int main(int argc, char *argv[])
{
char * tmp_env_var = NULL;
/* init enough of opal to use a few utilities */
if (OPAL_SUCCESS != opal_init_util(&argc, &argv)) {
fprintf(stderr, "OPAL failed to initialize -- ompi-probe aborting\n");
exit(1);
}
#if OPAL_ENABLE_FT_CR == 1
/* Disable the checkpoint notification routine for this
* tool. As we will never need to checkpoint this tool.
* Note: This must happen before opal_init().
*/
opal_cr_set_enabled(false);
/* Select the none component, since we don't actually use a checkpointer */
tmp_env_var = mca_base_param_env_var("crs");
opal_setenv(tmp_env_var,
"none",
true, &environ);
free(tmp_env_var);
tmp_env_var = NULL;
/* Mark as a tool program */
tmp_env_var = mca_base_param_env_var("opal_cr_is_tool");
opal_setenv(tmp_env_var,
"1",
true, &environ);
free(tmp_env_var);
#else
tmp_env_var = NULL; /* Silence compiler warning */
#endif
/* open up and select all the frameworks - this will generate the
* profiled output
*/
MPI_Init(NULL, NULL);
/* Finalize and clean up ourselves */
MPI_Finalize();
return 0;
}
/*
* Self is a special case. The 'fname' here is the command line that the user
* wishes to execute. This function takes this command line and adds
* -mca crs_self_do_restart 1
* Which will trigger the restart callback once the program has been run.
*
* For example, The user starts their program with:
* $ my_prog arg1 arg2
*
* They checkpoint it:
* $ opal_checkpoint -mca crs self 1234
*
* They restart it:
* $ opal_restart -mca crs self my_prog arg1 arg2
*
* fname is then:
* fname = "my_prog arg1 arg2"
*
* This funciton translates that to the command:
* cmd = "my_prog arg1 arg2 -mca crs self -mca crs_self_do_restart 1"
*
* Which will cause the program "my_prog" to call their restart function
* upon opal_init time.
*
* Note: The user could bypass the opal_restart routine safely by simply calling
* $ my_prog arg1 arg2 -mca crs self -mca crs_self_do_restart 1
* However, for consistency sake, we should not encourage this as it won't work for
* all of the other checkpointers.
*/
static int opal_crs_self_restart_cmd(opal_crs_self_snapshot_t *snapshot, char **cmd)
{
char * tmp_env_var = NULL;
opal_output_verbose(10, mca_crs_self_component.super.output_handle,
"crs:self: restart_cmd(%s, ---)", snapshot->cmd_line);
tmp_env_var = mca_base_param_env_var("crs");
opal_setenv(tmp_env_var,
"self",
true, &environ);
free(tmp_env_var);
tmp_env_var = NULL;
tmp_env_var = mca_base_param_env_var("crs_self_do_restart");
opal_setenv(tmp_env_var,
"1",
true, &environ);
free(tmp_env_var);
tmp_env_var = NULL;
tmp_env_var = mca_base_param_env_var("crs_self_prefix");
opal_setenv(tmp_env_var,
mca_crs_self_component.prefix,
true, &environ);
free(tmp_env_var);
tmp_env_var = NULL;
/* Instead of adding it to the command line, we should use the environment
* to pass the values. This allow sthe OPAL application to be braindead
* WRT MCA parameters
* add_args = strdup("-mca crs self -mca crs_self_do_restart 1");
*/
asprintf(cmd, "%s", snapshot->cmd_line);
return OPAL_SUCCESS;
}
int opal_cr_refresh_environ(int prev_pid) {
int val;
char *file_name = NULL;
struct stat file_status;
if( 0 >= prev_pid ) {
prev_pid = getpid();
}
/*
* Make sure the file exists. If it doesn't then this means 2 things:
* 1) We have already executed this function, and
* 2) The file has been deleted on the previous round.
*/
asprintf(&file_name, "%s/%s-%d", opal_tmp_directory(), OPAL_CR_BASE_ENV_NAME, prev_pid);
if(0 != stat(file_name, &file_status) ){
return OPAL_SUCCESS;
}
#if OPAL_ENABLE_CRDEBUG == 1
opal_unsetenv(mca_base_param_env_var("opal_cr_enable_crdebug"), &environ);
#endif
extract_env_vars(prev_pid, file_name);
#if OPAL_ENABLE_CRDEBUG == 1
mca_base_param_reg_int_name("opal_cr", "enable_crdebug",
"Enable checkpoint/restart debugging",
false, false,
0,
&val);
MPIR_debug_with_checkpoint = OPAL_INT_TO_BOOL(val);
opal_output_verbose(10, opal_cr_output,
"opal_cr: init: C/R Debugging Enabled [%s] (refresh)\n",
(MPIR_debug_with_checkpoint ? "True": "False"));
#else
val = 0; /* Silence Compiler warning */
#endif
if( NULL != file_name ){
free(file_name);
file_name = NULL;
}
return OPAL_SUCCESS;
}
int opal_crs_blcr_prelaunch(int32_t rank,
char *base_snapshot_dir,
char **app,
char **cwd,
char ***argv,
char ***env)
{
char * tmp_env_var = NULL;
tmp_env_var = mca_base_param_env_var("opal_cr_is_tool");
opal_setenv(tmp_env_var,
"0", true, env);
free(tmp_env_var);
tmp_env_var = NULL;
return OPAL_SUCCESS;
}
int opal_crs_self_prelaunch(int32_t rank,
char *base_snapshot_dir,
char **app,
char **cwd,
char ***argv,
char ***env)
{
char * tmp_env_var = NULL;
/*
* This function should never be called by a tool
*/
if( opal_cr_is_tool ) {
return OPAL_ERR_NOT_SUPPORTED;
}
tmp_env_var = mca_base_param_env_var("opal_cr_is_tool");
opal_setenv(tmp_env_var,
"0", true, env);
free(tmp_env_var);
tmp_env_var = NULL;
return OPAL_SUCCESS;
}
int orte_daemon(int argc, char *argv[])
{
int ret = 0;
opal_cmd_line_t *cmd_line = NULL;
char *rml_uri;
int i;
opal_buffer_t *buffer;
char hostname[100];
char *tmp_env_var = NULL;
/* initialize the globals */
memset(&orted_globals, 0, sizeof(orted_globals));
/* initialize the singleton died pipe to an illegal value so we can detect it was set */
orted_globals.singleton_died_pipe = -1;
/* init the failure orted vpid to an invalid value */
orted_globals.fail = ORTE_VPID_INVALID;
/* setup to check common command line options that just report and die */
cmd_line = OBJ_NEW(opal_cmd_line_t);
if (OPAL_SUCCESS != opal_cmd_line_create(cmd_line, orte_cmd_line_opts)) {
OBJ_RELEASE(cmd_line);
exit(1);
}
mca_base_cmd_line_setup(cmd_line);
if (ORTE_SUCCESS != (ret = opal_cmd_line_parse(cmd_line, false,
argc, argv))) {
char *args = NULL;
args = opal_cmd_line_get_usage_msg(cmd_line);
fprintf(stderr, "Usage: %s [OPTION]...\n%s\n", argv[0], args);
free(args);
OBJ_RELEASE(cmd_line);
return ret;
}
/*
* Since this process can now handle MCA/GMCA parameters, make sure to
* process them.
*/
mca_base_cmd_line_process_args(cmd_line, &environ, &environ);
/* Ensure that enough of OPAL is setup for us to be able to run */
/*
* NOTE: (JJH)
* We need to allow 'mca_base_cmd_line_process_args()' to process command
* line arguments *before* calling opal_init_util() since the command
* line could contain MCA parameters that affect the way opal_init_util()
* functions. AMCA parameters are one such option normally received on the
* command line that affect the way opal_init_util() behaves.
* It is "safe" to call mca_base_cmd_line_process_args() before
* opal_init_util() since mca_base_cmd_line_process_args() does *not*
* depend upon opal_init_util() functionality.
