getqueues.c

openPBS的开放源代码

/*
*         OpenPBS (Portable Batch System) v2.3 Software License
* 
* Copyright (c) 1999-2000 Veridian Information Solutions, Inc.
* All rights reserved.
* 
* ---------------------------------------------------------------------------
* For a license to use or redistribute the OpenPBS software under conditions
* other than those described below, or to purchase support for this software,
* please contact Veridian Systems, PBS Products Department ("Licensor") at:
* 
*    www.OpenPBS.org  [phone]                  [e-mail]
*                        877 902-4PBS (US toll-free)
* ---------------------------------------------------------------------------
* 
* This license covers use of the OpenPBS v2.3 software (the "Software") at
* your site or location, and, for certain users, redistribution of the
* Software to other sites and locations.  Use and redistribution of
* OpenPBS v2.3 in source and binary forms, with or without modification,
* are permitted provided that all of the following conditions are met.
* After December 31, 2001, only conditions 3-6 must be met:
* 
* 1. Commercial and/or non-commercial use of the Software is permitted
*    provided a current software registration is on file at www.OpenPBS.org.
*    If use of this software contributes to a publication, product, or
*    service, proper attribution must be given; see www.OpenPBS.org/credit.html
* 
* 2. Redistribution in any form is only permitted for non-commercial,
*    non-profit purposes.  There can be no charge for the Software or any
*    software incorporating the Software.  Further, there can be no
*    expectation of revenue generated as a consequence of redistributing
*    the Software.
* 
* 3. Any Redistribution of source code must retain the above copyright notice
*    and the acknowledgment contained in paragraph 6, this list of conditions
*    and the disclaimer contained in paragraph 7.
* 
* 4. Any Redistribution in binary form must reproduce the above copyright
*    notice and the acknowledgment contained in paragraph 6, this list of
*    conditions and the disclaimer contained in paragraph 7 in the
*    documentation and/or other materials provided with the distribution.
* 
* 5. Redistributions in any form must be accompanied by information on how to
*    obtain complete source code for the OpenPBS software and any
*    modifications and/or additions to the OpenPBS software.  The source code
*    must either be included in the distribution or be available for no more
*    than the cost of distribution plus a nominal fee, and all modifications
*    and additions to the Software must be freely redistributable by any party
*    (including Licensor) without restriction.
* 
* 6. All advertising materials mentioning features or use of the Software must
*    display the following acknowledgment:
* 
*     "This product includes software developed by NASA Ames Research Center,
*     Lawrence Livermore National Laboratory, and Veridian Information 
*     Solutions, Inc.
*     Visit www.OpenPBS.org for OpenPBS software support,
*     products, and information."
* 
* 7. DISCLAIMER OF WARRANTY
* 
* THIS SOFTWARE IS PROVIDED "AS IS" WITHOUT WARRANTY OF ANY KIND. ANY EXPRESS
* OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
* OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE, AND NON-INFRINGEMENT
* ARE EXPRESSLY DISCLAIMED.
* 
* IN NO EVENT SHALL VERIDIAN CORPORATION, ITS AFFILIATED COMPANIES, OR THE
* U.S. GOVERNMENT OR ANY OF ITS AGENCIES BE LIABLE FOR ANY DIRECT OR INDIRECT,
* INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
* LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA,
* OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
* LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
* NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE,
* EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
* 
* This license will be governed by the laws of the Commonwealth of Virginia,
* without reference to its choice of law rules.
*/
/* $Id: getqueues.c,v 1.2.2.1.2.4 2000/08/09 00:19:16 hender Exp $ */

#include <stdio.h>
#include <stdlib.h>
#include <errno.h>
#include <sys/types.h>
#include <sys/param.h>
#include <sys/sysmacros.h>
#include <string.h>

/* PBS header files */

#include "pbs_error.h"
#include "pbs_ifl.h"
#include "log.h"

/* Scheduler header files */

#include "toolkit.h"
#include "gblxvars.h"

extern int connector;

static int  queue_claim_jobs(Queue *queue, Job **joblist_ptr);
static int  queue_sanity(Queue *queue);
#ifdef NODEMASK
static int  find_nodemasks(Queue *queue, Resources *rsrcs);
#endif	/* NODEMASK */

