/**********************************************************************
This is the function used by the scheduler to get a complete list of
complexes for a given queue.
All templates are filled by actual values.
With this list the scheduler can decide whether a request matches this
queue.
**********************************************************************/
int queue_complexes2scheduler(lList **new_centry_list, lListElem *queue, const lList *exechost_list,
const lList *centry_list)
{
DENTER(BASIS_LAYER, "queue_complexes2scheduler");
lFreeList(new_centry_list);
*new_centry_list = get_attribute_list(host_list_locate(exechost_list, "global"),
queue ? host_list_locate(exechost_list, lGetHost(queue, QU_qhostname)) : NULL,
queue, centry_list);
DRETURN(0);
}
/****** qmaster/setup_qmaster/sge_setup_job_resend() ***************************
* NAME
* sge_setup_job_resend() -- Setup job resend events.
*
* SYNOPSIS
* void sge_setup_job_resend(void)
*
* FUNCTION
* Register a job resend event for each job or array task which does have a
* 'JTRANSFERING' status.
*
* INPUTS
* void - none
*
* RESULT
* void - none
*
* NOTES
* MT-NOTE: sge_setup_job_resend() is not MT safe
*
*******************************************************************************/
void sge_setup_job_resend(void)
{
lListElem *job = NULL;
object_description *object_base = object_type_get_object_description();
DENTER(TOP_LAYER, "sge_setup_job_resend");
job = lFirst(*object_base[SGE_TYPE_JOB].list);
while (NULL != job)
{
lListElem *task;
u_long32 job_num;
job_num = lGetUlong(job, JB_job_number);
task = lFirst(lGetList(job, JB_ja_tasks));
while (NULL != task)
{
if (lGetUlong(task, JAT_status) == JTRANSFERING)
{
lListElem *granted_queue, *qinstance, *host;
const char *qname;
u_long32 task_num, when;
te_event_t ev;
task_num = lGetUlong(task, JAT_task_number);
granted_queue = lFirst(lGetList(task, JAT_granted_destin_identifier_list));
qname = lGetString(granted_queue, JG_qname);
qinstance = cqueue_list_locate_qinstance(*object_base[SGE_TYPE_CQUEUE].list, qname);
host = host_list_locate(*object_base[SGE_TYPE_EXECHOST].list, lGetHost(qinstance, QU_qhostname));
when = lGetUlong(task, JAT_start_time);
when += MAX(load_report_interval(host), MAX_JOB_DELIVER_TIME);
ev = te_new_event((time_t)when, TYPE_JOB_RESEND_EVENT, ONE_TIME_EVENT, job_num, task_num, "job-resend_event");
te_add_event(ev);
te_free_event(&ev);
DPRINTF(("Did add job resend for "sge_u32"/"sge_u32" at %d\n", job_num, task_num, when));
}
task = lNext(task);
}
job = lNext(job);
}
DEXIT;
return;
} /* sge_setup_job_resend() */
/* provide a list of attributes containing all attributes for the given host */
int host_complexes2scheduler(
lList **new_centry_list,
lListElem *host,
lList *exechost_list,
lList *centry_list
) {
DENTER(TOP_LAYER, "host_comlexes2scheduler");
if (!host) {
DPRINTF(("!!missing host!!\n"));
}
/* build global complex and add it to result */
lFreeList(new_centry_list);
*new_centry_list = get_attribute_list(host_list_locate(exechost_list, "global"), host, NULL, centry_list);
DRETURN(0);
}
/****** sge_c_report() *******************************************************
* NAME
* sge_c_report() -- process execd load report
*
* SYNOPSIS
* void sge_c_report(char *rhost, char *commproc, int id, lList *report_list)
*
* FUNCTION
*
* INPUTS
* char *rhost
* char *commproc
* int id
* lList *report_list
*
* RESULT
* void - nothing
*
* NOTES
* MT-NOTE: sge_c_report() is MT safe
*
******************************************************************************/
void sge_c_report(sge_gdi_ctx_class_t *ctx, char *rhost, char *commproc, int id, lList *report_list, monitoring_t *monitor)
{
lListElem *hep = NULL;
u_long32 rep_type;
lListElem *report;
int ret = 0;
u_long32 this_seqno, last_seqno;
u_long32 rversion;
sge_pack_buffer pb;
bool is_pb_used = false;
