static void
_print_config(void)
{
int days, hours, mins, secs;
char name[128];
gethostname_short(name, sizeof(name));
printf("NodeName=%s ", name);
get_cpuinfo(&conf->actual_cpus,
&conf->actual_boards,
&conf->actual_sockets,
&conf->actual_cores,
&conf->actual_threads,
&conf->block_map_size,
&conf->block_map, &conf->block_map_inv);
printf("CPUs=%u Boards=%u Sockets=%u CoresPerSocket=%u "
"ThreadsPerCore=%u ",
conf->actual_cpus, conf->actual_boards, conf->actual_sockets,
conf->actual_cores, conf->actual_threads);
get_memory(&conf->real_memory_size);
get_tmp_disk(&conf->tmp_disk_space, "/tmp");
printf("RealMemory=%u TmpDisk=%u\n",
conf->real_memory_size, conf->tmp_disk_space);
get_up_time(&conf->up_time);
secs = conf->up_time % 60;
mins = (conf->up_time / 60) % 60;
hours = (conf->up_time / 3600) % 24;
days = (conf->up_time / 86400);
printf("UpTime=%u-%2.2u:%2.2u:%2.2u\n", days, hours, mins, secs);
}
void TimeOutTimer::OnTimer()
{
int timeout;
double last;
double curr = get_up_time();
MutexLock lock(mutex_);
map<uint64_t, weak_ptr<TermSession> >::iterator
it = termsession_list_.begin();
for (/* without initialize */; it != termsession_list_.end(); /* without condtion */) {
shared_ptr<TermSession> sp_it(it->second.lock());
if (sp_it) {
{
last = get_up_time();
timeout = kDefaultTimeout;
MutexLock lock(sp_it->termconn_mtx_);
if (sp_it->term_conn_) {
last = sp_it->term_conn_->last_heartbeat_time_;
timeout = sp_it->term_conn_->timeout_;
}
}
if ((int)(curr - last) > timeout) {
//超时处理...添加到工作队列里面？
logger_.Warn("[超时计时器]删除超时的终端会话，会话id: 0x"SFMT64X"CAId: "SFMT64U, sp_it->Id(), sp_it->CAId());
CASession *ca_session = sp_it->ca_session_;
if (psm_ctx_->busi_pool_->DelTermSession(sp_it) > 0) {
psm_ctx_->term_basic_func_svr_->NotifyAllTerminalStatusPChanged(ca_session, sp_it->Id());
}
MutexLock lock(sp_it->termconn_mtx_);
if (sp_it->term_conn_) {
sp_it->term_conn_->SetDirty();
}
termsession_list_.erase(it++);
} else {
it++;
}
} else {
//删除无效会话。
termsession_list_.erase(it++);
}
}
logger_.Trace("[超时计时器] 循环遍历处理时间%.3f", get_up_time() - curr);
}
/* % gcc -DDEBUG_MODULE get_mach_stat.c -I../../.. -g -DUSE_CPU_SPEED */
int
main(int argc, char * argv[])
{
int error_code;
uint16_t sockets, cores, threads;
uint16_t block_map_size;
uint16_t *block_map, *block_map_inv;
struct config_record this_node;
char node_name[MAX_SLURM_NAME];
float speed;
uint16_t testnumproc = 0;
uint32_t up_time = 0;
int days, hours, mins, secs;
char* _cpuinfo_path = "/proc/cpuinfo";
if (argc > 1) {
_cpuinfo_path = argv[1];
testnumproc = 1024; /* since may not match test host */
}
debug3("%s:", _cpuinfo_path);
error_code = get_mach_name(node_name);
if (error_code != 0)
exit(1); /* The show is all over without a node name */
