/****** test_sge_calendar/main() ***********************************************
* NAME
* main() -- calendar test
*
* SYNOPSIS
* int main(int argc, char* argv[])
*
* FUNCTION
* calendar test
*
* INPUTS
* int argc - nr. of args
* char* argv[] - args
*
* RESULT
* int - nr of failed tests
*
*******************************************************************************/
int main(int argc, char* argv[])
{
int test_counter = 0;
int failed = 0;
prof_mt_init();
lInit(nmv);
printf("==> category test <==\n");
printf("---------------------\n");
while (tests[test_counter].test_nr != -1) {
if (test(&(tests[test_counter]),
result_category[test_counter],
test_counter) != 0) {
failed++;
}
test_counter++;
}
if (failed == 0) {
printf("\n==> All tests are okay <==\n");
}
else {
printf("\n==> %d/%d test(s) failed <==\n", failed, test_counter);
}
return failed;
}
int main(int argc, char *argv[])
{
int num_objects;
int num_names;
int i;
bool uh, nuh;
if (argc != 1 && argc < 5) {
usage(argv[0]);
}
/* initialize globals */
lInit(my_nmv);
clk_tck = sysconf(_SC_CLK_TCK); /* JG: TODO: sge_sysconf? */
prof_mt_init();
/* we need random numbers */
srand(time(0));
if (argc == 1) {
num_objects = 1000;
num_names = 10;
uh = true;
nuh = true;
} else {
/* parse commandline options */
num_objects = atoi(argv[1]);
num_names = atoi(argv[2]);
uh = atoi(argv[3]) == 0 ? false : true;
nuh = atoi(argv[4]) == 0 ? false : true;
}
/* create name array */
names = sge_malloc (num_names * sizeof(const char *));
/* build random names */
for (i = 0; i < num_names; i++) {
const char *name = random_string(10);
names[i] = name;
}
/* output header */
printf(HEADER_FORMAT, "uh ", "nuh",
"create", "copy", "rau", "inu", "curo", "cnuro",
"dru", "(objs)", "dinu", "(objs)", "mem(kB)");
/* do tests */
do_test(uh, nuh, num_objects, num_names);
/* free names */
for (i = 0; i < num_names; i++) {
sge_free(&(names[i]));
}
return EXIT_SUCCESS;
}
int main(int argc, char *argv[])
{
lInit(nmv);
prof_mt_init();
printf("lWhat() ... %s\n",
test_lWhat() ? "Ok" : "Failed");
printf("lCountWhat(...) ... %s\n",
test_lCountWhat() ? "Ok" : "Failed");
printf("lReduceDescr(...) ... %s\n",
test_lReduceDescr() ? "Ok" : "Failed");
printf("lCopyWhat() ... %s\n",
test_lCopyWhat() ? "Ok" : "Failed");
printf("lIntVector2What() ... %s\n",
test_lIntVector2What() ? "Ok" : "Failed");
printf("lSelect() ... %s\n",
test_lWhat_lSelect() ? "Ok" : "Failed");
return EXIT_SUCCESS;
}
int
main(int argc, char **argv)
{
lList *opts_cmdline = NULL;
lList *opts_defaults = NULL;
lList *opts_scriptfile = NULL;
lList *opts_all = NULL;
lListElem *job = NULL;
lList *alp = NULL;
lListElem *ep;
int exit_status = 0;
int just_verify;
int tmp_ret;
int wait_for_job = 0, is_immediate = 0;
dstring session_key_out = DSTRING_INIT;
dstring diag = DSTRING_INIT;
dstring jobid = DSTRING_INIT;
u_long32 start, end, step;
u_long32 num_tasks;
int count, stat;
char *jobid_string = NULL;
bool has_terse;
drmaa_attr_values_t *jobids = NULL;
u_long32 prog_number = 0;
u_long32 myuid = 0;
const char *sge_root = NULL;
const char *cell_root = NULL;
const char *username = NULL;
const char *qualified_hostname = NULL;
const char *unqualified_hostname = NULL;
const char *mastername = NULL;
DENTER_MAIN(TOP_LAYER, "qsub");
prof_mt_init();
/* Set up the program information name */
sge_setup_sig_handlers(QSUB);
DPRINTF(("Initializing JAPI\n"));
if (japi_init(NULL, NULL, NULL, QSUB, false, NULL, &diag)
!= DRMAA_ERRNO_SUCCESS) {
fprintf(stderr, "\n");
fprintf(stderr, MSG_QSUB_COULDNOTINITIALIZEENV_S,
sge_dstring_get_string(&diag));
fprintf(stderr, "\n");
DEXIT;
SGE_EXIT((void**)&ctx, 1);
}
prog_number = ctx->get_who(ctx);
myuid = ctx->get_uid(ctx);
sge_root = ctx->get_sge_root(ctx);
cell_root = ctx->get_cell_root(ctx);
username = ctx->get_username(ctx);
qualified_hostname = ctx->get_qualified_hostname(ctx);
unqualified_hostname = ctx->get_unqualified_hostname(ctx);
mastername = ctx->get_master(ctx, false);
/*
* read switches from the various defaults files
*/
opt_list_append_opts_from_default_files(prog_number, cell_root, username, &opts_defaults, &alp, environ);
tmp_ret = answer_list_print_err_warn(&alp, NULL, NULL, MSG_WARNING);
if (tmp_ret > 0) {
DEXIT;
SGE_EXIT((void**)&ctx, tmp_ret);
}
/*
* append the commandline switches to the list
*/
opt_list_append_opts_from_qsub_cmdline(prog_number, &opts_cmdline, &alp,
argv + 1, environ);
tmp_ret = answer_list_print_err_warn(&alp, NULL, "qsub: ", MSG_QSUB_WARNING_S);
if (tmp_ret > 0) {
DEXIT;
SGE_EXIT((void**)&ctx, tmp_ret);
}
/*
* show usage if -help was in commandline
*/
if (opt_list_has_X(opts_cmdline, "-help")) {
sge_usage(QSUB, stdout);
DEXIT;
SGE_EXIT((void**)&ctx, 0);
}
/*
* We will only read commandline options from scripfile if the script
* itself should not be handled as binary
*/
if (opt_list_is_X_true(opts_cmdline, "-b") ||
(!opt_list_has_X(opts_cmdline, "-b") &&
opt_list_is_X_true(opts_defaults, "-b"))) {
DPRINTF(("Skipping options from script due to -b option\n"));
} else {
opt_list_append_opts_from_script(prog_number,
&opts_scriptfile, &alp,
opts_cmdline, environ);
tmp_ret = answer_list_print_err_warn(&alp, NULL, MSG_QSUB_COULDNOTREADSCRIPT_S,
MSG_WARNING);
if (tmp_ret > 0) {
DEXIT;
SGE_EXIT((void**)&ctx, tmp_ret);
}
}
/*
* Merge all commandline options and interprete them
*/
opt_list_merge_command_lines(&opts_all, &opts_defaults,
&opts_scriptfile, &opts_cmdline);
/*
* Check if -terse is requested
*/
has_terse = opt_list_has_X(opts_all, "-terse");
/* If "-sync y" is set, wait for the job to end. */
/* Remove all -sync switches since cull_parse_job_parameter()
* doesn't know what to do with them. */
while ((ep = lGetElemStr(opts_all, SPA_switch, "-sync"))) {
if (lGetInt(ep, SPA_argval_lIntT) == TRUE) {
wait_for_job = 1;
}
lRemoveElem(opts_all, &ep);
}
if (wait_for_job) {
DPRINTF(("Wait for job end\n"));
}
alp = cull_parse_job_parameter(myuid, username, cell_root, unqualified_hostname,
qualified_hostname, opts_all, &job);
tmp_ret = answer_list_print_err_warn(&alp, NULL, "qsub: ", MSG_WARNING);
if (tmp_ret > 0) {
DEXIT;
SGE_EXIT((void**)&ctx, tmp_ret);
}
if (set_sec_cred(sge_root, mastername, job, &alp) != 0) {
answer_list_output(&alp);
DEXIT;
SGE_EXIT((void**)&ctx, 1);
}
/* Check if job is immediate */
is_immediate = (int)JOB_TYPE_IS_IMMEDIATE(lGetUlong(job, JB_type));
DPRINTF(("Job is%s immediate\n", is_immediate ? "" : " not"));
DPRINTF(("Everything ok\n"));