bool send_tag_new_conf = false;
DENTER(TOP_LAYER, "sge_c_report");
if (lGetNumberOfElem(report_list) == 0) {
DPRINTF(("received empty report\n"));
if (rhost != NULL) {
WARNING((SGE_EVENT, MSG_QMASTER_RECEIVED_EMPTY_LOAD_REPORT_S, rhost));
} else {
WARNING((SGE_EVENT, MSG_QMASTER_RECEIVED_EMPTY_LOAD_REPORT_S, "unknown"));
}
DRETURN_VOID;
}
/* accept reports only from execd's */
if (strcmp(prognames[EXECD], commproc)) {
ERROR((SGE_EVENT, MSG_GOTSTATUSREPORTOFUNKNOWNCOMMPROC_S, commproc));
DRETURN_VOID;
}
/* do not process load reports from old execution daemons */
rversion = lGetUlong(lFirst(report_list), REP_version);
if (verify_request_version(NULL, rversion, rhost, commproc, id)) {
DRETURN_VOID;
}
this_seqno = lGetUlong(lFirst(report_list), REP_seqno);
/* need exec host for all types of reports */
if (!(hep = host_list_locate(*object_type_get_master_list(SGE_TYPE_EXECHOST), rhost))) {
ERROR((SGE_EVENT, MSG_GOTSTATUSREPORTOFUNKNOWNEXECHOST_S, rhost));
DRETURN_VOID;
}
/* prevent old reports being proceeded
frequent loggings of outdated reports can be an indication
of too high message traffic arriving at qmaster */
last_seqno = lGetUlong(hep, EH_report_seqno);
if ((this_seqno < last_seqno && (last_seqno - this_seqno) <= 9000) &&
!(last_seqno > 9990 && this_seqno < 10)) {
/* this must be an old report, log and then ignore it */
INFO((SGE_EVENT, MSG_QMASTER_RECEIVED_OLD_LOAD_REPORT_UUS,
sge_u32c(this_seqno), sge_u32c(last_seqno), rhost));
DRETURN_VOID;
}
lSetUlong(hep, EH_report_seqno, this_seqno);
/* RU: */
/* tag all reschedule_unknown list entries we hope to
hear about in that job report */
update_reschedule_unknown_list(ctx, hep);
/*
** process the reports one after the other
** usually there will be a load report
** and a configuration version report
*/
for_each(report, report_list) {
rep_type = lGetUlong(report, REP_type);
switch (rep_type) {
case NUM_REP_REPORT_LOAD:
case NUM_REP_FULL_REPORT_LOAD:
MONITOR_ELOAD(monitor);
/* Now handle execds load reports */
if (lGetUlong(hep, EH_lt_heard_from) == 0 && rep_type != NUM_REP_FULL_REPORT_LOAD) {
host_notify_about_full_load_report(ctx, hep);
} else {
if (!is_pb_used) {
is_pb_used = true;
init_packbuffer(&pb, 1024, 0);
}
sge_update_load_values(ctx, rhost, lGetList(report, REP_list));
if (mconf_get_simulate_execds()) {
lList *master_exechost_list = *object_type_get_master_list(SGE_TYPE_EXECHOST);
lListElem *shep;
lListElem *simhostElem=NULL;
for_each(shep, master_exechost_list) {
simhostElem = lGetSubStr(shep, CE_name, "load_report_host", EH_consumable_config_list);
if (simhostElem != NULL) {
const char *real_host = lGetString(simhostElem, CE_stringval);
if (real_host != NULL && sge_hostcmp(real_host, rhost) == 0) {
const char* sim_host = lGetHost(shep, EH_name);
lListElem *clp = NULL;
DPRINTF(("Copy load values of %s to simulated host %s\n",
rhost, sim_host));
for_each(clp, lGetList(report, REP_list)) {
if (strcmp(lGetHost(clp, LR_host), SGE_GLOBAL_NAME) != 0) {
lSetHost(clp, LR_host, sim_host);
}
}
sge_update_load_values(ctx, sim_host, lGetList(report, REP_list));
}
}
}
}
pack_ack(&pb, ACK_LOAD_REPORT, this_seqno, 0, NULL);
}
break;
case NUM_REP_REPORT_CONF:
MONITOR_ECONF(monitor);
if (sge_compare_configuration(hep, lGetList(report, REP_list)) != 0) {
DPRINTF(("%s: configuration on host %s is not up to date\n", SGE_FUNC, rhost));
send_tag_new_conf = true;
}
break;
case NUM_REP_REPORT_PROCESSORS:
/*
** save number of processors
*/
MONITOR_EPROC(monitor);
ret = update_license_data(ctx, hep, lGetList(report, REP_list));
if (ret) {
ERROR((SGE_EVENT, MSG_LICENCE_ERRORXUPDATINGLICENSEDATA_I, ret));
}
break;
case NUM_REP_REPORT_JOB:
MONITOR_EJOB(monitor);
if (!is_pb_used) {
is_pb_used = true;
init_packbuffer(&pb, 1024, 0);
}
process_job_report(ctx, report, hep, rhost, commproc, &pb, monitor);
break;
default:
