCESS)
		error("select_g_job_fini(%u): %m", job_ptr->job_id);

	agent_args = xmalloc(sizeof(agent_arg_t));
	if (timeout)
		agent_args->msg_type = REQUEST_KILL_TIMELIMIT;
	else
		agent_args->msg_type = REQUEST_TERMINATE_JOB;
	agent_args->retry = 0;	/* re_kill_job() resends as needed */
	agent_args->hostlist = hostlist_create("");
	kill_job = xmalloc(sizeof(kill_job_msg_t));
	last_node_update = time(NULL);
	kill_job->job_id  = job_ptr->job_id;
	kill_job->job_uid = job_ptr->user_id;
	kill_job->nodes   = xstrdup(job_ptr->nodes);
	kill_job->time    = time(NULL);
	kill_job->select_jobinfo = select_g_copy_jobinfo(
			job_ptr->select_jobinfo);

	for (i = 0; i < node_record_count; i++) {
		struct node_record *node_ptr = &node_record_table_ptr[i];
		if (bit_test(job_ptr->node_bitmap, i) == 0)
			continue;