*/
if (OPAL_SUCCESS != opal_init_util(&argc, &argv)) {
fprintf(stderr, "OPAL failed to initialize -- orted aborting\n");
exit(1);
}
/* save the environment for launch purposes. This MUST be
* done so that we can pass it to any local procs we
* spawn - otherwise, those local procs won't see any
* non-MCA envars that were set in the enviro when the
* orted was executed - e.g., by .csh
*/
orte_launch_environ = opal_argv_copy(environ);
/* purge any ess flag set in the environ when we were launched */
opal_unsetenv("OMPI_MCA_ess", &orte_launch_environ);
/* if orte_daemon_debug is set, let someone know we are alive right
* away just in case we have a problem along the way
*/
if (orted_globals.debug) {
gethostname(hostname, 100);
fprintf(stderr, "Daemon was launched on %s - beginning to initialize\n", hostname);
}
/* check for help request */
if (orted_globals.help) {
char *args = NULL;
args = opal_cmd_line_get_usage_msg(cmd_line);
orte_show_help("help-orted.txt", "orted:usage", false,
argv[0], args);
free(args);
return 1;
}
#if defined(HAVE_SETSID) && !defined(__WINDOWS__)
/* see if we were directed to separate from current session */
if (orted_globals.set_sid) {
setsid();
}
#endif /* !defined(__WINDOWS__) */
/* see if they want us to spin until they can connect a debugger to us */
i=0;
while (orted_spin_flag) {
i++;
if (1000 < i) i=0;
}
#if OPAL_ENABLE_FT_CR == 1
/* Mark as a tool program */
tmp_env_var = mca_base_param_env_var("opal_cr_is_tool");
opal_setenv(tmp_env_var,
"1",
true, &environ);
free(tmp_env_var);
#endif
tmp_env_var = NULL; /* Silence compiler warning */
/* if mapreduce set, flag it */
if (orted_globals.mapreduce) {
orte_map_reduce = true;
}
/* Set the flag telling OpenRTE that I am NOT a
* singleton, but am "infrastructure" - prevents setting
* up incorrect infrastructure that only a singleton would
* require.
*/
if (orted_globals.hnp) {
if (ORTE_SUCCESS != (ret = orte_init(&argc, &argv, ORTE_PROC_HNP))) {
ORTE_ERROR_LOG(ret);
return ret;
}
} else {
if (ORTE_SUCCESS != (ret = orte_init(&argc, &argv, ORTE_PROC_DAEMON))) {
ORTE_ERROR_LOG(ret);
return ret;
}
}
/* finalize the OPAL utils. As they are opened again from orte_init->opal_init
* we continue to have a reference count on them. So we have to finalize them twice...
*/
opal_finalize_util();
if ((int)ORTE_VPID_INVALID != orted_globals.fail) {
orted_globals.abort=false;
/* some vpid was ordered to fail. The value can be positive
* or negative, depending upon the desired method for failure,
* so need to check both here
*/
if (0 > orted_globals.fail) {
orted_globals.fail = -1*orted_globals.fail;
orted_globals.abort = true;
}
/* are we the specified vpid? */
if ((int)ORTE_PROC_MY_NAME->vpid == orted_globals.fail) {
/* if the user specified we delay, then setup a timer
* and have it kill us
*/
if (0 < orted_globals.fail_delay) {
ORTE_TIMER_EVENT(orted_globals.fail_delay, 0, shutdown_callback, ORTE_SYS_PRI);
} else {
opal_output(0, "%s is executing clean %s", ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
orted_globals.abort ? "abort" : "abnormal termination");
/* do -not- call finalize as this will send a message to the HNP
* indicating clean termination! Instead, just forcibly cleanup
* the local session_dir tree and exit
*/
orte_session_dir_cleanup(ORTE_JOBID_WILDCARD);
/* if we were ordered to abort, do so */
if (orted_globals.abort) {
abort();
}
/* otherwise, return with non-zero status */
ret = ORTE_ERROR_DEFAULT_EXIT_CODE;
goto DONE;
}
}
}
/* detach from controlling terminal
* otherwise, remain attached so output can get to us
*/
if(!orte_debug_flag &&
!orte_debug_daemons_flag &&
orted_globals.daemonize) {
opal_daemon_init(NULL);
}
/* insert our contact info into our process_info struct so we
* have it for later use and set the local daemon field to our name
*/
orte_process_info.my_daemon_uri = orte_rml.get_contact_info();
ORTE_PROC_MY_DAEMON->jobid = ORTE_PROC_MY_NAME->jobid;
ORTE_PROC_MY_DAEMON->vpid = ORTE_PROC_MY_NAME->vpid;
/* if I am also the hnp, then update that contact info field too */
if (ORTE_PROC_IS_HNP) {
orte_process_info.my_hnp_uri = orte_rml.get_contact_info();
ORTE_PROC_MY_HNP->jobid = ORTE_PROC_MY_NAME->jobid;
ORTE_PROC_MY_HNP->vpid = ORTE_PROC_MY_NAME->vpid;
}
/* setup the primary daemon command receive function */
ret = orte_rml.recv_buffer_nb(ORTE_NAME_WILDCARD, ORTE_RML_TAG_DAEMON,
ORTE_RML_PERSISTENT, orte_daemon_recv, NULL);
if (ret != ORTE_SUCCESS && ret != ORTE_ERR_NOT_IMPLEMENTED) {
ORTE_ERROR_LOG(ret);
goto DONE;
}
/* output a message indicating we are alive, our name, and our pid
* for debugging purposes
*/
if (orte_debug_daemons_flag) {
fprintf(stderr, "Daemon %s checking in as pid %ld on host %s\n",
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME), (long)orte_process_info.pid,
orte_process_info.nodename);
}
/* We actually do *not* want the orted to voluntarily yield() the
processor more than necessary. The orted already blocks when
it is doing nothing, so it doesn't use any more CPU cycles than
it should; but when it *is* doing something, we do not want it
to be unnecessarily delayed because it voluntarily yielded the
processor in the middle of its work.
For example: when a message arrives at the orted, we want the
OS to wake up the orted in a timely fashion (which most OS's
seem good about doing) and then we want the orted to process
the message as fast as possible. If the orted yields and lets
aggressive MPI applications get the processor back, it may be a
long time before the OS schedules the orted to run again
(particularly if there is no IO event to wake it up). Hence,
routed OOB messages (for example) may be significantly delayed
before being delivered to MPI processes, which can be
problematic in some scenarios (e.g., COMM_SPAWN, BTL's that
require OOB messages for wireup, etc.). */
opal_progress_set_yield_when_idle(false);
/* Change the default behavior of libevent such that we want to
continually block rather than blocking for the default timeout
and then looping around the progress engine again. There
should be nothing in the orted that cannot block in libevent
until "something" happens (i.e., there's no need to keep
cycling through progress because the only things that should
happen will happen in libevent). This is a minor optimization,
but what the heck... :-) */
opal_progress_set_event_flag(OPAL_EVLOOP_ONCE);
/* if requested, obtain and report a new process name and my uri to the indicated pipe */
if (orted_globals.uri_pipe > 0) {
orte_job_t *jdata;
orte_proc_t *proc;
orte_node_t *node;
orte_app_context_t *app;
char *tmp, *nptr, *sysinfo;
int32_t ljob;
/* setup the singleton's job */
jdata = OBJ_NEW(orte_job_t);
orte_plm_base_create_jobid(jdata);
ljob = ORTE_LOCAL_JOBID(jdata->jobid);
opal_pointer_array_set_item(orte_job_data, ljob, jdata);
/* must create a map for it (even though it has no
* info in it) so that the job info will be picked
* up in subsequent pidmaps or other daemons won't
* know how to route
*/
jdata->map = OBJ_NEW(orte_job_map_t);
/* setup an app_context for the singleton */
app = OBJ_NEW(orte_app_context_t);
app->app = strdup("singleton");
app->num_procs = 1;
opal_pointer_array_add(jdata->apps, app);
#if 0
/* run our local allocator to read the available
* allocation in case this singleton decides to
* comm_spawn other procs
*/
if (ORTE_SUCCESS != (ret = orte_ras.allocate(jdata))) {
ORTE_ERROR_LOG(ret);
/* don't quit as this would cause the singleton
* to hang!