if (lGetUlong(job, JB_verify)) {
cull_show_job(job, 0, false);
DEXIT;
SGE_EXIT((void**)&ctx, 0);
}
if (is_immediate || wait_for_job) {
pthread_t sigt;
qsub_setup_sig_handlers();
if (pthread_create(&sigt, NULL, sig_thread, (void *)NULL) != 0) {
fprintf(stderr, "\n");
fprintf(stderr, MSG_QSUB_COULDNOTINITIALIZEENV_S,
" error preparing signal handling thread");
fprintf(stderr, "\n");
exit_status = 1;
goto Error;
}
if (japi_enable_job_wait(username, unqualified_hostname, NULL, &session_key_out, error_handler, &diag) ==
DRMAA_ERRNO_DRM_COMMUNICATION_FAILURE) {
const char *msg = sge_dstring_get_string(&diag);
fprintf(stderr, "\n");
fprintf(stderr, MSG_QSUB_COULDNOTINITIALIZEENV_S,
msg?msg:" error starting event client thread");
fprintf(stderr, "\n");
exit_status = 1;
goto Error;
}
}
job_get_submit_task_ids(job, &start, &end, &step);
num_tasks = (end - start) / step + 1;
if (num_tasks > 1) {
int error = japi_run_bulk_jobs(&jobids, &job, start, end, step, &diag);
if (error != DRMAA_ERRNO_SUCCESS) {
/* No active session here means that japi_enable_job_wait() was
* interrupted by the signal handler, in which case we just break out
* quietly. */
if (error != DRMAA_ERRNO_NO_ACTIVE_SESSION) {
fprintf(stderr, MSG_QSUB_COULDNOTRUNJOB_S,
sge_dstring_get_string(&diag));
fprintf(stderr, "\n");
}
/* BUGFIX: Issuezilla #1013
* To quickly fix this issue, I'm mapping the JAPI/DRMAA error code
* back into a GDI error code. This is the easy solution. The
* correct solution would be to address issue #859, presumably by
* having JAPI reuse the GDI error codes instead of the JAPI error
* codes. */
if (error == DRMAA_ERRNO_TRY_LATER) {
exit_status = STATUS_NOTOK_DOAGAIN;
}
else {
exit_status = 1;
}
goto Error;
}
DPRINTF(("job id is: %ld\n", jobids->it.ji.jobid));
jobid_string = get_bulk_jobid_string((long)jobids->it.ji.jobid, start, end, step);
}
else if (num_tasks == 1) {
int error = japi_run_job(&jobid, &job, &diag);
if (error != DRMAA_ERRNO_SUCCESS) {
if (error != DRMAA_ERRNO_NO_ACTIVE_SESSION) {
fprintf(stderr, MSG_QSUB_COULDNOTRUNJOB_S,
sge_dstring_get_string(&diag));
fprintf(stderr, "\n");
}
/* BUGFIX: Issuezilla #1013
* To quickly fix this issue, I'm mapping the JAPI/DRMAA error code
* back into a GDI error code. This is the easy solution. The
* correct solution would be to address issue #859, presumably by
* having JAPI reuse the GDI error codes instead of the DRMAA error
* codes. */
if (error == DRMAA_ERRNO_TRY_LATER) {
exit_status = STATUS_NOTOK_DOAGAIN;
}
else {
exit_status = 1;
}
goto Error;
}
jobid_string = strdup(sge_dstring_get_string(&jobid));
DPRINTF(("job id is: %s\n", jobid_string));
sge_dstring_free(&jobid);
}
else {
fprintf(stderr, MSG_QSUB_COULDNOTRUNJOB_S, "invalid task structure");
fprintf(stderr, "\n");
exit_status = 1;
goto Error;
}
/* only success message is printed to stdout */
just_verify = (lGetUlong(job, JB_verify_suitable_queues)==JUST_VERIFY ||
lGetUlong(job, JB_verify_suitable_queues)==POKE_VERIFY);
DPRINTF(("Just verifying job\n"));
if (!just_verify) {
const char *output = sge_dstring_get_string(&diag);
/* print the tersed output */
if (has_terse) {
printf("%s", jobid_string);
} else if (output != NULL) {
printf("%s", output);
} else {
printf(MSG_QSUB_YOURJOBHASBEENSUBMITTED_SS, jobid_string, lGetString(job, JB_job_name));
}
printf("\n");
} else {
printf(MSG_JOB_VERIFYFOUNDQ);
printf("\n");
}
if ((wait_for_job || is_immediate) && !just_verify) {
int event;
if (is_immediate) {
fprintf(stderr, "%s\n", MSG_QSUB_WAITINGFORIMMEDIATEJOBTOBESCHEDULED);
/* We only need to wait for the first task to be scheduled to be able
* to say that the job is running. */
tmp_ret = japi_wait(DRMAA_JOB_IDS_SESSION_ANY, &jobid, &stat,
DRMAA_TIMEOUT_WAIT_FOREVER, JAPI_JOB_START, &event,
NULL, &diag);
if ((tmp_ret == DRMAA_ERRNO_SUCCESS) && (event == JAPI_JOB_START)) {
fprintf(stderr, "\n");
fprintf(stderr, MSG_QSUB_YOURIMMEDIATEJOBXHASBEENSUCCESSFULLYSCHEDULED_S,
jobid_string);
fprintf(stderr, "\n");
}
/* A job finish event here means that the job was rejected. */
else if ((tmp_ret == DRMAA_ERRNO_SUCCESS) &&
(event == JAPI_JOB_FINISH)) {
fprintf(stderr, "\n%s\n", MSG_QSUB_YOURQSUBREQUESTCOULDNOTBESCHEDULEDDTRYLATER);
exit_status = 1;
goto Error;
}
else {
/* Since we told japi_wait to wait forever, we know that if it gets
* a timeout, it's because it's been interrupted to exit, in which
* case we don't complain. Same for no active session. */
if ((tmp_ret != DRMAA_ERRNO_EXIT_TIMEOUT) &&
(tmp_ret != DRMAA_ERRNO_NO_ACTIVE_SESSION)) {
fprintf(stderr, "\n");
fprintf(stderr, MSG_QSUB_COULDNOTWAITFORJOB_S,
sge_dstring_get_string(&diag));
fprintf(stderr, "\n");
}
exit_status = 1;
goto Error;
}
}
if (wait_for_job) {
/* Rather than using japi_synchronize on ALL for bulk jobs, we use
* japi_wait on ANY num_tasks times because with synchronize, we would
* have to wait for all the tasks to finish before we know if any
* finished. */
for (count = 0; count < num_tasks; count++) {
/* Since there's only one running job in the session, we can just
* wait for ANY. */
if ((tmp_ret = japi_wait(DRMAA_JOB_IDS_SESSION_ANY, &jobid, &stat,
DRMAA_TIMEOUT_WAIT_FOREVER, JAPI_JOB_FINISH, &event,
NULL, &diag)) != DRMAA_ERRNO_SUCCESS) {
if ((tmp_ret != DRMAA_ERRNO_EXIT_TIMEOUT) &&
(tmp_ret != DRMAA_ERRNO_NO_ACTIVE_SESSION)) {
fprintf(stderr, "\n");
fprintf(stderr, MSG_QSUB_COULDNOTWAITFORJOB_S, sge_dstring_get_string(&diag));
fprintf(stderr, "\n");
}
exit_status = 1;
goto Error;
}
/* report how job finished */
/* If the job is an array job, use the first non-zero exit code as
* the exit code for qsub. */
if (exit_status == 0) {
exit_status = report_exit_status(stat,
sge_dstring_get_string(&jobid));
}
/* If we've already found a non-zero exit code, just print the exit
* info for the task. */
else {
report_exit_status(stat, sge_dstring_get_string(&jobid));
}
}
}
}
Error:
FREE(jobid_string);
lFreeList(&alp);
lFreeList(&opts_all);
if ((tmp_ret = japi_exit(JAPI_EXIT_NO_FLAG, &diag)) != DRMAA_ERRNO_SUCCESS) {
if (tmp_ret != DRMAA_ERRNO_NO_ACTIVE_SESSION) {
fprintf(stderr, "\n");
fprintf(stderr, MSG_QSUB_COULDNOTFINALIZEENV_S, sge_dstring_get_string(&diag));
fprintf(stderr, "\n");
}
else {
struct timespec ts;
/* We know that if we get a DRMAA_ERRNO_NO_ACTIVE_SESSION here, it's
* because the signal handler thread called japi_exit(). We know this
* because if the call to japi_init() fails, we just exit directly.