/*
 * schd_get_queue_limits - query queue information from the server.
 *
 * Returns 0 on success, -1 for "fatal errors", and 1 for a transient
 * error (i.e., the queue failed the sanity checks imposed by the
 * queue_sanity() function).
 */
int 
schd_get_queue_limits(Queue *queue)
{
    char   *id = "schd_get_queue_limits";
    int     moved = 0, istrue;
    Batch_Status *bs;
    AttrList *attr;
    static AttrList alist[] =
    {
	{&alist[1],  ATTR_start,	"", ""},
	{&alist[2],  ATTR_enable,	"", ""},
	{&alist[3],  ATTR_count,	"", ""},
	{&alist[4],  ATTR_maxuserrun,	"", ""},
	{&alist[5],  ATTR_rescavail,	"", ""},
	{&alist[6],  ATTR_rescassn,	"", ""},
	{&alist[7],  ATTR_rescdflt,	"", ""},
	{&alist[8],  ATTR_rescmax,	"", ""},
	{&alist[9],  ATTR_rescmin,	"", ""},
	{&alist[10], ATTR_acluren,	"", ""},
	{&alist[11], ATTR_acluser,	"", ""},
	{NULL,       ATTR_maxrun,	"", ""}
    };
    size_t  mem_default = UNSPECIFIED;
    size_t  mem_assn    = UNSPECIFIED;
    size_t  mem_max     = UNSPECIFIED;
    size_t  mem_min     = UNSPECIFIED;
    int     cpu_default = UNSPECIFIED;
    int     cpu_assn    = UNSPECIFIED;
    int     cpu_max     = UNSPECIFIED;
    int     cpu_min     = UNSPECIFIED;
    int     nodes_from_cpu, nodes_from_mem;

    queue->running	= UNSPECIFIED;
    queue->queued	= UNSPECIFIED;
    queue->maxrun	= UNSPECIFIED;
    queue->userrun      = UNSPECIFIED;

    queue->nodes_max	= UNSPECIFIED;
    queue->nodes_min	= UNSPECIFIED;
    queue->nodes_default= UNSPECIFIED;
    queue->nodes_assn	= UNSPECIFIED;
    queue->nodes_rsvd	= UNSPECIFIED;
    queue->wallt_max	= UNSPECIFIED;
    queue->wallt_min	= UNSPECIFIED;
    queue->wallt_default= UNSPECIFIED;

    queue->flags	= 0;
#ifdef NODEMASK
    BITFIELD_CLRALL(&queue->queuemask);
    BITFIELD_CLRALL(&queue->availmask);
#endif	/* NODEMASK */
    queue->rsrcs	= NULL;
    
    if (queue->jobs) {
	DBPRT(("%s: found jobs on queue '%s'!  Freeing them...\n", id, 
	    queue->qname));
	schd_free_jobs(queue->jobs);
    }

    if (queue->useracl) {
	DBPRT(("%s: found user ACL list on queue '%s'!  Freeing it...\n", id, 
	    queue->qname));
	schd_free_useracl(queue->useracl);
    }

    queue->jobs         = NULL;
    queue->useracl	= NULL;

    /* Ask the server for information about the specified queue. */

    if ((bs = pbs_statque(connector, queue->qname, alist, NULL)) == NULL) {
	sprintf(log_buffer, "pbs_statque failed, \"%s\" %d",
		queue->qname, pbs_errno);
	log_record(PBSEVENT_ERROR, PBS_EVENTCLASS_SERVER, id, log_buffer);
	DBPRT(("%s: %s\n", id, log_buffer));
	return (-1);
    }

    /* Process the list of attributes returned by the server. */
	
    for (attr = bs->attribs; attr != NULL; attr = attr->next) {

	/* Is queue started? */
	if (!strcmp(attr->name, ATTR_start)) {
	    if (schd_val2bool(attr->value, &istrue) == 0) {
		if (istrue)	/* if true, queue is not stopped. */
		    queue->flags &= ~QFLAGS_STOPPED;
		else 
		    queue->flags |= QFLAGS_STOPPED;
	    } else {
		DBPRT(("%s: couldn't parse attr %s value %s to boolean\n", id,
		    attr->name, attr->value));
	    }
	    continue;
	}

	/* Is queue enabled? */
	if (!strcmp(attr->name, ATTR_enable)) {
	    if (schd_val2bool(attr->value, &istrue) == 0) {
		if (istrue)	/* if true, queue is not disabled. */
		    queue->flags &= ~QFLAGS_DISABLED;
		else 
		    queue->flags |= QFLAGS_DISABLED;
	    } else {
		DBPRT(("%s: couldn't parse attr %s value %s to boolean\n", id,
		    attr->name, attr->value));
	    }
	    continue;
	}

	/* How many jobs are queued and running? */
	if (!strcmp(attr->name, ATTR_count)) {
	    queue->queued = schd_how_many(attr->value, SC_QUEUED);
	    queue->running = schd_how_many(attr->value, SC_RUNNING);
	    continue;
	}

	/* Queue-wide maximum number of jobs running. */
	if (!strcmp(attr->name, ATTR_maxrun)) {
	    queue->maxrun = atoi(attr->value);
	    continue;
	}