*/
}
#endif
/* setup a proc object for the singleton - since we
* -must- be the HNP, and therefore we stored our
* node on the global node pool, and since the singleton
* -must- be on the same node as us, indicate that
*/
proc = OBJ_NEW(orte_proc_t);
proc->name.jobid = jdata->jobid;
proc->name.vpid = 0;
proc->alive = true;
proc->state = ORTE_PROC_STATE_RUNNING;
proc->app_idx = 0;
/* obviously, it is on my node */
node = (orte_node_t*)opal_pointer_array_get_item(orte_node_pool, 0);
proc->node = node;
OBJ_RETAIN(node); /* keep accounting straight */
opal_pointer_array_add(jdata->procs, proc);
jdata->num_procs = 1;
/* and obviously it is one of my local procs */
OBJ_RETAIN(proc);
opal_pointer_array_add(orte_local_children, proc);
jdata->num_local_procs = 1;
/* set the trivial */
proc->local_rank = 0;
proc->node_rank = 0;
proc->app_rank = 0;
proc->state = ORTE_PROC_STATE_RUNNING;
proc->alive = true;
proc->app_idx = 0;
proc->local_proc = true;
#if OPAL_HAVE_HWLOC
proc->bind_idx = 0;
#endif
/* the singleton will use the first three collectives
* for its modex/barriers
*/
orte_grpcomm_base.coll_id += 3;
/* need to setup a pidmap for it */
jdata->pmap = (opal_byte_object_t*)malloc(sizeof(opal_byte_object_t));
if (ORTE_SUCCESS != (ret = orte_util_encode_pidmap(jdata->pmap))) {
ORTE_ERROR_LOG(ret);
goto DONE;
}
/* if we don't yet have a daemon map, then we have to generate one
* to pass back to it
*/
if (NULL == orte_odls_globals.dmap) {
orte_odls_globals.dmap = (opal_byte_object_t*)malloc(sizeof(opal_byte_object_t));
/* construct a nodemap */
if (ORTE_SUCCESS != (ret = orte_util_encode_nodemap(orte_odls_globals.dmap))) {
ORTE_ERROR_LOG(ret);
goto DONE;
}
}
/* create a string that contains our uri + the singleton's name + sysinfo */
orte_util_convert_process_name_to_string(&nptr, &proc->name);
orte_util_convert_sysinfo_to_string(&sysinfo, orte_local_cpu_type, orte_local_cpu_model);
asprintf(&tmp, "%s[%s][%s]", orte_process_info.my_daemon_uri, nptr, sysinfo);
free(nptr);
free(sysinfo);
/* pass that info to the singleton */
#ifndef __WINDOWS__
write(orted_globals.uri_pipe, tmp, strlen(tmp)+1); /* need to add 1 to get the NULL */
#else
send(orted_globals.uri_pipe, tmp, strlen(tmp)+1, 0); /* need to add 1 to get the NULL */
#endif
/* cleanup */
free(tmp);
}
/* if we were given a pipe to monitor for singleton termination, set that up */
if (orted_globals.singleton_died_pipe > 0) {
/* register shutdown handler */
pipe_handler = (opal_event_t*)malloc(sizeof(opal_event_t));
opal_event_set(orte_event_base, pipe_handler,
orted_globals.singleton_died_pipe,
OPAL_EV_READ,
pipe_closed,
pipe_handler);
opal_event_add(pipe_handler, NULL);
}
/* If I have a parent, then save his contact info so
* any messages we send can flow thru him.
*/
mca_base_param_reg_string_name("orte", "parent_uri",
"URI for the parent if tree launch is enabled.",
true, false, NULL, &rml_uri);
if (NULL != rml_uri) {
orte_process_name_t parent;
/* set the contact info into the hash table */
if (ORTE_SUCCESS != (ret = orte_rml.set_contact_info(rml_uri))) {
ORTE_ERROR_LOG(ret);
free(rml_uri);
goto DONE;
}
ret = orte_rml_base_parse_uris(rml_uri, &parent, NULL );
if( ORTE_SUCCESS != ret ) {
ORTE_ERROR_LOG(ret);
free(rml_uri);
goto DONE;
}
free(rml_uri);
/* tell the routed module that we have a path
* back to the HNP
*/
if (ORTE_SUCCESS != (ret = orte_routed.update_route(ORTE_PROC_MY_HNP, &parent))) {
ORTE_ERROR_LOG(ret);
goto DONE;
}
/* set the lifeline to point to our parent so that we
* can handle the situation if that lifeline goes away
*/
if (ORTE_SUCCESS != (ret = orte_routed.set_lifeline(&parent))) {
ORTE_ERROR_LOG(ret);
goto DONE;
}
}
/* if we are not the HNP...the only time we will be an HNP
* is if we are launched by a singleton to provide support
* for it
*/
if (!ORTE_PROC_IS_HNP) {
/* send the information to the orted report-back point - this function
* will process the data, but also counts the number of
* orteds that reported back so the launch procedure can continue.
* We need to do this at the last possible second as the HNP
* can turn right around and begin issuing orders to us
*/
buffer = OBJ_NEW(opal_buffer_t);
/* insert our name for rollup purposes */
if (ORTE_SUCCESS != (ret = opal_dss.pack(buffer, ORTE_PROC_MY_NAME, 1, ORTE_NAME))) {
ORTE_ERROR_LOG(ret);
OBJ_RELEASE(buffer);
goto DONE;
}
/* for now, always include our contact info, even if we are using
* static ports. Eventually, this will be removed
*/
rml_uri = orte_rml.get_contact_info();
if (ORTE_SUCCESS != (ret = opal_dss.pack(buffer, &rml_uri, 1, OPAL_STRING))) {
ORTE_ERROR_LOG(ret);
OBJ_RELEASE(buffer);
goto DONE;
}
/* include our node name */
opal_dss.pack(buffer, &orte_process_info.nodename, 1, OPAL_STRING);
#if OPAL_HAVE_HWLOC
/* add the local topology */
if (NULL != opal_hwloc_topology &&
(1 == ORTE_PROC_MY_NAME->vpid || orte_hetero_nodes)) {
if (ORTE_SUCCESS != (ret = opal_dss.pack(buffer, &opal_hwloc_topology, 1, OPAL_HWLOC_TOPO))) {
ORTE_ERROR_LOG(ret);
}
}
#endif
if ((orte_static_ports || orte_use_common_port) && !orted_globals.tree_spawn) {
/* use the rollup collective to send our data to the HNP
* so we minimize the HNP bottleneck
*/
orte_grpcomm_collective_t *coll;
coll = OBJ_NEW(orte_grpcomm_collective_t);
/* get the list of contributors we need from the routed module */
orte_routed.get_routing_list(ORTE_GRPCOMM_COLL_PEERS, coll);
/* add the collective to our list */
opal_list_append(&orte_grpcomm_base.active_colls, &coll->super);
/* send the buffer to ourselves to start the collective */
if (0 > (ret = orte_rml.send_buffer_nb(ORTE_PROC_MY_NAME, buffer,
ORTE_RML_TAG_ROLLUP, 0,
rml_cbfunc, NULL))) {
ORTE_ERROR_LOG(ret);
OBJ_RELEASE(buffer);
goto DONE;
}
} else {
/* send directly to the HNP's callback */
if (0 > (ret = orte_rml.send_buffer_nb(ORTE_PROC_MY_HNP, buffer,
ORTE_RML_TAG_ORTED_CALLBACK, 0,
rml_cbfunc, NULL))) {
ORTE_ERROR_LOG(ret);
OBJ_RELEASE(buffer);
goto DONE;
}
}
}
if (orte_debug_daemons_flag) {
opal_output(0, "%s orted: up and running - waiting for commands!", ORTE_NAME_PRINT(ORTE_PROC_MY_NAME));
}
/* loop the event lib until an exit event is detected */
while (orte_event_base_active) {
opal_event_loop(orte_event_base, OPAL_EVLOOP_ONCE);
}
/* ensure all local procs are dead */
orte_odls.kill_local_procs(NULL);
DONE:
/* update the exit status, in case it wasn't done */
ORTE_UPDATE_EXIT_STATUS(orte_exit_status);
/* cleanup and leave */
orte_finalize();
if (orte_debug_flag) {
fprintf(stderr, "exiting with status %d\n", orte_exit_status);
}
exit(orte_exit_status);
}
static int ckpt_init(int argc, char *argv[]) {
int exit_status = ORTE_SUCCESS, ret;
char * tmp_env_var = NULL;
listener_started = false;
/*
* Make sure to init util before parse_args
* to ensure installdirs is setup properly
* before calling mca_base_open();
*/
if( ORTE_SUCCESS != (ret = opal_init_util()) ) {
return ret;
}
/*
* Parse Command Line Arguments
*/
if (ORTE_SUCCESS != (ret = parse_args(argc, argv))) {
return ret;
}
/* Disable the checkpoint notification routine for this
* tool. As we will never need to checkpoint this tool.