error_code += get_procs(&this_node.cpus);
error_code += get_cpuinfo(MAX(this_node.cpus, testnumproc),
&this_node.sockets,
&this_node.cores,
&this_node.threads,
&block_map_size,
&block_map, &block_map_inv);
xfree(block_map); /* not used here */
xfree(block_map_inv); /* not used here */
error_code += get_memory(&this_node.real_memory);
error_code += get_tmp_disk(&this_node.tmp_disk, "/tmp");
error_code += get_up_time(&up_time);
#ifdef USE_CPU_SPEED
error_code += get_speed(&speed);
#endif
debug3("");
debug3("NodeName=%s CPUs=%u Sockets=%u Cores=%u Threads=%u",
node_name, this_node.cpus,
this_node.sockets, this_node.cores, this_node.threads);
debug3("\tRealMemory=%u TmpDisk=%u Speed=%f",
this_node.real_memory, this_node.tmp_disk, speed);
secs = up_time % 60;
mins = (up_time / 60) % 60;
hours = (up_time / 3600) % 24;
days = (up_time / 86400);
debug3("\tUpTime=%u=%u-%2.2u:%2.2u:%2.2u",
up_time, days, hours, mins, secs);
if (error_code != 0)
debug3("get_mach_stat error_code=%d encountered", error_code);
exit (error_code);
}
weak_ptr<TermSession> BusinessPool::GenTermSession(PtLoginRequest *msg, TermConnection *conn, LoginError *error)
{
if (!started_) {
logger_.Warn("PSM Business Pool have stoped, but generate terminal session");
*error = LoginWorkPoolNotStarted;
return weak_ptr<TermSession>();
}
shared_ptr<TermSession> sp_ts(new TermSession(logger_, msg, conn, error));
if (!sp_ts->valid()) {
logger_.Warn("PSM Business Pool generate terminal session user cert failure");
return weak_ptr<TermSession>();
}
// valid ok.
// generate terminal session id.
uint64_t ts_id = genTermSessionId(get_up_time(), global_cnt_++, psm_ctx_->ip_addr(), sp_ts->CAId());
logger_.Trace("PSM Business Pool generate terminal session id: 0x" SFMT64X, ts_id);
sp_ts->Id(ts_id);
CASessionMgr *cs_mgr = pool_relation_ca_session_mgr_[getIdByTermSessionId(sp_ts->Id())];
CASession *cs = cs_mgr->FindCASessionById(sp_ts->CAId());
if (cs == NULL) {
cs = cs_mgr->Create(sp_ts->CAId());
cs_mgr->Attach(cs);
}
sp_ts->ca_session_ = cs;
cs->Add(sp_ts);
cs_mgr->AttachTermSessionInfo(sp_ts->Id(), cs);
//设置Session描述符
{
ByteStream agreekey;
for ( int i=0; i<8; i++ ) {
agreekey.Add((uint8_t*)"\x0F\x0E\x0D\x0C\x0B\x0A\x09\x08\x07\x06\x05\x04\x03\x02\x01\x00",16);
}
sp_ts->session_info_desc_.agreement_key_ = agreekey;
sp_ts->session_info_desc_.session_id_ = sp_ts->Id();
sp_ts->session_info_desc_.timeout_ = 15;
sp_ts->session_info_desc_.interval_time_ = 10;
sp_ts->session_info_desc_.valid_ = true;
}
//设置终端描述符
sp_ts->terminal_info_desc_.session_id_ = sp_ts->Id();
*error = LoginOK;
return sp_ts;
}
/*
* Read the slurm configuration file (slurm.conf) and substitute some
* values into the slurmd configuration in preference of the defaults.