DPRINTF(("received invalid report type %ld\n", (long) rep_type));
}
/****** qquota_output/qquota_output() ********************************************
* NAME
* qquota_output() -- qquota output function
*
* SYNOPSIS
* bool qquota_output(void *ctx, lList *host_list, lList *resource_match_list,
* lList *user_list, lList *pe_list, lList *project_list, lList
* *cqueue_list, lList **alpp, report_handler_t* report_handler)
*
* FUNCTION
* print resource quota rule and the limit
*
* INPUTS
* void *ctx - gdi handler
* lList *host_list - selected hosts
* lList *resource_match_list - selected resources
* lList *user_list - selected users
* lList *pe_list - selecte pes
* lList *project_list - selected projects
* lList *cqueue_list - selected cluster queues
* lList **alpp - answer list
* report_handler_t* report_handler - report handler for xml output
*
* RESULT
* bool - true on success
* false on error
*
* NOTES
* MT-NOTE: qquota_output() is MT safe
*
*******************************************************************************/
bool qquota_output(sge_gdi_ctx_class_t *ctx, lList *host_list, lList *resource_match_list, lList *user_list,
lList *pe_list, lList *project_list, lList *cqueue_list, lList **alpp,
report_handler_t* report_handler)
{
lList *rqs_list = NULL;
lList *centry_list = NULL;
lList *userset_list = NULL;
lList *hgroup_list = NULL;
lList *exechost_list = NULL;
lListElem* global_host = NULL;
lListElem* exec_host = NULL;
lList* printed_rules = NULL; /* Hash list of already printed resource quota rules (possible with -u user1,user2,user3...) */
bool ret = true;
int xml_ret = 0;
qquota_filter_t qquota_filter = { "*",
"*",
"*",
"*",
"*" };
dstring rule_name = DSTRING_INIT;
DENTER(TOP_LAYER, "qquota_output");
/* If no user is requested on command line we set the current user as default */
qquota_filter.user = ctx->get_username(ctx);
ret = get_all_lists(ctx, &rqs_list, ¢ry_list, &userset_list, &hgroup_list, &exechost_list, host_list, alpp);
if (ret == true) {
lListElem *rqs = NULL;
printed_rules = lCreateList("rule_hash", ST_Type);
global_host = host_list_locate(exechost_list, SGE_GLOBAL_NAME);
if (report_handler != NULL) {
xml_ret = report_handler->report_started(report_handler, alpp);
if (xml_ret != QQUOTA_SUCCESS) {
ret = false;
goto qquota_output_error;
}
}
for_each(rqs, rqs_list) {
lListElem *rule = NULL;
int rule_count = 1;
if (lGetBool(rqs, RQS_enabled) == false) {
continue;
}
for_each(rule, lGetList(rqs, RQS_rule)) {
lListElem *user_ep = lFirst(user_list);
lListElem *project_ep = lFirst(project_list);
lListElem *pe_ep = lFirst(pe_list);
lListElem *queue_ep = lFirst(cqueue_list);
lListElem *host_ep = lFirst(host_list);
do {
if (user_ep != NULL) {
qquota_filter.user = lGetString(user_ep, ST_name);
}
do {
if (project_ep != NULL) {
qquota_filter.project = lGetString(project_ep, ST_name);
}
do {
if (pe_ep != NULL) {
qquota_filter.pe = lGetString(pe_ep, ST_name);
}
do {
if (queue_ep != NULL) {
qquota_filter.queue = lGetString(queue_ep, ST_name);
}
do {
if (host_ep != NULL) {
qquota_filter.host = lGetString(host_ep, ST_name);
}
if (rqs_is_matching_rule(rule, qquota_filter.user, NULL, qquota_filter.project,
qquota_filter.pe, qquota_filter.host,
qquota_filter.queue, userset_list, hgroup_list)) {
lListElem *limit = NULL;
for_each(limit, lGetList(rule, RQR_limit)) {
const char *limit_name = lGetString(limit, RQRL_name);
lList *rue_list = lGetList(limit, RQRL_usage);
lListElem *raw_centry = centry_list_locate(centry_list, limit_name);
lListElem *rue_elem = NULL;
if (raw_centry == NULL) {
/* undefined centries can be ignored */
DPRINTF(("centry %s not defined -> IGNORING\n", limit_name));
continue;
}
if ((resource_match_list != NULL) &&
((centry_list_locate(resource_match_list, limit_name) == NULL) &&
(centry_list_locate(resource_match_list, lGetString(raw_centry, CE_shortcut)) == NULL))) {