* Note: This must happen before opal_init().
*/
opal_cr_set_enabled(false);
/* Select the none component, since we don't actually use a checkpointer */
tmp_env_var = mca_base_param_env_var("crs");
opal_setenv(tmp_env_var,
"none",
true, &environ);
free(tmp_env_var);
tmp_env_var = NULL;
/***************************
* We need all of OPAL and the TOOLS portion of ORTE - this
* sets us up so we can talk to any HNP over the wire
***************************/
if (ORTE_SUCCESS != (ret = orte_init(ORTE_TOOL))) {
exit_status = ret;
goto cleanup;
}
/*
* Setup ORTE Output handle from the verbose argument
*/
if( orte_checkpoint_globals.verbose ) {
orte_checkpoint_globals.output = opal_output_open(NULL);
opal_output_set_verbosity(orte_checkpoint_globals.output, 10);
} else {
orte_checkpoint_globals.output = 0; /* Default=STDERR */
}
/*
* Start the listener
*/
if( ORTE_SUCCESS != (ret = start_listener() ) ) {
exit_status = ret;
}
cleanup:
return exit_status;
}
static int parse_args(int argc, char *argv[]) {
int i, ret, len, exit_status = ORTE_SUCCESS ;
opal_cmd_line_t cmd_line;
char **app_env = NULL, **global_env = NULL;
char * tmp_env_var = NULL;
/* Init structure */
memset(&orte_checkpoint_globals, 0, sizeof(orte_checkpoint_globals_t));
orte_checkpoint_globals.help = false;
orte_checkpoint_globals.pid = -1;
orte_checkpoint_globals.term = false;
orte_checkpoint_globals.verbose = false;
orte_checkpoint_globals.req_hnp = ORTE_JOBID_INVALID;
orte_checkpoint_globals.nowait = false;
orte_checkpoint_globals.status = false;
orte_checkpoint_globals.output = -1;
orte_checkpoint_globals.ckpt_status = ORTE_SNAPC_CKPT_STATE_NONE;
/* Parse the command line options */
opal_cmd_line_create(&cmd_line, cmd_line_opts);
mca_base_open();
mca_base_cmd_line_setup(&cmd_line);
ret = opal_cmd_line_parse(&cmd_line, true, argc, argv);
/**
* Put all of the MCA arguments in the environment
*/
mca_base_cmd_line_process_args(&cmd_line, &app_env, &global_env);
len = opal_argv_count(app_env);
for(i = 0; i < len; ++i) {
putenv(app_env[i]);
}
len = opal_argv_count(global_env);
for(i = 0; i < len; ++i) {
putenv(global_env[i]);
}
tmp_env_var = mca_base_param_env_var("opal_cr_is_tool");
opal_setenv(tmp_env_var,
"1",
true, &environ);
free(tmp_env_var);
tmp_env_var = NULL;
/**
* Now start parsing our specific arguments
*/
/* get the remaining bits */
opal_cmd_line_get_tail(&cmd_line, &argc, &argv);
#if OPAL_ENABLE_FT == 0
/* Warn and exit if not configured with Checkpoint/Restart */
{
char *args = NULL;
args = opal_cmd_line_get_usage_msg(&cmd_line);
orte_show_help("help-orte-checkpoint.txt", "usage-no-cr",
true, args);
free(args);
exit_status = ORTE_ERROR;
goto cleanup;
}
#endif
if (OPAL_SUCCESS != ret ||
orte_checkpoint_globals.help ||
(0 >= argc && ORTE_JOBID_INVALID == orte_checkpoint_globals.req_hnp)) {
char *args = NULL;
args = opal_cmd_line_get_usage_msg(&cmd_line);
orte_show_help("help-orte-checkpoint.txt", "usage", true,
args);
free(args);
exit_status = ORTE_ERROR;
goto cleanup;
}
/*
* If the user did not supply an hnp jobid, then they must
* supply the PID of MPIRUN
*/
if(0 >= argc &&
ORTE_JOBID_INVALID != orte_checkpoint_globals.req_hnp) {
exit_status = ORTE_SUCCESS;
goto cleanup;
}
orte_checkpoint_globals.pid = atoi(argv[0]);
if ( 0 >= orte_checkpoint_globals.pid ) {
orte_show_help("help-orte-checkpoint.txt", "invalid_pid", true,
orte_checkpoint_globals.pid);
exit_status = ORTE_ERROR;
goto cleanup;
}
/*
* JJH: No wait is currently not implemented or tested
*/
if(orte_checkpoint_globals.nowait) {
orte_checkpoint_globals.nowait = false;
orte_show_help("help-orte-checkpoint.txt", "not_impl",
true,
"Disconnected checkpoint");
}
if(orte_checkpoint_globals.verbose) {
orte_checkpoint_globals.status = true;
}
cleanup:
return exit_status;
}
static int parse_args(int argc, char *argv[]) {
int i, ret, len, exit_status = ORTE_SUCCESS ;
opal_cmd_line_t cmd_line;
char **app_env = NULL, **global_env = NULL;
char * tmp_env_var = NULL;
char *argv0 = NULL;
/* Init structure */
memset(&orte_migrate_globals, 0, sizeof(orte_migrate_globals_t));
orte_migrate_globals.help = false;
orte_migrate_globals.pid = -1;
orte_migrate_globals.verbose = false;
orte_migrate_globals.verbose_level = 0;
orte_migrate_globals.status = false;
orte_migrate_globals.output = -1;
orte_migrate_globals.off_nodes = NULL;
orte_migrate_globals.off_procs = NULL;
orte_migrate_globals.onto_nodes = NULL;
#if OPAL_ENABLE_FT_CR == 0
/* Warn and exit if not configured with Migrate/Restart */
{
char *str, *args = NULL;
args = opal_cmd_line_get_usage_msg(&cmd_line);
str = opal_show_help_string("help-orte-migrate.txt", "usage-no-cr",
true, args);
if (NULL != str) {
printf("%s", str);
free(str);
}
free(args);
exit_status = ORTE_ERROR;
goto cleanup;
}
#endif
/* Parse the command line options */
opal_cmd_line_create(&cmd_line, cmd_line_opts);
mca_base_open();
mca_base_cmd_line_setup(&cmd_line);
ret = opal_cmd_line_parse(&cmd_line, false, argc, argv);
if (OPAL_SUCCESS != ret) {
if (OPAL_ERR_SILENT != ret) {
fprintf(stderr, "%s: command line error (%s)\n", argv[0],
opal_strerror(ret));
}
exit_status = 1;
goto cleanup;
}
if (orte_migrate_globals.help) {
char *str, *args = NULL;
args = opal_cmd_line_get_usage_msg(&cmd_line);
str = opal_show_help_string("help-orte-migrate.txt", "usage", true,
args);
if (NULL != str) {
printf("%s", str);
free(str);
}
free(args);
/* If we show the help message, that should be all we do */
exit(0);
}
/**
* Put all of the MCA arguments in the environment
*/
mca_base_cmd_line_process_args(&cmd_line, &app_env, &global_env);
len = opal_argv_count(app_env);
for(i = 0; i < len; ++i) {
putenv(app_env[i]);
}
len = opal_argv_count(global_env);
for(i = 0; i < len; ++i) {
putenv(global_env[i]);
}
tmp_env_var = mca_base_param_env_var("opal_cr_is_tool");
opal_setenv(tmp_env_var,
"1",
true, &environ);
free(tmp_env_var);
tmp_env_var = NULL;
/**
* Now start parsing our specific arguments
*/
/* get the remaining bits */
argv0 = strdup(argv[0]);
opal_cmd_line_get_tail(&cmd_line, &argc, &argv);
if (NULL == orte_migrate_globals.off_nodes &&
NULL == orte_migrate_globals.off_procs) {
fprintf(stderr, "%s: Nothing to do\n", argv0);
fprintf(stderr, "Type '%s --help' for usage.\n", argv0);
exit_status = 1;
goto cleanup;
}
if(orte_migrate_globals.verbose_level < 0 ) {
orte_migrate_globals.verbose_level = 0;
}
if(orte_migrate_globals.verbose_level > 0) {