*/
static void
_read_config(void)
{
char *path_pubkey = NULL;
slurm_ctl_conf_t *cf = NULL;
uint16_t tmp16 = 0;
#ifndef HAVE_FRONT_END
bool cr_flag = false, gang_flag = false;
#endif
cf = slurm_conf_lock();
slurm_mutex_lock(&conf->config_mutex);
if (conf->conffile == NULL)
conf->conffile = xstrdup(cf->slurm_conf);
conf->slurm_user_id = cf->slurm_user_id;
conf->cr_type = cf->select_type_param;
path_pubkey = xstrdup(cf->job_credential_public_certificate);
if (!conf->logfile)
conf->logfile = xstrdup(cf->slurmd_logfile);
#ifndef HAVE_FRONT_END
if (!strcmp(cf->select_type, "select/cons_res"))
cr_flag = true;
if (cf->preempt_mode & PREEMPT_MODE_GANG)
gang_flag = true;
#endif
slurm_conf_unlock();
/* node_name may already be set from a command line parameter */
if (conf->node_name == NULL)
conf->node_name = slurm_conf_get_nodename(conf->hostname);
/* if we didn't match the form of the hostname already
* stored in conf->hostname, check to see if we match any
* valid aliases */
if (conf->node_name == NULL)
conf->node_name = slurm_conf_get_aliased_nodename();
if (conf->node_name == NULL)
conf->node_name = slurm_conf_get_nodename("localhost");
if (conf->node_name == NULL)
fatal("Unable to determine this slurmd's NodeName");
_massage_pathname(&conf->logfile);
/* set node_addr if relevant */
if ((conf->node_addr == NULL) &&
(conf->node_addr = slurm_conf_get_nodeaddr(conf->hostname)) &&
(strcmp(conf->node_addr, conf->hostname) == 0)) {
xfree(conf->node_addr); /* Sets to NULL */
}
conf->port = slurm_conf_get_port(conf->node_name);
slurm_conf_get_cpus_bsct(conf->node_name,
&conf->conf_cpus, &conf->conf_boards,
&conf->conf_sockets, &conf->conf_cores,
&conf->conf_threads);
/* store hardware properties in slurmd_config */
xfree(conf->block_map);
xfree(conf->block_map_inv);
_update_logging();
_update_nice();
get_cpuinfo(&conf->actual_cpus,
&conf->actual_boards,
&conf->actual_sockets,
&conf->actual_cores,
&conf->actual_threads,
&conf->block_map_size,
&conf->block_map, &conf->block_map_inv);
#ifdef HAVE_FRONT_END
/*
* When running with multiple frontends, the slurmd S:C:T values are not
* relevant, hence ignored by both _register_front_ends (sets all to 1)
* and validate_nodes_via_front_end (uses slurm.conf values).
* Report actual hardware configuration, irrespective of FastSchedule.
*/
conf->cpus = conf->actual_cpus;
conf->boards = conf->actual_boards;
conf->sockets = conf->actual_sockets;
conf->cores = conf->actual_cores;
conf->threads = conf->actual_threads;
#else
/* If the actual resources on a node differ than what is in
* the configuration file and we are using
* cons_res or gang scheduling we have to use what is in the
* configuration file because the slurmctld creates bitmaps
* for scheduling before these nodes check in.
*/
if (((cf->fast_schedule == 0) && !cr_flag && !gang_flag) ||
((cf->fast_schedule == 1) &&
(conf->actual_cpus < conf->conf_cpus))) {
conf->cpus = conf->actual_cpus;
conf->boards = conf->actual_boards;
conf->sockets = conf->actual_sockets;
conf->cores = conf->actual_cores;
conf->threads = conf->actual_threads;
} else {
conf->cpus = conf->conf_cpus;
conf->boards = conf->conf_boards;
conf->sockets = conf->conf_sockets;
conf->cores = conf->conf_cores;
conf->threads = conf->conf_threads;
}
if ((conf->cpus != conf->actual_cpus) ||