	/* Per-user maximum number of jobs running. */
	if (!strcmp(attr->name, ATTR_maxuserrun)) {
	    queue->userrun = atoi(attr->value);
	    continue;
	}

	/* Is there an enabled user access control list on this queue? */
	if (!strcmp(attr->name, ATTR_acluren)) {
	    if (schd_val2bool(attr->value, &istrue) == 0) {
		if (istrue)	/* if true, queue has an ACL */
		    queue->flags |= QFLAGS_USER_ACL;
		else 
		    queue->flags &= ~QFLAGS_USER_ACL;
	    } else {
		DBPRT(("%s: couldn't parse attr %s value %s to boolean\n", id,
		    attr->name, attr->value));
	    }
	    continue;
	}

	if (!strcmp(attr->name, ATTR_acluser)) {
	    if (queue->useracl) {
		DBPRT(("queue %s acluser already set!\n", queue->qname));
		schd_free_useracl(queue->useracl);
	    }
	    queue->useracl = schd_create_useracl(attr->value);
	    continue;
	}

	/* Queue maximum resource usage. */
	if (!strcmp(attr->name, ATTR_rescmax)) {
	    if (!strcmp("mem", attr->resource)) {
		mem_max = schd_val2byte(attr->value);
		continue;
	    }

	    if (!strcmp("ncpus", attr->resource)) {
		cpu_max = atoi(attr->value);
		continue;
	    }

	    if (!strcmp("walltime", attr->resource)) {
		queue->wallt_max = schd_val2sec(attr->value);
		continue;
	    }

#ifdef NODEMASK
	    if (!strcmp("nodemask", attr->resource)) {
		if (schd_str2mask(attr->value, &queue->queuemask)) {
		    (void)sprintf(log_buffer, "couldn't convert nodemask %s", 
			attr->value);
		    log_record(PBSEVENT_SYSTEM, PBS_EVENTCLASS_SERVER, id,
			log_buffer);
		} else 
		    queue->flags |= QFLAGS_NODEMASK;	/* Valid nodemask. */
	    }
#endif	/* NODEMASK */

	    continue;
	}

	/* Queue minimum resource usage. */
	if (!strcmp(attr->name, ATTR_rescmin)) {
	    if (!strcmp("mem", attr->resource)) {
		mem_min = schd_val2byte(attr->value);
		continue;
	    }

	    if (!strcmp("ncpus", attr->resource)) {
		cpu_min = atoi(attr->value);
		continue;
	    }

	    if (!strcmp("walltime", attr->resource)) {
		queue->wallt_min = schd_val2sec(attr->value);
		continue;
	    }

	    continue;
	}

	/* Queue assigned (in use) resource usage. */
	if (!strcmp(attr->name, ATTR_rescassn)) {
	    if (!strcmp("mem", attr->resource)) {
		mem_assn = schd_val2byte(attr->value);
		continue;
	    }

	    if (!strcmp("ncpus", attr->resource)) {
		cpu_assn = atoi(attr->value);
	    }

	    continue;
	}

	if (!strcmp(attr->name, ATTR_rescdflt)) {
	    if (!strcmp("mem", attr->resource))	{
		mem_default = schd_val2byte(attr->value);
		continue;
	    }

	    if (!strcmp("ncpus", attr->resource)) {
		cpu_default = atoi(attr->value);
		continue;
	    }

	    if (!strcmp("walltime", attr->resource))
		queue->wallt_default = schd_val2sec(attr->value);
	}

	/* Ignore anything else */
    }

    pbs_statfree(bs);

    /*
     * Calculate values for queue node limits, given memory and cpu values.
     * Note any discrepancies.
     */
    nodes_from_cpu = NODES_FROM_CPU(cpu_default);
    nodes_from_mem = NODES_FROM_MEM(mem_default);
    if (nodes_from_cpu != nodes_from_mem) {
	sprintf(log_buffer, 
	    "%s: Queue '%s' default cpu/mem (%d/%s) convert to %d != %d nodes",
	    id, queue->qname, cpu_default, schd_byte2val(mem_default), 
	    nodes_from_cpu, nodes_from_mem);
	log_record(PBSEVENT_ERROR, PBS_EVENTCLASS_SERVER, id, log_buffer);
	DBPRT(("%s: %s\n", id, log_buffer));
    }

    nodes_from_cpu = NODES_FROM_CPU(cpu_max);
    nodes_from_mem = NODES_FROM_MEM(mem_max);
    if (nodes_from_cpu != nodes_from_mem) {
	sprintf(log_buffer, 
	    "%s: Queue '%s' maximum cpu/mem (%d/%s) convert to %d != %d nodes",
	    id, queue->qname, cpu_max, schd_byte2val(mem_max), 
	    nodes_from_cpu, nodes_from_mem);
	log_record(PBSEVENT_ERROR, PBS_EVENTCLASS_SERVER, id, log_buffer);
	DBPRT(("%s: %s\n", id, log_buffer));
    }