orte_migrate_globals.verbose = true;
}
/*
* If the user did not supply an hnp jobid, then they must
* supply the PID of MPIRUN
*/
if(0 >= argc ) {
fprintf(stderr, "%s: Nothing to do\n", argv[0]);
fprintf(stderr, "Type '%s --help' for usage.\n", argv[0]);
exit_status = ORTE_ERROR;
goto cleanup;
}
orte_migrate_globals.pid = atoi(argv[0]);
if ( 0 >= orte_migrate_globals.pid ) {
opal_show_help("help-orte-migrate.txt", "invalid_pid", true,
orte_migrate_globals.pid);
exit_status = ORTE_ERROR;
goto cleanup;
}
if(orte_migrate_globals.verbose) {
orte_migrate_globals.status = true;
}
if(orte_migrate_globals.verbose) {
pretty_print_migration();
}
cleanup:
if (NULL != argv0) {
free(argv0);
}
return exit_status;
}
/* When working in this function, ALWAYS jump to "cleanup" if
* you encounter an error so that orterun will be woken up and
* the job can cleanly terminate
*/
static void launch_daemons(int fd, short args, void *cbdata)
{
orte_job_map_t *map = NULL;
orte_app_context_t *app;
orte_node_t *node;
int proc_vpid_index;
char *param;
char **env = NULL;
char *var;
char **argv = NULL;
char **nodeargv;
int argc = 0;
int rc;
orte_std_cntr_t i;
char *bin_base = NULL, *lib_base = NULL;
tm_event_t *tm_events = NULL;
tm_task_id *tm_task_ids = NULL;
bool failed_launch = true;
mode_t current_umask;
char *nodelist;
char* vpid_string;
orte_job_t *daemons, *jdata;
orte_state_caddy_t *state = (orte_state_caddy_t*)cbdata;
jdata = state->jdata;
/* if we are launching debugger daemons, then just go
* do it - no new daemons will be launched
*/
if (ORTE_JOB_CONTROL_DEBUGGER_DAEMON & jdata->controls) {
jdata->state = ORTE_JOB_STATE_DAEMONS_LAUNCHED;
ORTE_ACTIVATE_JOB_STATE(jdata, ORTE_JOB_STATE_DAEMONS_REPORTED);
OBJ_RELEASE(state);
return;
}
/* setup the virtual machine */
daemons = orte_get_job_data_object(ORTE_PROC_MY_NAME->jobid);
if (ORTE_SUCCESS != (rc = orte_plm_base_setup_virtual_machine(jdata))) {
ORTE_ERROR_LOG(rc);
goto cleanup;
}
/* if we don't want to launch, then don't attempt to
* launch the daemons - the user really wants to just
* look at the proposed process map
*/
if (orte_do_not_launch) {
/* set the state to indicate the daemons reported - this
* will trigger the daemons_reported event and cause the
* job to move to the following step
*/
jdata->state = ORTE_JOB_STATE_DAEMONS_LAUNCHED;
ORTE_ACTIVATE_JOB_STATE(jdata, ORTE_JOB_STATE_DAEMONS_REPORTED);
OBJ_RELEASE(state);
return;
}
/* Get the map for this job */
if (NULL == (map = daemons->map)) {
ORTE_ERROR_LOG(ORTE_ERR_NOT_FOUND);
rc = ORTE_ERR_NOT_FOUND;
goto cleanup;
}
if (0 == map->num_new_daemons) {
/* set the state to indicate the daemons reported - this
* will trigger the daemons_reported event and cause the
* job to move to the following step
*/
jdata->state = ORTE_JOB_STATE_DAEMONS_LAUNCHED;
if (ORTE_JOB_STATE_DAEMONS_REPORTED == daemons->state) {
ORTE_ACTIVATE_JOB_STATE(jdata, ORTE_JOB_STATE_DAEMONS_REPORTED);
}
OBJ_RELEASE(state);
return;
}
OPAL_OUTPUT_VERBOSE((1, orte_plm_globals.output,
"%s plm:tm: launching vm",
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME)));
/* Allocate a bunch of TM events to use for tm_spawn()ing */
tm_events = malloc(sizeof(tm_event_t) * map->num_new_daemons);
if (NULL == tm_events) {
rc = ORTE_ERR_OUT_OF_RESOURCE;
ORTE_ERROR_LOG(rc);
goto cleanup;
}
tm_task_ids = malloc(sizeof(tm_task_id) * map->num_new_daemons);
if (NULL == tm_task_ids) {
rc = ORTE_ERR_OUT_OF_RESOURCE;
ORTE_ERROR_LOG(rc);
goto cleanup;
}
/* add the daemon command (as specified by user) */
orte_plm_base_setup_orted_cmd(&argc, &argv);
/* create a list of nodes in this launch */
nodeargv = NULL;
for (i = 0; i < map->nodes->size; i++) {
if (NULL == (node = (orte_node_t*)opal_pointer_array_get_item(map->nodes, i))) {
continue;
}
/* if this daemon already exists, don't launch it! */
if (node->daemon_launched) {
continue;
}
/* add to list */
opal_argv_append_nosize(&nodeargv, node->name);
}
nodelist = opal_argv_join(nodeargv, ',');
opal_argv_free(nodeargv);
/* Add basic orted command line options */
orte_plm_base_orted_append_basic_args(&argc, &argv, "tm",
&proc_vpid_index,
nodelist);
free(nodelist);
if (0 < opal_output_get_verbosity(orte_plm_globals.output)) {
param = opal_argv_join(argv, ' ');
OPAL_OUTPUT_VERBOSE((1, orte_plm_globals.output,
"%s plm:tm: final top-level argv:\n\t%s",
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
(NULL == param) ? "NULL" : param));
if (NULL != param) free(param);
}
rc = plm_tm_connect();
if (ORTE_SUCCESS != rc) {
goto cleanup;
}
connected = true;
/* Figure out the basenames for the libdir and bindir. There is a
lengthy comment about this in plm_rsh_module.c explaining all
the rationale for how / why we're doing this. */
lib_base = opal_basename(opal_install_dirs.libdir);
bin_base = opal_basename(opal_install_dirs.bindir);
/* setup environment */
env = opal_argv_copy(orte_launch_environ);
/* enable local launch by the orteds */
var = mca_base_param_env_var ("plm");
opal_setenv(var, "rsh", true, &env);
free(var);
/* add our umask -- see big note in orted.c */
current_umask = umask(0);
umask(current_umask);
asprintf(&var, "0%o", current_umask);
opal_setenv("ORTE_DAEMON_UMASK_VALUE", var, true, &env);
free(var);
/* If we have a prefix, then modify the PATH and
LD_LIBRARY_PATH environment variables. We only allow
a single prefix to be specified. Since there will
always be at least one app_context, we take it from
there
*/
app = (orte_app_context_t*)opal_pointer_array_get_item(jdata->apps, 0);
if (NULL != app->prefix_dir) {
char *newenv;
for (i = 0; NULL != env && NULL != env[i]; ++i) {
/* Reset PATH */
if (0 == strncmp("PATH=", env[i], 5)) {
asprintf(&newenv, "%s/%s:%s",
app->prefix_dir, bin_base, env[i] + 5);
OPAL_OUTPUT_VERBOSE((1, orte_plm_globals.output,
"%s plm:tm: resetting PATH: %s",
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
newenv));
opal_setenv("PATH", newenv, true, &env);
free(newenv);
}
/* Reset LD_LIBRARY_PATH */
else if (0 == strncmp("LD_LIBRARY_PATH=", env[i], 16)) {
asprintf(&newenv, "%s/%s:%s",
app->prefix_dir, lib_base, env[i] + 16);
OPAL_OUTPUT_VERBOSE((1, orte_plm_globals.output,
"%s plm:tm: resetting LD_LIBRARY_PATH: %s",
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
newenv));
opal_setenv("LD_LIBRARY_PATH", newenv, true, &env);
free(newenv);
}
}
}
/* Iterate through each of the nodes and spin
* up a daemon.