* If the call to japi_init() succeeds, then we have an active session,
* so coming here because of an error would not result in the
* DRMAA_ERRNO_NO_ACTIVE_SESSION error. */
DPRINTF(("Sleeping for 15 seconds to wait for the exit to finish.\n"));
sge_relative_timespec(15, &ts);
sge_mutex_lock("qsub_exit_mutex", SGE_FUNC, __LINE__, &exit_mutex);
while (!exited) {
if (pthread_cond_timedwait(&exit_cv, &exit_mutex, &ts) == ETIMEDOUT) {
DPRINTF(("Exit has not finished after 15 seconds. Exiting.\n"));
break;
}
}
sge_mutex_unlock("qsub_exit_mutex", SGE_FUNC, __LINE__, &exit_mutex);
}
}
sge_prof_cleanup();
/* This is an exit() instead of an SGE_EXIT() because when the qmaster is
* supended, SGE_EXIT() hangs. */
exit(exit_status);
DEXIT;
return exit_status;
}
extern int main(int argc, char** argv)
{
struct sigaction sa;
struct timeval now;
time_t last_time = 0;
time_t shutdown_time = 0;
int i,first_message_sent = 0;
int no_output = 0;
cl_byte_t *reference = NULL;
prof_mt_init();
if (argc < 4) {
printf("syntax: debug_level vmaster_port vmaster_host [no_output]\n");
exit(1);
}
if (argc >= 5) {
if (strcmp(argv[4],"no_output") == 0) {
printf("virtual event client: no_output option set\n");
no_output = 1;
}
}
/* setup signalhandling */
memset(&sa, 0, sizeof(sa));
sa.sa_handler = sighandler_client; /* one handler for all signals */
sigemptyset(&sa.sa_mask);
sigaction(SIGINT, &sa, NULL);
sigaction(SIGTERM, &sa, NULL);
sigaction(SIGHUP, &sa, NULL);
sigaction(SIGPIPE, &sa, NULL);
gettimeofday(&now,NULL);
shutdown_time = now.tv_sec + SGE_TEST_VIRTUAL_CLIENT_SHUTDOWN_TIMEOUT;
printf("startup commlib ...\n");
cl_com_setup_commlib(CL_NO_THREAD , (cl_log_t)atoi(argv[1]), NULL);
printf("setting up handle for connect port %d\n", atoi(argv[2]) );
handle=cl_com_create_handle(NULL,CL_CT_TCP,CL_CM_CT_MESSAGE , CL_FALSE, atoi(argv[2]) , CL_TCP_DEFAULT,"virtual_event_client", 0, 1,0 );
if (handle == NULL) {
printf("could not get handle\n");
exit(1);
}
printf("local hostname is \"%s\"\n", handle->local->comp_host);
printf("local component is \"%s\"\n", handle->local->comp_name);
printf("local component id is \"%ld\"\n", handle->local->comp_id);
cl_com_get_connect_port(handle, &i);
printf("connecting to port \"%d\" on host \"%s\"\n", i, argv[3]);
printf("virtual event client is running ...\n");
while(do_shutdown == 0) {
int retval = 0;
cl_com_message_t* message = NULL;
cl_com_endpoint_t* sender = NULL;
gettimeofday(&now,NULL);
if (now.tv_sec != last_time && !no_output) {
printf("virtual event client["pid_t_fmt"] message count[sent |%d|] events[received |%d|]...\n",
getpid(), snd_messages, events_received);
last_time = now.tv_sec;
}
if (now.tv_sec > shutdown_time ) {
printf("shutting down test - timeout\n");
do_shutdown = 1;
}
retval = cl_commlib_receive_message(handle, NULL, NULL, 0, /* handle, comp_host, comp_name , comp_id, */
CL_TRUE, 0, /* syncron, response_mid */
&message, &sender );
if ( retval != CL_RETVAL_OK) {
if ( retval == CL_RETVAL_CONNECTION_NOT_FOUND ) {
#if 0
printf("opening connection to %s/%s/%d\n", argv[3], "virtual_master", 1);
#endif
/* shutdown when virtual qmaster is not running anymore */
if (events_received > 0) {
do_shutdown = 1;
printf("lost connection to %s/%s/%d\n", argv[3], "virtual_master", 1);
}
retval = cl_commlib_open_connection(handle, argv[3], "virtual_master", 1);
if (retval == CL_RETVAL_OK) {
first_message_sent = 0;
}
}
} else {
#if 0
printf("received message from %s/%s/%ld: \"%s\" (%ld bytes)\n", sender->comp_host,sender->comp_name,sender->comp_id, message->message, message->message_length);
#endif
gettimeofday(&now,NULL);
shutdown_time = now.tv_sec + SGE_TEST_VIRTUAL_CLIENT_SHUTDOWN_TIMEOUT;
events_received++;
cl_com_free_message(&message);
cl_com_free_endpoint(&sender);
}
#if 0
printf("status: %s\n",cl_get_error_text(retval));
#endif
if ( first_message_sent == 0) {
char data[6] = "event";
reference = (cl_byte_t *)data;
first_message_sent = 1;
retval = cl_commlib_send_message(handle, argv[3], "virtual_master", 1,
CL_MIH_MAT_ACK,
&reference, 6,
NULL, 0, 0 , CL_TRUE, CL_TRUE );
if (retval == CL_RETVAL_OK) {
snd_messages++;
}
#if 0
printf("sending a event hello message: %s\n",cl_get_error_text(retval) );
#endif
}
}
printf("shutdown commlib ...\n");
cl_com_cleanup_commlib();
printf("main done\n");
return 0;
}
/****** qmaster/sge_qmaster_main/main() ****************************************
* NAME
* main() -- qmaster entry point
*
* SYNOPSIS
* int main(int argc, char* argv[])
*
* FUNCTION
* Qmaster entry point.
*
* NOTE: The main thread must block all signals before any additional thread
* is created. Failure to do so will ruin signal handling!
*
* INPUTS
* int argc - number of commandline arguments
* char* argv[] - commandline arguments
*
* RESULT
* 0 - success
*
* NOTES
* We check whether 'SGE_ROOT' is set before we daemonize. Once qmaster is
* a daemon, we are no longer connected to a terminal and hence can not
* output an error message to stdout or stderr.