(conf->sockets != conf->actual_sockets) ||
(conf->cores != conf->actual_cores) ||
(conf->threads != conf->actual_threads)) {
if (cf->fast_schedule) {
info("Node configuration differs from hardware: "
"CPUs=%u:%u(hw) Boards=%u:%u(hw) "
"SocketsPerBoard=%u:%u(hw) CoresPerSocket=%u:%u(hw) "
"ThreadsPerCore=%u:%u(hw)",
conf->cpus, conf->actual_cpus,
conf->boards, conf->actual_boards,
conf->sockets, conf->actual_sockets,
conf->cores, conf->actual_cores,
conf->threads, conf->actual_threads);
} else if ((cf->fast_schedule == 0) && (cr_flag || gang_flag)) {
error("You are using cons_res or gang scheduling with "
"Fastschedule=0 and node configuration differs "
"from hardware. The node configuration used "
"will be what is in the slurm.conf because of "
"the bitmaps the slurmctld must create before "
"the slurmd registers.\n"
" CPUs=%u:%u(hw) Boards=%u:%u(hw) "
"SocketsPerBoard=%u:%u(hw) CoresPerSocket=%u:%u(hw) "
"ThreadsPerCore=%u:%u(hw)",
conf->cpus, conf->actual_cpus,
conf->boards, conf->actual_boards,
conf->sockets, conf->actual_sockets,
conf->cores, conf->actual_cores,
conf->threads, conf->actual_threads);
}
}
#endif
get_memory(&conf->real_memory_size);
get_up_time(&conf->up_time);
cf = slurm_conf_lock();
get_tmp_disk(&conf->tmp_disk_space, cf->tmp_fs);
_free_and_set(&conf->epilog, xstrdup(cf->epilog));
_free_and_set(&conf->prolog, xstrdup(cf->prolog));
_free_and_set(&conf->tmpfs, xstrdup(cf->tmp_fs));
_free_and_set(&conf->health_check_program,
xstrdup(cf->health_check_program));
_free_and_set(&conf->spooldir, xstrdup(cf->slurmd_spooldir));
_massage_pathname(&conf->spooldir);
_free_and_set(&conf->pidfile, xstrdup(cf->slurmd_pidfile));
_massage_pathname(&conf->pidfile);
_free_and_set(&conf->select_type, xstrdup(cf->select_type));
_free_and_set(&conf->task_prolog, xstrdup(cf->task_prolog));
_free_and_set(&conf->task_epilog, xstrdup(cf->task_epilog));
_free_and_set(&conf->pubkey, path_pubkey);
conf->debug_flags = cf->debug_flags;
conf->propagate_prio = cf->propagate_prio_process;
_free_and_set(&conf->job_acct_gather_freq,
xstrdup(cf->job_acct_gather_freq));
conf->acct_freq_task = (uint16_t)NO_VAL;
tmp16 = acct_gather_parse_freq(PROFILE_TASK,
conf->job_acct_gather_freq);
if (tmp16 != -1)
conf->acct_freq_task = tmp16;
_free_and_set(&conf->acct_gather_energy_type,
xstrdup(cf->acct_gather_energy_type));
_free_and_set(&conf->acct_gather_filesystem_type,
xstrdup(cf->acct_gather_filesystem_type));
_free_and_set(&conf->acct_gather_infiniband_type,
xstrdup(cf->acct_gather_infiniband_type));
_free_and_set(&conf->acct_gather_profile_type,
xstrdup(cf->acct_gather_profile_type));
_free_and_set(&conf->job_acct_gather_type,
xstrdup(cf->job_acct_gather_type));
if ( (conf->node_name == NULL) ||
(conf->node_name[0] == '\0') )
fatal("Node name lookup failure");
if (cf->control_addr == NULL)
fatal("Unable to establish controller machine");
if (cf->slurmctld_port == 0)
fatal("Unable to establish controller port");
conf->slurmd_timeout = cf->slurmd_timeout;
conf->use_pam = cf->use_pam;
conf->task_plugin_param = cf->task_plugin_param;
slurm_mutex_unlock(&conf->config_mutex);
slurm_conf_unlock();
}
static void
_fill_registration_msg(slurm_node_registration_status_msg_t *msg)
{
List steps;
ListIterator i;
step_loc_t *stepd;
int n;
char *arch, *os;
struct utsname buf;
static bool first_msg = true;