*/
for (i = 0; i < map->nodes->size; i++) {
if (NULL == (node = (orte_node_t*)opal_pointer_array_get_item(map->nodes, i))) {
continue;
}
/* if this daemon already exists, don't launch it! */
if (node->daemon_launched) {
continue;
}
OPAL_OUTPUT_VERBOSE((1, orte_plm_globals.output,
"%s plm:tm: launching on node %s",
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
node->name));
/* setup process name */
rc = orte_util_convert_vpid_to_string(&vpid_string, node->daemon->name.vpid);
if (ORTE_SUCCESS != rc) {
opal_output(0, "plm:tm: unable to get daemon vpid as string");
exit(-1);
}
free(argv[proc_vpid_index]);
argv[proc_vpid_index] = strdup(vpid_string);
free(vpid_string);
/* exec the daemon */
if (0 < opal_output_get_verbosity(orte_plm_globals.output)) {
param = opal_argv_join(argv, ' ');
OPAL_OUTPUT_VERBOSE((1, orte_plm_globals.output,
"%s plm:tm: executing:\n\t%s",
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
(NULL == param) ? "NULL" : param));
if (NULL != param) free(param);
}
rc = tm_spawn(argc, argv, env, node->launch_id, tm_task_ids + launched, tm_events + launched);
if (TM_SUCCESS != rc) {
orte_show_help("help-plm-tm.txt", "tm-spawn-failed",
true, argv[0], node->name, node->launch_id);
rc = ORTE_ERROR;
goto cleanup;
}
launched++;
}
/* indicate that the daemons for this job were launched */
state->jdata->state = ORTE_JOB_STATE_DAEMONS_LAUNCHED;
daemons->state = ORTE_JOB_STATE_DAEMONS_LAUNCHED;
/* flag that launch was successful, so far as we currently know */
failed_launch = false;
OPAL_OUTPUT_VERBOSE((1, orte_plm_globals.output,
"%s plm:tm:launch: finished spawning orteds",
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME)));
cleanup:
/* cleanup */
OBJ_RELEASE(state);
/* check for failed launch - if so, force terminate */
if (failed_launch) {
ORTE_TERMINATE(ORTE_ERROR_DEFAULT_EXIT_CODE);
}
}
/*
* setup env and argv for specified process
*/
static int setup_proc_env_and_argv(orte_job_t* jdata, orte_app_context_t* app,
orte_proc_t* proc, char ***pargv, char ***penv)
{
char* param;
char* param2;
char* value;
char* vp_id_str;
char* job_id_str;
int rc;
int i, num_nodes;
/* obtain app->argv */
if (!(app->argv)) {
opal_output(0, "%s plm::yarn::setup_proc_env_and_argv: app->argv is null",
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME));
return ORTE_ERROR;
}
*pargv = opal_argv_copy(app->argv);
if (ORTE_SUCCESS != orte_util_convert_jobid_to_string(&job_id_str, jdata->jobid)) {
ORTE_ERROR_LOG(ORTE_ERR_OUT_OF_RESOURCE);
return ORTE_ERR_OUT_OF_RESOURCE;
}
if (ORTE_SUCCESS != orte_util_convert_vpid_to_string(&vp_id_str, proc->name.vpid)) {
ORTE_ERROR_LOG(ORTE_ERR_OUT_OF_RESOURCE);
return ORTE_ERR_OUT_OF_RESOURCE;
}
// add stdout, stderr to app
opal_argv_append_nosize(pargv, "1><LOG_DIR>/stdout");
opal_argv_append_nosize(pargv, "2><LOG_DIR>/stderr");
// add java executor to app
opal_argv_prepend_nosize(pargv, vp_id_str);
opal_argv_prepend_nosize(pargv, job_id_str);
opal_argv_prepend_nosize(pargv, "com.pivotal.hamster.yarnexecutor.YarnExecutor");
opal_argv_prepend_nosize(pargv, "hamster-core.jar");
opal_argv_prepend_nosize(pargv, "-cp");
opal_argv_prepend_nosize(pargv, getenv("HAMSTER_JAVA_OPT")==NULL ? "-Xmx32M -Xms8M" : getenv("HAMSTER_JAVA_OPT"));
opal_argv_prepend_nosize(pargv, "$JAVA_HOME/bin/java");
/* obtain app->env */
*penv = opal_environ_merge(environ, app->env);
if (!proc->node) {
opal_output(0, "%s plm::yarn::setup_proc_env_and_argv: node of proc[%d] is NULL",
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME), proc->name.vpid);
return ORTE_ERROR;
}
if (!proc->node->daemon) {
opal_output(0, "%s plm::yarn::setup_proc_env_and_argv: daemon of node[%s] is NULL",
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME), proc->node->name);
return ORTE_ERROR;
}
/* set the app_context number into the environment */ //??
param = mca_base_param_env_var("orte_app_num");
asprintf(¶m2, "%ld", (long)app->idx);
opal_setenv(param, param2, true, penv);
free(param);
free(param2);
// pass the daemon's name
param = mca_base_param_env_var("orte_local_daemon_uri");
opal_setenv(param, proc->node->daemon->rml_uri, true, penv);
free(param);
/* pass my contact info */
param = mca_base_param_env_var("orte_hnp_uri");
opal_setenv(param, orte_process_info.my_hnp_uri, true, penv);
free(param);
/* pass the jobid */
param = mca_base_param_env_var("orte_ess_jobid");
opal_setenv(param, job_id_str, true, penv);
free(param);
free(job_id_str);
/* pass the rank */
param = mca_base_param_env_var("orte_ess_vpid");
opal_setenv(param, vp_id_str, true, penv);
free(param);
opal_setenv("OMPI_COMM_WORLD_RANK", vp_id_str, true, penv);
free(vp_id_str); /* done with this now */
/* pass local rank */
asprintf(&value, "%lu", (unsigned long) proc->local_rank);
opal_setenv("OMPI_COMM_WORLD_LOCAL_RANK", value, true, penv);
free(value);
/* pass node rank */
asprintf(&value, "%lu", (unsigned long) proc->node_rank);
opal_setenv("OMPI_COMM_WORLD_NODE_RANK", value, true, penv);
/* set an mca param for it too */
param = mca_base_param_env_var("orte_ess_node_rank");
opal_setenv(param, value, true, penv);
free(param);
free(value);
/* pass a param telling the child what model of cpu we are on,
* if we know it
*/
if (NULL != orte_local_cpu_type) {
param = mca_base_param_env_var("orte_cpu_type");
/* do not overwrite what the user may have provided */
opal_setenv(param, orte_local_cpu_type, false, penv);
free(param);
}
if (NULL != orte_local_cpu_model) {
param = mca_base_param_env_var("orte_cpu_model");
/* do not overwrite what the user may have provided */
opal_setenv(param, orte_local_cpu_model, false, penv);
free(param);
}
/* pass the number of nodes involved in this job */
param = mca_base_param_env_var("orte_num_nodes");
/* we have to count the number of nodes as the size of orte_node_pool
* is only guaranteed to be equal or larger than that number - i.e.,
* the pointer_array increases the size by a block each time, so some
* of the locations are left empty
*/
num_nodes = 0;
for (i = 0; i < orte_node_pool->size; i++) {
if (NULL != opal_pointer_array_get_item(orte_node_pool, i)) {
num_nodes++;
}
}
asprintf(&value, "%d", num_nodes);
opal_setenv(param, value, true, penv);
free(param);
free(value);
/* setup yield schedule */
param = mca_base_param_env_var("mpi_yield_when_idle");
opal_setenv(param, "0", false, penv);
free(param);
/* set MPI universe envar */
orte_ess_env_put(jdata->num_procs, proc->node->num_procs, penv);
asprintf(&value, "%ld", (long) jdata->num_procs);
opal_setenv("OMPI_UNIVERSE_SIZE", value, true, penv);
free(value);
/* pass collective ids for the std MPI operations */
param = mca_base_param_env_var("orte_peer_modex_id");
asprintf(&value, "%d", jdata->peer_modex);
opal_setenv(param, value, true, penv);
free(param);
free(value);
param = mca_base_param_env_var("orte_peer_init_barrier_id");
asprintf(&value, "%d", jdata->peer_init_barrier);
opal_setenv(param, value, true, penv);
free(param);
free(value);
param = mca_base_param_env_var("orte_peer_fini_barrier_id");
asprintf(&value, "%d", jdata->peer_fini_barrier);
opal_setenv(param, value, true, penv);
free(param);
free(value);
/* finally, we will set/unset some mca param to select modules */
opal_unsetenv("OMPI_MCA_plm", penv);
opal_unsetenv("OMPI_MCA_ras", penv);
opal_unsetenv("OMPI_MCA_ess", penv);
opal_unsetenv("OMPI_MCA_state", penv);
opal_unsetenv("OMPI_MCA_errmgr", penv);
return 0;
}
static int orte_ps_init(int argc, char *argv[]) {
int ret;
char * tmp_env_var = NULL;
/*
* Make sure to init util before parse_args
* to ensure installdirs is setup properly
* before calling mca_base_open();
*/
if( ORTE_SUCCESS != (ret = opal_init_util(&argc, &argv)) ) {
return ret;
}
/*
* Parse Command Line Arguments
*/
if (ORTE_SUCCESS != (ret = parse_args(argc, argv))) {
return ret;
}
/*
* Setup OPAL Output handle from the verbose argument
*/
if( orte_ps_globals.verbose ) {
orte_ps_globals.output = opal_output_open(NULL);
opal_output_set_verbosity(orte_ps_globals.output, 10);
} else {
orte_ps_globals.output = 0; /* Default=STDERR */
}
#if OPAL_ENABLE_FT_CR == 1
/* Disable the checkpoint notification routine for this
* tool. As we will never need to checkpoint this tool.