*
* We need to inovke 'prepare_enroll()' *before* the user id is switched via
* 'become_admin_user()'. This is because qmaster must be able to bind a so
* called reserved port (requires root privileges) if configured to do so.
*
*******************************************************************************/
int main(int argc, char* argv[])
{
int max_enroll_tries;
int ret_val;
int file_descriptor_settings_result = 0;
bool has_daemonized = false;
sge_gdi_ctx_class_t *ctx = NULL;
u_long32 start_time = sge_get_gmt();
monitoring_t monitor;
DENTER_MAIN(TOP_LAYER, "qmaster");
sge_monitor_init(&monitor, "MAIN", NONE_EXT, MT_WARNING, MT_ERROR);
prof_mt_init();
sge_get_root_dir(true, NULL, 0, true);
#ifdef __SGE_COMPILE_WITH_GETTEXT__
sge_init_language_func((gettext_func_type)gettext, (setlocale_func_type)setlocale, (bindtextdomain_func_type)bindtextdomain, (textdomain_func_type)textdomain);
sge_init_language(NULL,NULL);
#endif
/*
* qmaster doesn't support any commandline anymore,
* but we should show version string and -help option
*/
if (argc != 1) {
sigset_t sig_set;
sigfillset(&sig_set);
pthread_sigmask(SIG_SETMASK, &sig_set, NULL);
sge_qmaster_thread_init(&ctx, QMASTER, MAIN_THREAD, true);
sge_process_qmaster_cmdline(argv);
SGE_EXIT((void**)&ctx, 1);
}
/*
* daemonize qmaster
* set file descriptor limits
* and initialize libraries to be used in multi threaded environment
* also take care that finished child processed of this process become
* zombie jobs
*/
has_daemonized = sge_daemonize_qmaster();
file_descriptor_settings_result = set_file_descriptor_limit();
#if !defined(INTERIX) && !defined(CYGWIN)
init_sig_action_and_mask();
#endif
/* init qmaster threads without becomming admin user */
sge_qmaster_thread_init(&ctx, QMASTER, MAIN_THREAD, false);
ctx->set_daemonized(ctx, has_daemonized);
/* this must be done as root user to be able to bind ports < 1024 */
max_enroll_tries = 30;
while (cl_com_get_handle(prognames[QMASTER],1) == NULL) {
ctx->prepare_enroll(ctx);
max_enroll_tries--;
if (max_enroll_tries <= 0) {
/* exit after 30 seconds */
CRITICAL((SGE_EVENT, MSG_QMASTER_COMMUNICATION_ERRORS ));
SGE_EXIT((void**)&ctx, 1);
}
if (cl_com_get_handle(prognames[QMASTER],1) == NULL) {
/* sleep when prepare_enroll() failed */
sleep(1);
}
}
/*
* now the commlib up and running. Set qmaster application status function
* (commlib callback function for qping status information response
* messages (SIRM))
*/
ret_val = cl_com_set_status_func(sge_qmaster_application_status);
if (ret_val != CL_RETVAL_OK) {
ERROR((SGE_EVENT, cl_get_error_text(ret_val)));
}
/*
* now we become admin user change into the correct root directory set the
* the target for logging messages
*/
sge_become_admin_user(ctx->get_admin_user(ctx));
sge_chdir_exit(ctx->get_qmaster_spool_dir(ctx), 1);
log_state_set_log_file(ERR_FILE);
ctx->set_exit_func(ctx, sge_exit_func);
#if defined(SOLARIS)
/* Init shared SMF libs if necessary */
if (sge_smf_used() == 1 && sge_smf_init_libs() != 0) {
SGE_EXIT((void**)&ctx, 1);
}
#endif
/*
* We do increment the heartbeat manually here. This is the 'startup heartbeat'.
* The first time the hearbeat will be incremented through the heartbeat event
* handler is after about HEARTBEAT_INTERVAL seconds. The hardbeat event handler
* is setup during the initialisazion of the timer thread.
*/
inc_qmaster_heartbeat(QMASTER_HEARTBEAT_FILE, HEARTBEAT_INTERVAL, NULL);
/*
* Event master module has to be initialized already here because
* sge_setup_qmaster() might already access it although event delivery
* thread is not running.
*
* Corresponding shutdown is done in sge_event_master_terminate();
*
* EB: In my opinion the init function should called in
* sge_event_master_initialize(). Is it possible to move that call?
*/
sge_event_master_init();
sge_setup_qmaster(ctx, argv);
#ifndef USE_POLL
if (file_descriptor_settings_result == 1) {
WARNING((SGE_EVENT, MSG_QMASTER_FD_SETSIZE_LARGER_THAN_LIMIT_U, sge_u32c(FD_SETSIZE)));
WARNING((SGE_EVENT, MSG_QMASTER_FD_SETSIZE_COMPILE_MESSAGE1_U, sge_u32c(FD_SETSIZE - 20)));
WARNING((SGE_EVENT, MSG_QMASTER_FD_SETSIZE_COMPILE_MESSAGE2));
WARNING((SGE_EVENT, MSG_QMASTER_FD_SETSIZE_COMPILE_MESSAGE3));
}
#endif
/*
* Setup all threads and initialize corresponding modules.
* Order is important!
*/
sge_signaler_initialize(ctx);
sge_event_master_initialize(ctx);
sge_timer_initialize(ctx, &monitor);
sge_worker_initialize(ctx);
#if 0
sge_test_initialize(ctx);
#endif
sge_listener_initialize(ctx);
sge_scheduler_initialize(ctx, NULL);
#ifndef NO_JNI
sge_jvm_initialize(ctx, NULL);
#endif
INFO((SGE_EVENT, "qmaster startup took "sge_u32" seconds", sge_get_gmt() - start_time));
/*
* Block till signal from signal thread arrives us
*/
sge_thread_wait_for_signal();
/*
* Shutdown all threads and shutdown corresponding modules.
* Order is important!