    nodes_from_cpu = NODES_FROM_CPU(cpu_min);
    nodes_from_mem = NODES_FROM_MEM(mem_min);
    if (nodes_from_cpu != nodes_from_mem) {
	sprintf(log_buffer, 
	    "%s: Queue '%s' minimum cpu/mem (%d/%s) convert to %d != %d nodes",
	    id, queue->qname, cpu_min, schd_byte2val(mem_min), 
	    nodes_from_cpu, nodes_from_mem);
	log_record(PBSEVENT_ERROR, PBS_EVENTCLASS_SERVER, id, log_buffer);
	DBPRT(("%s: %s\n", id, log_buffer));
    }

    /*
     * Note: The assigned cpus and memory need not be exactly the same
     * node equivalency.
     */
    if ((cpu_default != UNSPECIFIED) && (mem_default != UNSPECIFIED))
	queue->nodes_default = NODES_REQD(cpu_default, mem_default);
    if ((cpu_max != UNSPECIFIED) && (mem_max != UNSPECIFIED))
	queue->nodes_max     = NODES_REQD(cpu_max, mem_max);
    if ((cpu_min != UNSPECIFIED) && (mem_min != UNSPECIFIED))
	queue->nodes_min     = NODES_REQD(cpu_min, mem_min);
    if ((cpu_assn != UNSPECIFIED) && (mem_assn != UNSPECIFIED))
	queue->nodes_assn    = NODES_REQD(cpu_assn, mem_assn);

    /*
     * Move any jobs on this queue from the global list onto the queue's
     * list.  Keep track of when the longest-running job will end, and set
     * the 'empty_by' field to that value.  Maintain the ordering as it was
     * in "schd_AllJobs".
     */

    if (schd_AllJobs)
	moved = queue_claim_jobs(queue, &schd_AllJobs);
    
    if (moved < 0) {
	sprintf(log_buffer, "%s: WARNING! Queue '%s' failed to claim jobs.",
	    id, queue->qname);
	log_record(PBSEVENT_ERROR, PBS_EVENTCLASS_SERVER, id, log_buffer);
	DBPRT(("%s: %s\n", id, log_buffer));
    }

    if (queue->nodes_assn == UNSPECIFIED)
	queue->nodes_assn = 0;
    if (queue->running    == UNSPECIFIED)
	queue->running    = 0;

    /*
     * Find out if the queue is idle, and if it was not before, set the idle
     * time to now.  If there are running jobs, the queue is not idle at the
     * start of this iteration - set idle_since to 0.
     */
    if (queue->running) {
	queue->idle_since = 0;
    } else {
	if (queue->idle_since == 0)
	    queue->idle_since = schd_TimeNow;
    }

    /*
     * Get the resources for this queue from the resource monitor (if
     * available).  If the resmom is not accessible, disable the queue.
     * If the resources were received okay, compute the available node
     * masks from the resources and jobs.
     * Don't bother with resources for the special or submit queues.
     */
    if ((strcmp(queue->qname, schd_SubmitQueue->queue->qname) != 0) ||
	((schd_SpecialQueue != NULL) &&
          (!strcmp(queue->qname, schd_SpecialQueue->queue->qname))) ) {

	queue->rsrcs = schd_get_resources(queue->exechost);
	if (queue->rsrcs != NULL) {
	    /* Account for this queue's resources. */
	    queue->rsrcs->nodes_alloc += queue->nodes_assn;
	    queue->rsrcs->njobs       += queue->running;

	    /*
	     * If the HPM counters do not appear to be in use on this host, 
	     * check for jobs on the queue that are using hpm.  If so, set
	     * the 'HPM_IN_USE' flag on the resources.  This will prevent the
	     * HPM counters from being released to global mode at the end 
	     * of the scheduling run (c.f. cleanup.c).
	     * The 'HPM_IN_USE' flag will also be asserted if a job is run 
	     * that uses the HPM counters.
	     */
	    if (schd_MANAGE_HPM && 
		!(queue->rsrcs->flags & RSRCS_FLAGS_HPM_IN_USE))
	    {
		if (schd_hpm_job_count(queue->jobs))
		    queue->rsrcs->flags |= RSRCS_FLAGS_HPM_IN_USE;
	    }

#ifdef NODEMASK
	    /* And find the nodemasks for the queue and resources. */
	    find_nodemasks(queue, queue->rsrcs);
#endif	/* NODEMASK */