* Note: This must happen before opal_init().
*/
opal_cr_set_enabled(false);
/* Select the none component, since we don't actually use a checkpointer */
tmp_env_var = mca_base_param_env_var("crs");
opal_setenv(tmp_env_var,
"none",
true, &environ);
free(tmp_env_var);
tmp_env_var = NULL;
tmp_env_var = mca_base_param_env_var("opal_cr_is_tool");
opal_setenv(tmp_env_var,
"1",
true, &environ);
free(tmp_env_var);
#else
tmp_env_var = NULL; /* Silence compiler warning */
#endif
/***************************
* We need all of OPAL and the TOOL portion of ORTE
***************************/
ret = orte_init(&argc, &argv, ORTE_PROC_TOOL);
return ret;
}
/*
* This utility will do a brute force clean of a node. It will
* attempt to clean up any files in the user's session directory.
* It will also look for any orted and orterun processes that are
* not part of this job, and kill them off.
*/
int
main(int argc, char *argv[])
{
int ret = ORTE_SUCCESS;
char *tmp_env_var;
/* This is needed so we can print the help message */
if (ORTE_SUCCESS != (ret = opal_init_util(&argc, &argv))) {
return ret;
}
if (ORTE_SUCCESS != (ret = parse_args(argc, argv))) {
return ret;
}
#if OPAL_ENABLE_FT_CR == 1
/* Disable the checkpoint notification routine for this
* tool. As we will never need to checkpoint this tool.
* Note: This must happen before opal_init().
*/
opal_cr_set_enabled(false);
/* Select the none component, since we don't actually use a checkpointer */
tmp_env_var = mca_base_param_env_var("crs");
opal_setenv(tmp_env_var,
"none",
true, &environ);
free(tmp_env_var);
tmp_env_var = NULL;
tmp_env_var = mca_base_param_env_var("opal_cr_is_tool");
opal_setenv(tmp_env_var,
"1", true, NULL);
free(tmp_env_var);
#else
tmp_env_var = NULL; /* Silence compiler warning */
#endif
if (ORTE_SUCCESS != (ret = orte_init(&argc, &argv, ORTE_PROC_TOOL))) {
return ret;
}
/*
* Clean out all session directories - we don't have to protect
* our own session directory because (since we are a tool) we
* didn't create one!
*/
if (orte_clean_globals.verbose) {
fprintf(stderr, "orte-clean: cleaning session dir tree %s\n",
orte_process_info.top_session_dir);
}
opal_os_dirpath_destroy(orte_process_info.top_session_dir, true, NULL);
/* now kill any lingering procs, if we can */
#if !defined(__WINDOWS__)
kill_procs();
#endif /* !defined(__WINDOWS__) */
orte_finalize();
return ORTE_SUCCESS;
}
int main(int argc, char *argv[])
{
int ret = 0;
opal_cmd_line_t *cmd_line = NULL;
char *rml_uri;
char * tmp_env_var = NULL;
/* init enough of opal to process cmd lines */
if (OPAL_SUCCESS != opal_init_util(&argc, &argv)) {
fprintf(stderr, "OPAL failed to initialize -- orted aborting\n");
exit(1);
}
/* setup to check common command line options that just report and die */
cmd_line = OBJ_NEW(opal_cmd_line_t);
opal_cmd_line_create(cmd_line, ompi_server_cmd_line_opts);
mca_base_cmd_line_setup(cmd_line);
if (OPAL_SUCCESS != (ret = opal_cmd_line_parse(cmd_line, false,
argc, argv))) {
if (OPAL_ERR_SILENT != ret) {
fprintf(stderr, "%s: command line error (%s)\n", argv[0],
opal_strerror(ret));
}
return 1;
}
/* check for help request */
if (help) {
char *str, *args = NULL;
args = opal_cmd_line_get_usage_msg(cmd_line);
str = opal_show_help_string("help-ompi-server.txt",
"ompiserver:usage", false,
argv[0], args);
if (NULL != str) {
printf("%s", str);
free(str);
}
free(args);
/* If we show the help message, that should be all we do */
return 0;
}
/*
* Since this process can now handle MCA/GMCA parameters, make sure to
* process them.
*/
mca_base_cmd_line_process_args(cmd_line, &environ, &environ);
/* if debug is set, then set orte_debug_flag so that the data server
* code will output
*/
if (debug) {
putenv("OMPI_MCA_orte_debug=1");
}
/* detach from controlling terminal
* otherwise, remain attached so output can get to us
*/
if(debug == false &&
no_daemonize == false) {
opal_daemon_init(NULL);
}
#if OPAL_ENABLE_FT_CR == 1
/* Disable the checkpoint notification routine for this
* tool. As we will never need to checkpoint this tool.
* Note: This must happen before opal_init().
*/
opal_cr_set_enabled(false);
/* Select the none component, since we don't actually use a checkpointer */
tmp_env_var = mca_base_param_env_var("crs");
opal_setenv(tmp_env_var,
"none",
true, &environ);
free(tmp_env_var);
tmp_env_var = NULL;
/* Mark as a tool program */
tmp_env_var = mca_base_param_env_var("opal_cr_is_tool");
opal_setenv(tmp_env_var,
"1",
true, &environ);
free(tmp_env_var);
#else
tmp_env_var = NULL; /* Silence compiler warning */
#endif
/* don't want session directories */
orte_create_session_dirs = false;
/* Perform the standard init, but flag that we are an HNP */
if (ORTE_SUCCESS != (ret = orte_init(&argc, &argv, ORTE_PROC_HNP))) {
fprintf(stderr, "ompi-server: failed to initialize -- aborting\n");
exit(1);
}
/* report out our URI, if we were requested to do so, using syntax
* proposed in an email thread by Jeff Squyres
*/
if (NULL != report_uri) {
rml_uri = orte_rml.get_contact_info();
if (0 == strcmp(report_uri, "-")) {
/* if '-', then output to stdout */
printf("%s\n", rml_uri);
} else if (0 == strcmp(report_uri, "+")) {
/* if '+', output to stderr */
fprintf(stderr, "%s\n", rml_uri);
} else {
/* treat it as a filename and output into it */
FILE *fp;
fp = fopen(report_uri, "w");
if (NULL == fp) {
fprintf(stderr, "ompi-server: failed to open designated file %s -- aborting\n", report_uri);
orte_finalize();
exit(1);
}
fprintf(fp, "%s\n", rml_uri);
fclose(fp);
}
free(rml_uri);
}
/* setup the data server to listen for commands */
if (ORTE_SUCCESS != (ret = orte_data_server_init())) {
fprintf(stderr, "ompi-server: failed to start data server -- aborting\n");
orte_finalize();
exit(1);
}
/* setup to listen for commands sent specifically to me */
ret = orte_rml.recv_buffer_nb(ORTE_NAME_WILDCARD, ORTE_RML_TAG_DAEMON,
ORTE_RML_NON_PERSISTENT, orte_daemon_recv, NULL);
if (ret != ORTE_SUCCESS && ret != ORTE_ERR_NOT_IMPLEMENTED) {
ORTE_ERROR_LOG(ret);
orte_finalize();
exit(1);
}
/* Set signal handlers to catch kill signals so we can properly clean up
* after ourselves.