static time_t slurmd_start_time = 0;
Buf gres_info;
msg->node_name = xstrdup (conf->node_name);
msg->cpus = conf->cpus;
msg->boards = conf->boards;
msg->sockets = conf->sockets;
msg->cores = conf->cores;
msg->threads = conf->threads;
msg->real_memory = conf->real_memory_size;
msg->tmp_disk = conf->tmp_disk_space;
msg->hash_val = slurm_get_hash_val();
get_cpu_load(&msg->cpu_load);
gres_info = init_buf(1024);
if (gres_plugin_node_config_pack(gres_info) != SLURM_SUCCESS)
error("error packing gres configuration");
else
msg->gres_info = gres_info;
get_up_time(&conf->up_time);
msg->up_time = conf->up_time;
if (slurmd_start_time == 0)
slurmd_start_time = time(NULL);
msg->slurmd_start_time = slurmd_start_time;
if (first_msg) {
first_msg = false;
info("CPUs=%u Boards=%u Sockets=%u Cores=%u Threads=%u "
"Memory=%u TmpDisk=%u Uptime=%u",
msg->cpus, msg->boards, msg->sockets, msg->cores,
msg->threads, msg->real_memory, msg->tmp_disk,
msg->up_time);
} else {
debug3("CPUs=%u Boards=%u Sockets=%u Cores=%u Threads=%u "
"Memory=%u TmpDisk=%u Uptime=%u",
msg->cpus, msg->boards, msg->sockets, msg->cores,
msg->threads, msg->real_memory, msg->tmp_disk,
msg->up_time);
}
uname(&buf);
if ((arch = getenv("SLURM_ARCH")))
msg->arch = xstrdup(arch);
else
msg->arch = xstrdup(buf.machine);
if ((os = getenv("SLURM_OS")))
msg->os = xstrdup(os);
else
msg->os = xstrdup(buf.sysname);
if (msg->startup) {
if (switch_g_alloc_node_info(&msg->switch_nodeinfo))
error("switch_g_alloc_node_info: %m");
if (switch_g_build_node_info(msg->switch_nodeinfo))
error("switch_g_build_node_info: %m");
}
steps = stepd_available(conf->spooldir, conf->node_name);
msg->job_count = list_count(steps);
msg->job_id = xmalloc(msg->job_count * sizeof(*msg->job_id));
/* Note: Running batch jobs will have step_id == NO_VAL */
msg->step_id = xmalloc(msg->job_count * sizeof(*msg->step_id));
i = list_iterator_create(steps);
n = 0;
while ((stepd = list_next(i))) {
int fd;
fd = stepd_connect(stepd->directory, stepd->nodename,
stepd->jobid, stepd->stepid);
if (fd == -1) {
--(msg->job_count);
continue;
}
if (stepd_state(fd) == SLURMSTEPD_NOT_RUNNING) {
debug("stale domain socket for stepd %u.%u ",
stepd->jobid, stepd->stepid);
--(msg->job_count);
close(fd);
continue;
}
close(fd);
if (stepd->stepid == NO_VAL)
debug("found apparently running job %u", stepd->jobid);
else
debug("found apparently running step %u.%u",
stepd->jobid, stepd->stepid);
msg->job_id[n] = stepd->jobid;
msg->step_id[n] = stepd->stepid;
n++;
}
list_iterator_destroy(i);
list_destroy(steps);
if (!msg->energy)
msg->energy = acct_gather_energy_alloc();
acct_gather_energy_g_get_data(ENERGY_DATA_STRUCT, msg->energy);
msg->timestamp = time(NULL);
return;
}
int SMSvcApplyWork::SendResponed( unsigned int ret_code )
{
// 参数错误，直接给终端应答
PbSvcApplyResponse svcapply_response(ret_code, 0);
if ( this->pkg_->sequence_no_desc_.valid_ ) svcapply_response.Add(this->pkg_->sequence_no_desc_);
if ( this->pkg_->test_data_desc_.valid_ )
{
PB_TestDataDescriptor test_data_desc(this->pkg_->test_data_desc_.request_str_, to_string("%.3lf",get_up_time()));
svcapply_response.Add(test_data_desc);
}
ByteStream response_pkg = svcapply_response.Serialize();
if ( this->conn_->Write(response_pkg.GetBuffer(), response_pkg.Size()) )
{
//TODO:暂时默认发送成功
return 0;
}
return -1;
}