*/
#ifndef NO_JNI
sge_jvm_terminate(ctx, NULL);
#endif
sge_scheduler_terminate(ctx, NULL);
sge_listener_terminate();
#if 0
sge_test_terminate(ctx);
#endif
sge_worker_terminate(ctx);
sge_timer_terminate();
sge_event_master_terminate();
sge_signaler_terminate();
/*
* Remaining shutdown operations
*/
sge_clean_lists();
sge_monitor_free(&monitor);
sge_shutdown((void**)&ctx, sge_qmaster_get_exit_state());
sge_prof_cleanup();
DEXIT;
return 0;
} /* main() */
extern int main(int argc, char** argv)
{
struct sigaction sa;
static int runs = 100;
cl_com_handle_t* handle = NULL;
cl_com_message_t* message = NULL;
cl_com_endpoint_t* sender = NULL;
int bytes_received = 0;
unsigned long total_bytes_sent = 0;
unsigned long total_bytes_received = 0;
unsigned long total_connections = 0;
unsigned long total_connection_sum = 0;
#if 1
char* welcome_text = "Welcome to the tcp framework module!Welcome to the tcp framework module!Welcome to the tcp framework module!\nWelcome to the tcp framework module!Welcome to the tcp framework module!Welcome to the tcp framework module!\nWelcome to the tcp framework module!Welcome to the tcp framework module!Welcome to the tcp framework module!\nWelcome to the tcp framework module!Welcome to the tcp framework module!Welcome to the tcp framework module!\nWelcome to the tcp framework module!Welcome to the tcp framework module!Welcome to the tcp framework module!\nWelcome to the tcp framework module!Welcome to the tcp framework module!Welcome to the tcp framework module!\nWelcome to the tcp framework module!Welcome to the tcp framework module!Welcome to the tcp framework module!\nWelcome to the tcp framework module!Welcome to the tcp framework module!Welcome to the tcp framework module!\nWelcome to the tcp framework module!Welcome to the tcp framework module!Welcome to the tcp framework module!\nWelcome to the tcp framework module!Welcome to the tcp framework module!Welcome to the tcp framework module!\nWelcome to the tcp framework module!Welcome to the tcp framework module!Welcome to the tcp framework module!\nWelcome to the tcp framework module!Welcome to the tcp framework module!Welcome to the tcp framework module!\nWelcome to the tcp framework module!Welcome to the tcp framework module!Welcome to the tcp framework module!\nWelcome to the tcp framework module!Welcome to the tcp framework module!Welcome to the tcp framework module!\nWelcome to the tcp framework module!Welcome to the tcp framework module!Welcome to the tcp framework module!\nWelcome to the tcp framework module!Welcome to the tcp framework module!Welcome to the tcp framework module!\nWelcome to the tcp framework module!Welcome to the tcp framework module!Welcome to the tcp framework module!\nWelcome to the tcp framework module!Welcome to the tcp framework module!Welcome to the tcp framework module!\nWelcome to the tcp framework module!Welcome to the tcp framework module!Welcome to the tcp framework module!\nWelcome to the tcp framework module!Welcome to the tcp framework module!Welcome to the tcp framework module!\n";
#else
char* welcome_text = "This message is from thread 1";
#endif
struct timeval now;
long start_time;
long appl_start_time;
long end_time;
int retval = CL_RETVAL_PARAMS;
int welcome_text_size;
int close_connection = 0;
unsigned long last_mid = 0;
int i;
cl_framework_t framework = CL_CT_TCP;
cl_xml_connection_autoclose_t autoclose;
/* setup signalhandling */
memset(&sa, 0, sizeof(sa));
sa.sa_handler = sighandler_client; /* one handler for all signals */
sigemptyset(&sa.sa_mask);
sigaction(SIGINT, &sa, NULL);
sigaction(SIGTERM, &sa, NULL);
sigaction(SIGHUP, &sa, NULL);
sigaction(SIGPIPE, &sa, NULL);
if (argc < 7) {
printf("wrong parameters, param1 = server host, param2 = port number, param3 = client id, param4=debug_level, param5=sleep time, param6=do_close, [param7=framework(TCP/SSL)]\n");
exit(1);
}
prof_mt_init();
cl_com_setup_commlib(CL_NO_THREAD, (cl_log_t)atoi(argv[4]), NULL);
if (atoi(argv[6]) != 0) {
close_connection = 1;
}
if (argv[7]) {
framework = CL_CT_UNDEFINED;
if (strcmp(argv[7], "TCP") == 0) {
framework=CL_CT_TCP;
printf("using TCP framework\n");
}
if (strcmp(argv[7], "SSL") == 0) {
framework=CL_CT_SSL;
printf("using SSL framework\n");
}
if (framework == CL_CT_UNDEFINED) {
printf("unexpected framework type\n");
exit(1);
}
}
cl_com_set_alias_file("./alias_file");
if ( framework == CL_CT_SSL) {
cl_ssl_setup_t ssl_config;
ssl_config.ssl_method = CL_SSL_v23; /* v23 method */
ssl_config.ssl_CA_cert_pem_file = getenv("SSL_CA_CERT_FILE"); /* CA certificate file */
ssl_config.ssl_CA_key_pem_file = NULL; /* private certificate file of CA (not used) */
ssl_config.ssl_cert_pem_file = getenv("SSL_CERT_FILE"); /* certificates file */
ssl_config.ssl_key_pem_file = getenv("SSL_KEY_FILE"); /* key file */
ssl_config.ssl_rand_file = getenv("SSL_RAND_FILE"); /* rand file (if RAND_status() not ok) */
ssl_config.ssl_crl_file = getenv("SSL_CRL_FILE"); /* revocation list file */
ssl_config.ssl_reconnect_file = NULL; /* file for reconnect data (not used) */
ssl_config.ssl_refresh_time = 0; /* key alive time for connections (not used) */
ssl_config.ssl_password = NULL; /* password for encrypted keyfiles (not used) */
ssl_config.ssl_verify_func = NULL; /* function callback for peer user/name check */
if (ssl_config.ssl_CA_cert_pem_file == NULL ||
ssl_config.ssl_cert_pem_file == NULL ||
ssl_config.ssl_key_pem_file == NULL ||
ssl_config.ssl_rand_file == NULL) {
printf("please set the following environment variables:\n");
printf("SSL_CA_CERT_FILE = CA certificate file\n");
printf("SSL_CERT_FILE = certificates file\n");
printf("SSL_KEY_FILE = key file\n");
printf("(optional) SSL_RAND_FILE = rand file (if RAND_status() not ok)\n");
}
cl_com_specify_ssl_configuration(&ssl_config);
}
CL_LOG_STR(CL_LOG_INFO,"connection to server on host", argv[1]);
CL_LOG_INT(CL_LOG_INFO,"using port",atoi(argv[2]));
CL_LOG_STR(CL_LOG_INFO,"component is","client");
CL_LOG_INT(CL_LOG_INFO,"id ist",atoi(argv[3]));
#define SELECT_TIMEOUT 1
#if 0
#define CREATE_SERVICE
#endif
#ifdef CREATE_SERVICE
cl_com_append_known_endpoint_from_name(argv[1], "server", 1,atoi(argv[2]),CL_CM_AC_DISABLED , 1);