DPRINTF(("centry %s was not requested on CLI -> IGNORING\n", limit_name));
continue;
}
if (lGetString(rule, RQR_name)) {
sge_dstring_sprintf(&rule_name, "%s/%s", lGetString(rqs, RQS_name), lGetString(rule, RQR_name));
} else {
sge_dstring_sprintf(&rule_name, "%s/%d", lGetString(rqs, RQS_name), rule_count);
}
if (lGetUlong(raw_centry, CE_consumable)) {
/* for consumables we need to walk through the utilization and search for matching values */
DPRINTF(("found centry %s - consumable\n", limit_name));
for_each(rue_elem, rue_list) {
u_long32 dominant = 0;
const char *rue_name = lGetString(rue_elem, RUE_name);
char *cp = NULL;
stringT user, project, pe, queue, host;
dstring limit_str = DSTRING_INIT;
dstring value_str = DSTRING_INIT;
qquota_filter_t qf = { NULL, NULL, NULL, NULL, NULL };
/* check user name */
cp = qquota_get_next_filter(user, rue_name);
/* usergroups have the same beginning character @ as host groups */
if (is_hgroup_name(qquota_filter.user)) {
lListElem *ugroup = NULL;
if ((ugroup = userset_list_locate(userset_list, &qquota_filter.user[1])) != NULL) {
if (sge_contained_in_access_list(user, NULL, ugroup, NULL) == 0) {
continue;
}
}
} else {
if ((strcmp(user, "-") != 0) && (strcmp(qquota_filter.user, "*") != 0)
&& (fnmatch(qquota_filter.user, user, 0) != 0)) {
continue;
}
}
/* check project */
cp = qquota_get_next_filter(project, cp);
if ((strcmp(project, "-") != 0) && (strcmp(qquota_filter.project, "*") != 0)
&& (fnmatch(qquota_filter.project, project, 0) != 0)) {
continue;
}
/* check parallel environment */
cp = qquota_get_next_filter(pe, cp);
if ((strcmp(pe, "-") != 0) && (strcmp(qquota_filter.pe, "*") != 0) &&
(fnmatch(qquota_filter.pe, pe, 0) != 0) ) {
continue;
}
/* check cluster queue */
cp = qquota_get_next_filter(queue, cp);
if ((strcmp(queue, "-") != 0) && (strcmp(qquota_filter.queue, "*") != 0) &&
(fnmatch(qquota_filter.queue, queue, 0) != 0)) {
continue;
}
/* check host name */
cp = qquota_get_next_filter(host, cp);
if (is_hgroup_name(qquota_filter.host)) {
lListElem *hgroup = NULL;
if ((hgroup = hgroup_list_locate(hgroup_list, qquota_filter.host)) != NULL) {
lList *host_list = NULL;
hgroup_find_all_references(hgroup, NULL, hgroup_list, &host_list, NULL);
if (host_list == NULL && lGetElemHost(host_list, HR_name, host) == NULL) {
lFreeList(&host_list);
continue;
}
lFreeList(&host_list);
}
} else {
if ((strcmp(host, "-") != 0) && (strcmp(qquota_filter.host, "*") != 0) &&
(fnmatch(qquota_filter.host, host, 0) != 0) ) {
continue;
}
}
if (lGetBool(limit, RQRL_dynamic)) {
exec_host = host_list_locate(exechost_list, host);
sge_dstring_sprintf(&limit_str, "%d", (int)scaled_mixed_load(lGetString(limit, RQRL_value),
global_host, exec_host, centry_list));
} else {
lSetDouble(raw_centry, CE_pj_doubleval, lGetDouble(limit, RQRL_dvalue));
sge_get_dominant_stringval(raw_centry, &dominant, &limit_str);
}
lSetDouble(raw_centry,CE_pj_doubleval, lGetDouble(rue_elem, RUE_utilized_now));
sge_get_dominant_stringval(raw_centry, &dominant, &value_str);
qf.user = user;
qf.project = project;
qf.pe = pe;
qf.queue = queue;
qf.host = host;
ret = qquota_print_out_rule(rule, rule_name, limit_name,
sge_dstring_get_string(&value_str), sge_dstring_get_string(&limit_str),
qf, raw_centry, report_handler, printed_rules, alpp);
sge_dstring_free(&limit_str);
sge_dstring_free(&value_str);
}
} else {
/* static values */
qquota_filter_t qf = { NULL, NULL, NULL, NULL, NULL };
DPRINTF(("found centry %s - static value\n", limit_name));
ret = qquota_print_out_rule(rule, rule_name, limit_name,
NULL, lGetString(limit, RQRL_value),
qf, raw_centry, report_handler, printed_rules, alpp);
}
}
}
} while ((host_ep = lNext(host_ep)));
} while ((queue_ep = lNext(queue_ep)));
} while ((pe_ep = lNext(pe_ep)));
} while ((project_ep = lNext(project_ep)));