*/
opal_event_set(opal_event_base, &term_handler, SIGTERM, OPAL_EV_SIGNAL,
shutdown_callback, NULL);
opal_event_add(&term_handler, NULL);
opal_event_set(opal_event_base, &int_handler, SIGINT, OPAL_EV_SIGNAL,
shutdown_callback, NULL);
opal_event_add(&int_handler, NULL);
/* We actually do *not* want the server to voluntarily yield() the
processor more than necessary. The server already blocks when
it is doing nothing, so it doesn't use any more CPU cycles than
it should; but when it *is* doing something, we do not want it
to be unnecessarily delayed because it voluntarily yielded the
processor in the middle of its work.
For example: when a message arrives at the server, we want the
OS to wake up the server in a timely fashion (which most OS's
seem good about doing) and then we want the server to process
the message as fast as possible. If the server yields and lets
aggressive MPI applications get the processor back, it may be a
long time before the OS schedules the server to run again
(particularly if there is no IO event to wake it up). Hence,
publish and lookup (for example) may be significantly delayed
before being delivered to MPI processes, which can be
problematic in some scenarios (e.g., COMM_SPAWN). */
opal_progress_set_yield_when_idle(false);
/* Change the default behavior of libevent such that we want to
continually block rather than blocking for the default timeout
and then looping around the progress engine again. There
should be nothing in the server that cannot block in libevent
until "something" happens (i.e., there's no need to keep
cycling through progress because the only things that should
happen will happen in libevent). This is a minor optimization,
but what the heck... :-) */
opal_progress_set_event_flag(OPAL_EVLOOP_ONCE);
if (debug) {
opal_output(0, "%s ompi-server: up and running!", ORTE_NAME_PRINT(ORTE_PROC_MY_NAME));
}
/* wait to hear we are done */
while (orte_event_base_active) {
opal_event_loop(orte_event_base, OPAL_EVLOOP_ONCE);
}
/* should never get here, but if we do... */
/* Finalize and clean up ourselves */
if (ORTE_SUCCESS != (ret = orte_finalize())) {
ORTE_ERROR_LOG(ret);
}
return ret;
}
static int parse_args(int argc, char *argv[]) {
int i, ret, len, exit_status = ORTE_SUCCESS ;
opal_cmd_line_t cmd_line;
char **app_env = NULL, **global_env = NULL;
char * tmp_env_var = NULL;
char *argv0 = NULL;
/* Init structure */
memset(&orte_checkpoint_globals, 0, sizeof(orte_checkpoint_globals_t));
orte_checkpoint_globals.help = false;
orte_checkpoint_globals.pid = -1;
orte_checkpoint_globals.verbose = false;
orte_checkpoint_globals.verbose_level = 0;
orte_checkpoint_globals.req_hnp = ORTE_JOBID_INVALID;
orte_checkpoint_globals.nowait = false;
orte_checkpoint_globals.status = false;
orte_checkpoint_globals.output = -1;
orte_checkpoint_globals.ckpt_status = ORTE_SNAPC_CKPT_STATE_NONE;
orte_checkpoint_globals.list_only = false;
#if OPAL_ENABLE_CRDEBUG == 1
orte_checkpoint_globals.enable_crdebug = false;
#endif
orte_checkpoint_globals.options = OBJ_NEW(opal_crs_base_ckpt_options_t);
orte_checkpoint_globals.term = false;
orte_checkpoint_globals.stop = false;
#if OPAL_ENABLE_CRDEBUG == 1
orte_checkpoint_globals.attach_debugger = false;
orte_checkpoint_globals.detach_debugger = false;
#endif
#if OPAL_ENABLE_FT_CR == 0
/* Warn and exit if not configured with Checkpoint/Restart */
{
char *str, *args = NULL;
args = opal_cmd_line_get_usage_msg(&cmd_line);
str = opal_show_help_string("help-orte-checkpoint.txt", "usage-no-cr",
true, args);
if (NULL != str) {
printf("%s", str);
free(str);
}
free(args);
exit_status = ORTE_ERROR;
goto cleanup;
}
#endif
/* Parse the command line options */
opal_cmd_line_create(&cmd_line, cmd_line_opts);
mca_base_open();
mca_base_cmd_line_setup(&cmd_line);
ret = opal_cmd_line_parse(&cmd_line, true, argc, argv);
if (OPAL_SUCCESS != ret) {
if (OPAL_ERR_SILENT != ret) {
fprintf(stderr, "%s: command line error (%s)\n", argv[0],
opal_strerror(ret));
}
exit_status = 1;
goto cleanup;
}
if (orte_checkpoint_globals.help) {
char *str, *args = NULL;
args = opal_cmd_line_get_usage_msg(&cmd_line);
str = opal_show_help_string("help-orte-checkpoint.txt", "usage", true,
args);
if (NULL != str) {
printf("%s", str);
free(str);
}
free(args);
/* If we show the help message, that should be all we do */
exit(0);
}
/**
* Put all of the MCA arguments in the environment
*/
mca_base_cmd_line_process_args(&cmd_line, &app_env, &global_env);
len = opal_argv_count(app_env);
for(i = 0; i < len; ++i) {
putenv(app_env[i]);
}
len = opal_argv_count(global_env);
for(i = 0; i < len; ++i) {
putenv(global_env[i]);
}
tmp_env_var = mca_base_param_env_var("opal_cr_is_tool");
opal_setenv(tmp_env_var,
"1",
true, &environ);
free(tmp_env_var);
tmp_env_var = NULL;
/**
* Now start parsing our specific arguments
*/
/* get the remaining bits */
argv0 = strdup(argv[0]);
opal_cmd_line_get_tail(&cmd_line, &argc, &argv);
if(orte_checkpoint_globals.list_only ) {
exit_status = ORTE_SUCCESS;
goto cleanup;
}
if (0 >= argc && ORTE_JOBID_INVALID == orte_checkpoint_globals.req_hnp) {
fprintf(stderr, "%s: Nothing to do\n", argv0);
fprintf(stderr, "Type '%s --help' for usage.\n", argv0);
exit_status = 1;
goto cleanup;
}
orte_checkpoint_globals.options->term = orte_checkpoint_globals.term;
orte_checkpoint_globals.options->stop = orte_checkpoint_globals.stop;
#if OPAL_ENABLE_CRDEBUG == 1
orte_checkpoint_globals.options->attach_debugger = orte_checkpoint_globals.attach_debugger;
orte_checkpoint_globals.options->detach_debugger = orte_checkpoint_globals.detach_debugger;
#endif
if(orte_checkpoint_globals.verbose_level < 0 ) {
orte_checkpoint_globals.verbose_level = 0;
}
if(orte_checkpoint_globals.verbose_level > 0) {
orte_checkpoint_globals.verbose = true;
}
/*
* If the user did not supply an hnp jobid, then they must
* supply the PID of MPIRUN
*/
if(0 >= argc &&
ORTE_JOBID_INVALID != orte_checkpoint_globals.req_hnp) {
exit_status = ORTE_SUCCESS;
goto cleanup;
}
orte_checkpoint_globals.pid = atoi(argv[0]);
if ( 0 >= orte_checkpoint_globals.pid ) {
opal_show_help("help-orte-checkpoint.txt", "invalid_pid", true,
orte_checkpoint_globals.pid);
exit_status = ORTE_ERROR;
goto cleanup;
}
/*
* JJH: No wait is currently not implemented or tested
*/
if(orte_checkpoint_globals.nowait) {
orte_checkpoint_globals.nowait = false;
opal_show_help("help-orte-checkpoint.txt", "not_impl",
true,
"Disconnected checkpoint");
}
if(orte_checkpoint_globals.verbose) {
orte_checkpoint_globals.status = true;
}
cleanup:
if (NULL != argv0) {
free(argv0);
}
return exit_status;
}