handle=cl_com_create_handle(NULL,framework,CL_CM_CT_MESSAGE , CL_TRUE, 0 , CL_TCP_DEFAULT, "client", atoi(argv[3]),SELECT_TIMEOUT,0 );
if (handle == NULL) {
printf("could not get handle\n");
exit(1);
} else {
int my_port;
cl_com_get_service_port(handle,&my_port);
printf("I'm reachable at port %d!\n", my_port);
}
#else
handle=cl_com_create_handle(NULL,framework,CL_CM_CT_MESSAGE , CL_FALSE, atoi(argv[2]) , CL_TCP_DEFAULT,"client", atoi(argv[3]),SELECT_TIMEOUT,0 );
if (handle == NULL) {
printf("could not get handle\n");
exit(1);
}
#endif
cl_com_set_auto_close_mode(handle,CL_CM_AC_ENABLED );
cl_com_get_auto_close_mode(handle,&autoclose);
if (autoclose != CL_CM_AC_ENABLED ) {
printf("could not enable autoclose\n");
exit(1);
}
printf("local hostname is \"%s\"\n", handle->local->comp_host);
printf("local component is \"%s\"\n", handle->local->comp_name);
printf("local component id is \"%ld\"\n", handle->local->comp_id);
#ifdef CREATE_SERVICE
cl_com_get_known_endpoint_port_from_name(argv[1], "server", 1, &i);
printf("connecting to port \"%d\" on host \"%s\"\n", i, argv[1]);
#else
cl_com_get_connect_port(handle, &i);
printf("connecting to port \"%d\" on host \"%s\"\n", i, argv[1]);
#endif
gettimeofday(&now,NULL);
start_time = now.tv_sec;
appl_start_time = start_time;
welcome_text_size = strlen(welcome_text) + 1;
if (getenv("CL_RUNS")) {
runs = atoi(getenv("CL_RUNS"));
} else {
runs = -1; /* disable runs shutdown */
}
while(do_shutdown != 1) {
unsigned long mid = 0;
int my_sent_error = 0;
CL_LOG(CL_LOG_INFO,"main loop");
if (runs > 0) {
runs--;
}
/* printf("sending to \"%s\" ...\n", argv[1]); */
/* CL_LOG(CL_LOG_ERROR,"sending ack message ..."); */
my_sent_error = cl_commlib_send_message(handle, argv[1], "server", 1, CL_MIH_MAT_ACK, (cl_byte_t**)&welcome_text , welcome_text_size, &mid ,0,0, CL_TRUE, CL_FALSE);
if ( retval == CL_RETVAL_CONNECTION_NOT_FOUND ) {
CL_LOG(CL_LOG_ERROR,"after new connection");
}
if (last_mid >= mid || mid == 1) {
total_connections++;
total_connection_sum++;
}
last_mid = mid;
#if 1
if (my_sent_error != CL_RETVAL_OK) {
printf("cl_commlib_send_message() returned %s\n", cl_get_error_text(my_sent_error));
#ifdef CREATE_SERVICE
cl_com_get_known_endpoint_port_from_name(argv[1], "server", 1, &i);
printf("connecting to port \"%d\" on host \"%s\"\n", i, argv[1]);
#else
cl_com_get_connect_port(handle, &i);
printf("connecting to port \"%d\" on host \"%s\"\n", i, argv[1]);
#endif
/* exit(1); */
#if CL_DO_SLOW
sleep(atoi(argv[5]));
#endif
cl_commlib_trigger(handle,1);
continue;
}
#endif
#if 1
retval = CL_RETVAL_PARAMS;
while (retval != CL_RETVAL_OK ) {
while ( (retval=cl_commlib_receive_message(handle,NULL, NULL, 0, CL_FALSE, 0, &message, &sender)) == CL_RETVAL_OK) {
if (message != NULL) {
cl_com_free_endpoint(&sender);
cl_com_free_message(&message);
} else {
break;
}
}
if ( retval == CL_RETVAL_CONNECTION_NOT_FOUND ) {
CL_LOG(CL_LOG_ERROR,"connection not found (1)");
break;
}
retval = cl_commlib_check_for_ack(handle, argv[1], "server", 1, mid, CL_TRUE );
if (retval != CL_RETVAL_MESSAGE_WAIT_FOR_ACK && retval != CL_RETVAL_OK) {
printf("retval of cl_commlib_check_for_ack(%ld) is %s\n",mid,cl_get_error_text(retval));
/* exit(1); */
break;
}
if (retval == CL_RETVAL_OK) {
CL_LOG_INT(CL_LOG_INFO,"received ack for message mid", (int)mid);
} else {
cl_commlib_trigger(handle, 1);
}
if ( retval == CL_RETVAL_CONNECTION_NOT_FOUND ) {
CL_LOG(CL_LOG_ERROR,"connection not found (2)");
break;
}
/* printf("retval of cl_commlib_check_for_ack(%ld) is %s\n",mid,cl_get_error_text(retval)); */
}
if (retval == CL_RETVAL_CONNECTION_NOT_FOUND) {
cl_commlib_trigger(handle, 1);
continue;
}
#endif
total_bytes_sent = total_bytes_sent + welcome_text_size;
CL_LOG_INT(CL_LOG_INFO,"bytes sent:", welcome_text_size);
bytes_received = 0;
while (bytes_received != welcome_text_size ) {
cl_commlib_trigger(handle, 1);
CL_LOG_INT(CL_LOG_INFO,"waiting for mid .... ", (int)mid);
retval = cl_commlib_receive_message(handle,NULL, NULL, 0, CL_FALSE, mid, &message, &sender);
CL_LOG_STR(CL_LOG_INFO,"waiting for bytes ...", cl_get_error_text(retval));
if (retval == CL_RETVAL_CONNECTION_NOT_FOUND) {
#if CL_DO_SLOW
/* CL_LOG(CL_LOG_ERROR,"connection not found"); */
if (atoi(argv[5]) > 0) {
printf("sleeping...\n");
sleep(atoi(argv[5]));
}
#endif
break;
}
if (message != NULL) {
/* printf("received message from \"%s\"\n", sender->comp_host); */
CL_LOG_INT(CL_LOG_INFO,"bytes received:", (int)message->message_length);
if (strcmp((char*)message->message, welcome_text) != 0) {
printf("------------------------> message transfer error\n");
}
total_bytes_received = total_bytes_received + message->message_length;
bytes_received = bytes_received + (int)message->message_length;
cl_com_free_endpoint(&sender);
cl_com_free_message(&message);
}
while ( (retval = cl_commlib_receive_message(handle,NULL, NULL, 0, CL_FALSE, 0, &message, &sender)) == CL_RETVAL_OK) {
if (message != NULL) {
cl_com_free_endpoint(&sender);
cl_com_free_message(&message);
} else {
break;
}
}
if (retval == CL_RETVAL_CONNECTION_NOT_FOUND) {
CL_LOG(CL_LOG_ERROR,"connection not found (3)");
break;
}
#if CL_DO_SLOW
sleep(atoi(argv[5]));
#endif
if (do_shutdown == 1) {
break;
}
}
#if CL_DO_SLOW
sleep(atoi(argv[5]));
#endif
gettimeofday(&now,NULL);
end_time = now.tv_sec;
if (end_time - start_time >= 2 ) {
cl_com_connection_t* con = NULL;
cl_connection_list_elem_t* elem = NULL;
/* printf("Kbit/s sent: %.2f ", ((total_bytes_sent * 8.0)/1024.0) / (double)(end_time - start_time));
printf("Kbit/s read: %.2f ", ((total_bytes_received * 8.0)/1024.0) / (double)(end_time - start_time)); */
printf("KBit/s : %.2f ", (((total_bytes_received + total_bytes_sent) * 8.0)/1024.0) / (double)(end_time - start_time));
printf("connections/s: %.2f ", (double) total_connections / (double)(end_time - start_time));
printf("avg. connections/s: %.2f", (double) total_connection_sum / (double)(end_time - appl_start_time));
cl_raw_list_lock(handle->connection_list);
elem = cl_connection_list_get_first_elem(handle->connection_list);
if (elem != NULL) {
con = elem->connection;
if (elem->connection->local) {
printf("[for comp host, comp name, comp id: \"%s\", \"%s\", \"%ld\"] \n",con->local->comp_host, con->local->comp_name , con->local->comp_id);
}
} else {
printf(" \n");
}
cl_raw_list_unlock(handle->connection_list);
start_time = now.tv_sec;
total_bytes_sent = 0;
total_bytes_received = 0;
total_connections = 0;
fflush(stdout);
}
if (close_connection != 0) {
while (cl_commlib_shutdown_handle(handle, CL_TRUE) == CL_RETVAL_MESSAGE_IN_BUFFER) {
printf("got message\n");
message = NULL;
cl_commlib_receive_message(handle,NULL, NULL, 0, CL_FALSE, 0, &message, &sender);
if (message != NULL) {
cl_com_free_endpoint(&sender);
cl_com_free_message(&message);
} else {
printf("error shutdown handle");
exit(-1);
break;
}
}
#ifdef CREATE_SERVICE
handle=cl_com_create_handle(NULL,framework,CL_CM_CT_MESSAGE , CL_TRUE, 0 , CL_TCP_DEFAULT, "client", atoi(argv[3]),SELECT_TIMEOUT,0 );
if (handle == NULL) {
printf("could not get handle\n");
exit(-1);
} else {
int my_port;
cl_com_get_service_port(handle,&my_port);
printf("I'm reachable at port %d!\n", my_port);
}
#else
handle=cl_com_create_handle(NULL,framework,CL_CM_CT_MESSAGE , CL_FALSE, atoi(argv[2]) , CL_TCP_DEFAULT, "client", atoi(argv[3]), SELECT_TIMEOUT,0 );
if (handle == NULL) {
printf("could not get handle\n");
exit(-1);
}
#endif
cl_com_set_auto_close_mode(handle,CL_CM_AC_ENABLED );
cl_com_get_auto_close_mode(handle,&autoclose);
if (autoclose != CL_CM_AC_ENABLED ) {
printf("could not enable autoclose\n");
exit(1);
}
}
if (runs == 0) {
do_shutdown = 1;
}
}
printf("do_shutdown received\n");
fflush(stdout);
while (cl_commlib_shutdown_handle(handle, CL_TRUE) == CL_RETVAL_MESSAGE_IN_BUFFER) {
printf("got message\n");
message = NULL;
cl_commlib_receive_message(handle,NULL, NULL, 0, CL_FALSE, 0, &message, &sender);
if (message != NULL) {
cl_com_free_endpoint(&sender);
cl_com_free_message(&message);
} else {
break;
}
}
printf("cleanup commlib ...\n");
cl_com_cleanup_commlib();
printf("main done\n");
fflush(stdout);
return 0;
}
int main(int argc, char **argv) {
int ret = STATUS_OK;
lList *alp = NULL;
lList *request_list = NULL;
lList *cmdline = NULL;
lListElem *aep;
int all_jobs = 0;
int all_users = 0;
u_long32 gdi_cmd = SGE_GDI_MOD;
int tmp_ret;
int me_who;
sge_gdi_ctx_class_t *ctx = NULL;
DENTER_MAIN(TOP_LAYER, "qalter");
prof_mt_init();
/*
** get command name: qalter or qresub
*/
if (!strcmp(sge_basename(argv[0], '/'), "qresub")) {
DPRINTF(("QRESUB\n"));
me_who = QRESUB;
} else if (!strcmp(sge_basename(argv[0], '/'), "qhold")) {
DPRINTF(("QHOLD\n"));
me_who = QHOLD;
} else if (!strcmp(sge_basename(argv[0], '/'), "qrls")) {
DPRINTF(("QRLS\n"));
me_who = QRLS;
} else {
DPRINTF(("QALTER\n"));
me_who = QALTER;
}
log_state_set_log_gui(1);
sge_setup_sig_handlers(me_who);
if (sge_gdi2_setup(&ctx, me_who, MAIN_THREAD, &alp) != AE_OK) {
answer_list_output(&alp);
SGE_EXIT((void**)&ctx, 1);
}
/*
** begin to work
*/
opt_list_append_opts_from_qalter_cmdline(me_who, &cmdline, &alp, argv + 1, environ);
tmp_ret = answer_list_print_err_warn(&alp, MSG_QALTER, MSG_QALTER, MSG_QALTERWARNING);
if (tmp_ret > 0) {
SGE_EXIT((void**)&ctx, tmp_ret);
}
/* handling the case that no command line parameter was specified */
if ((me_who == QHOLD || me_who == QRLS) && lGetNumberOfElem(cmdline) == 1) {
/* -h option is set implicitly for QHOLD and QRLS */
sge_usage(me_who, stderr);
fprintf(stderr, "%s\n", MSG_PARSE_NOOPTIONARGUMENT);
SGE_EXIT((void**)&ctx, 1);
} else if ((me_who == QRESUB || me_who == QALTER) && lGetNumberOfElem(cmdline) == 0) {
/* qresub and qalter have nothing set */
sge_usage(me_who, stderr);
fprintf(stderr, "%s\n", MSG_PARSE_NOOPTIONARGUMENT);
SGE_EXIT((void**)&ctx, 1);
} else if (opt_list_has_X(cmdline, "-help")) {
/* -help was specified */
sge_usage(me_who, stdout);
SGE_EXIT((void**)&ctx, 0);
}
alp = qalter_parse_job_parameter(me_who, cmdline, &request_list, &all_jobs,
&all_users);
DPRINTF(("all_jobs = %d, all_user = %d\n", all_jobs, all_users));
if (request_list && verify) {
/*
got a request list containing one element
for each job to be modified
save jobid all fields contain the same fields
so we may use show_job() with the first job
in our list
The jobid's in our request list get printed before
show_job()
*/
cull_show_job(lFirst(request_list), FLG_QALTER, false);
sge_prof_cleanup();
SGE_EXIT((void**)&ctx, 0);
}
tmp_ret = answer_list_print_err_warn(&alp, NULL, NULL, MSG_WARNING);
if (tmp_ret > 0) {
SGE_EXIT((void**)&ctx, tmp_ret);
}
if ((me_who == QALTER) ||
(me_who == QHOLD) ||
(me_who == QRLS)
) {
DPRINTF(("QALTER\n"));
gdi_cmd = SGE_GDI_MOD;
} else if (me_who == QRESUB){
DPRINTF(("QRESUB\n"));
gdi_cmd = SGE_GDI_COPY;
} else {
printf("unknown binary name.\n");
SGE_EXIT((void**)&ctx, 1);
}
if (all_jobs)
gdi_cmd |= SGE_GDI_ALL_JOBS;
if (all_users)
gdi_cmd |= SGE_GDI_ALL_USERS;
alp = ctx->gdi(ctx, SGE_JB_LIST, gdi_cmd, &request_list, NULL, NULL);
for_each (aep, alp) {
printf("%s\n", lGetString(aep, AN_text));
if (ret == STATUS_OK) {
ret = lGetUlong(aep, AN_status);
}
}
extern int main(int argc, char** argv) {
struct sigaction sa;
int test_result = 0;
bool issue_found = false;
long i = 0;
long issue = -1;
bool do_all = false;
if (argc == 1) {
do_all = true;
} else if (argc != 2) {
usage();
exit(1);
}
if (do_all == false) {
issue = atol(argv[1]);
i=0;
while(issue_list[i] != 0) {
if (issue_list[i] == issue) {
issue_found = true;
break;
}
i++;
}
if (issue_found == false) {
printf("Issue test %ld not found!\n\n", issue);
usage();
exit(1);
}
}
prof_mt_init();
/* setup signalhandling */
memset(&sa, 0, sizeof(sa));
sa.sa_handler = sighandler_issue_tests; /* one handler for all signals */
sigemptyset(&sa.sa_mask);
sigaction(SIGINT, &sa, NULL);
sigaction(SIGTERM, &sa, NULL);
sigaction(SIGHUP, &sa, NULL);
sigaction(SIGPIPE, &sa, NULL);
printf("commlib setup ...\n");
cl_com_setup_commlib(CL_RW_THREAD, CL_LOG_OFF, NULL);
if (do_all == false) {
test_result = run_test(i);
} else {
i=0;
while (issue_list[i] != 0 && test_result == 0 && do_shutdown == 0) {
test_result = run_test(i);
i++;
}
if (do_shutdown != 0) {
test_result = 100;
}
}
printf("commlib cleanup ...\n");
cl_com_cleanup_commlib();
printf("main done\n");
return test_result;
}
/*-------------------------------------------------------------------------*/
int main(int argc, char **argv)
{
int ret;
int my_pid;
int ret_val;
int printed_points = 0;
int max_enroll_tries;
static char tmp_err_file_name[SGE_PATH_MAX];
time_t next_prof_output = 0;
int execd_exit_state = 0;
lList **master_job_list = NULL;
sge_gdi_ctx_class_t *ctx = NULL;
lList *alp = NULL;
DENTER_MAIN(TOP_LAYER, "execd");
#if defined(LINUX)
gen_procList ();
#endif
prof_mt_init();
set_thread_name(pthread_self(),"Execd Thread");
prof_set_level_name(SGE_PROF_CUSTOM1, "Execd Thread", NULL);
prof_set_level_name(SGE_PROF_CUSTOM2, "Execd Dispatch", NULL);
#ifdef __SGE_COMPILE_WITH_GETTEXT__
/* init language output for gettext() , it will use the right language */
sge_init_language_func((gettext_func_type) gettext,
(setlocale_func_type) setlocale,
(bindtextdomain_func_type) bindtextdomain,
(textdomain_func_type) textdomain);
sge_init_language(NULL,NULL);
#endif /* __SGE_COMPILE_WITH_GETTEXT__ */
/* This needs a better solution */
umask(022);
/* Initialize path for temporary logging until we chdir to spool */
my_pid = getpid();
sprintf(tmp_err_file_name,"%s."sge_U32CFormat"", TMP_ERR_FILE_EXECD, sge_u32c(my_pid));
log_state_set_log_file(tmp_err_file_name);
/* exit func for SGE_EXIT() */
sge_sig_handler_in_main_loop = 0;
sge_setup_sig_handlers(EXECD);
if (sge_setup2(&ctx, EXECD, MAIN_THREAD, &alp, false) != AE_OK) {
answer_list_output(&alp);
SGE_EXIT((void**)&ctx, 1);
}
ctx->set_exit_func(ctx, execd_exit_func);
#if defined(SOLARIS)
/* Init shared SMF libs if necessary */
if (sge_smf_used() == 1 && sge_smf_init_libs() != 0) {
SGE_EXIT((void**)&ctx, 1);
}
#endif
/* prepare daemonize */
if (!getenv("SGE_ND")) {
sge_daemonize_prepare(ctx);
}
if ((ret=sge_occupy_first_three())>=0) {
CRITICAL((SGE_EVENT, MSG_FILE_REDIRECTFD_I, ret));
SGE_EXIT((void**)&ctx, 1);
}
lInit(nmv);
/* unset XAUTHORITY if set */
if (getenv("XAUTHORITY") != NULL) {
sge_unsetenv("XAUTHORITY");
}
parse_cmdline_execd(argv);
/* exit if we can't get communication handle (bind port) */
max_enroll_tries = 30;
while (cl_com_get_handle(prognames[EXECD],1) == NULL) {
ctx->prepare_enroll(ctx);
max_enroll_tries--;
if (max_enroll_tries <= 0 || shut_me_down) {
/* exit after 30 seconds */
if (printed_points != 0) {
printf("\n");
}
CRITICAL((SGE_EVENT, MSG_COM_ERROR));
SGE_EXIT((void**)&ctx, 1);
}
if (cl_com_get_handle(prognames[EXECD],1) == NULL) {
/* sleep when prepare_enroll() failed */
sleep(1);
if (max_enroll_tries < 27) {
printf(".");
printed_points++;
fflush(stdout);
}
}
}
if (printed_points != 0) {
printf("\n");
}
/*
* now the commlib up and running. Set execd application status function
* ( commlib callback function for qping status information response
* messages (SIRM) )
*/
ret_val = cl_com_set_status_func(sge_execd_application_status);
if (ret_val != CL_RETVAL_OK) {
ERROR((SGE_EVENT, cl_get_error_text(ret_val)) );
}
/* test connection */
{
cl_com_SIRM_t* status = NULL;
ret_val = cl_commlib_get_endpoint_status(ctx->get_com_handle(ctx),
(char *)ctx->get_master(ctx, true),
(char*)prognames[QMASTER], 1, &status);
if (ret_val != CL_RETVAL_OK) {
ERROR((SGE_EVENT, cl_get_error_text(ret_val)));
ERROR((SGE_EVENT, MSG_CONF_NOCONFBG));
}
cl_com_free_sirm_message(&status);
}
/* finalize daeamonize */
if (!getenv("SGE_ND")) {
sge_daemonize_finalize(ctx);
}
/* daemonizes if qmaster is unreachable */
sge_setup_sge_execd(ctx, tmp_err_file_name);
/* are we using qidle or not */
sge_ls_qidle(mconf_get_use_qidle());
sge_ls_gnu_ls(1);
DPRINTF(("use_qidle: %d\n", mconf_get_use_qidle()));
/* test load sensor (internal or external) */
{
lList *report_list = sge_build_load_report(ctx->get_qualified_hostname(ctx), ctx->get_binary_path(ctx));
lFreeList(&report_list);
}
/* here we have to wait for qmaster registration */
while (sge_execd_register_at_qmaster(ctx, false) != 0) {
if (sge_get_com_error_flag(EXECD, SGE_COM_ACCESS_DENIED, true)) {
/* This is no error */
DPRINTF(("***** got SGE_COM_ACCESS_DENIED from qmaster *****\n"));
}
if (sge_get_com_error_flag(EXECD, SGE_COM_ENDPOINT_NOT_UNIQUE, false)) {
execd_exit_state = SGE_COM_ENDPOINT_NOT_UNIQUE;
break;
}
if (shut_me_down != 0) {
break;
}
sleep(30);
}
/*
* Terminate on SIGTERM or hard communication error
*/
if (execd_exit_state != 0 || shut_me_down != 0) {
sge_shutdown((void**)&ctx, execd_exit_state);
DRETURN(execd_exit_state);
}
/*
* We write pid file when we are connected to qmaster. Otherwise an old
* execd might overwrite our pidfile.
*/
sge_write_pid(EXECD_PID_FILE);
/*
* At this point we are sure we are the only sge_execd and we are connected
* to the current qmaster. First we have to report any reaped children
* that might exist.
*/
starting_up();
/*
* Log a warning message if execd hasn't been started by a superuser
*/
if (!sge_is_start_user_superuser()) {
WARNING((SGE_EVENT, MSG_SWITCH_USER_NOT_ROOT));
}
#ifdef COMPILE_DC
if (ptf_init()) {
CRITICAL((SGE_EVENT, MSG_EXECD_NOSTARTPTF));
SGE_EXIT((void**)&ctx, 1);
}
INFO((SGE_EVENT, MSG_EXECD_STARTPDCANDPTF));
#endif
master_job_list = object_type_get_master_list(SGE_TYPE_JOB);
*master_job_list = lCreateList("Master_Job_List", JB_Type);
job_list_read_from_disk(master_job_list, "Master_Job_List",
0, SPOOL_WITHIN_EXECD,
job_initialize_job);
/* clean up jobs hanging around (look in active_dir) */
clean_up_old_jobs(ctx, 1);
execd_trash_load_report();
sge_set_flush_lr_flag(true);
sge_sig_handler_in_main_loop = 1;
if (thread_prof_active_by_id(pthread_self())) {
prof_start(SGE_PROF_CUSTOM1, NULL);
prof_start(SGE_PROF_CUSTOM2, NULL);
prof_start(SGE_PROF_GDI_REQUEST, NULL);
} else {
prof_stop(SGE_PROF_CUSTOM1, NULL);
prof_stop(SGE_PROF_CUSTOM2, NULL);
prof_stop(SGE_PROF_GDI_REQUEST, NULL);
}
PROF_START_MEASUREMENT(SGE_PROF_CUSTOM1);
/* Start dispatching */
execd_exit_state = sge_execd_process_messages(ctx);
/*
* This code is only reached when dispatcher terminates and execd goes down.
*/
/* log if we received SIGPIPE signal */
if (sge_sig_handler_sigpipe_received) {
sge_sig_handler_sigpipe_received = 0;
INFO((SGE_EVENT, "SIGPIPE received\n"));
}
#if defined(LINUX)
free_procList();
#endif
lFreeList(master_job_list);
PROF_STOP_MEASUREMENT(SGE_PROF_CUSTOM1);
if (prof_is_active(SGE_PROF_ALL)) {
time_t now = (time_t)sge_get_gmt();
if (now > next_prof_output) {
prof_output_info(SGE_PROF_ALL, false, "profiling summary:\n");
prof_reset(SGE_PROF_ALL,NULL);
next_prof_output = now + 60;
}
}
sge_prof_cleanup();
sge_shutdown((void**)&ctx, execd_exit_state);
DRETURN(execd_exit_state);
}
