int main( int argc, char *argv[] )
{
signed char c;
int work_queue_master_mode = WORK_QUEUE_MASTER_MODE_STANDALONE;
char *project = NULL;
int priority = 0;
const char *progname = "wavefront";
debug_config(progname);
struct option long_options[] = {
{"help", no_argument, 0, 'h'},
{"version", no_argument, 0, 'v'},
{"debug", required_argument, 0, 'd'},
{"advertise", no_argument, 0, 'a'},
{"project-name", required_argument, 0, 'N'},
{"debug-file", required_argument, 0, 'o'},
{"port", required_argument, 0, 'p'},
{"priority", required_argument, 0, 'P'},
{"estimated-time", required_argument, 0, 't'},
{"random-port", required_argument, 0, 'Z'},
{"bitmap", required_argument, 0, 'B'},
{0,0,0,0}
};
while((c=getopt_long(argc,argv,"aB:d:hN:p:P:o:v:Z:", long_options, NULL)) >= 0) {
switch(c) {
case 'a':
break;
case 'd':
debug_flags_set(optarg);
break;
case 'h':
show_help(progname);
exit(0);
break;
case 'N':
work_queue_master_mode = WORK_QUEUE_MASTER_MODE_CATALOG;
free(project);
project = xxstrdup(optarg);
break;
case 'p':
port = atoi(optarg);
break;
case 'P':
priority = atoi(optarg);
break;
case 'o':
debug_config_file(optarg);
break;
case 'v':
cctools_version_print(stdout, progname);
exit(0);
break;
case 'Z':
port_file = optarg;
port = 0;
break;
case 'B':
progress_bitmap_file = optarg;
break;
default:
show_help(progname);
return 1;
}
}
cctools_version_debug(D_DEBUG, argv[0]);
if( (argc-optind)!=5 ) {
show_help(progname);
exit(1);
}
function = argv[optind];
xsize=atoi(argv[optind+1]);
ysize=atoi(argv[optind+2]);
infile=argv[optind+3];
outfile=argv[optind+4];
start_time = time(0);
last_display_time = 0;
cells_total = xsize*ysize;
xsize++;
ysize++;
array = text_array_create(xsize,ysize);
if(!text_array_load(array,infile)) {
fprintf(stderr,"couldn't load %s: %s",infile,strerror(errno));
return 1;
}
int count = text_array_load(array,outfile);
if(count>0) printf("recovered %d results from %s\n",count,outfile);
logfile = fopen(outfile,"a");
if(!logfile) {
fprintf(stderr,"couldn't open %s for append: %s\n",outfile,strerror(errno));
return 1;
}
if(work_queue_master_mode == WORK_QUEUE_MASTER_MODE_CATALOG && !project) {
fprintf(stderr, "wavefront: wavefront master running in catalog mode. Please use '-N' option to specify the name of this project.\n");
fprintf(stderr, "wavefront: Run \"%s -h\" for help with options.\n", argv[0]);
return 1;
}
queue = work_queue_create(port);
//Read the port the queue is actually running, in case we just called
//work_queue_create(LINK_PORT_ANY)
port = work_queue_port(queue);
if(!queue) {
fprintf(stderr,"%s: could not create work queue on port %d: %s\n",progname,port,strerror(errno));
return 1;
}
if(port_file)
opts_write_port_file(port_file, port);
// advanced work queue options
work_queue_specify_master_mode(queue, work_queue_master_mode);
work_queue_specify_name(queue, project);
work_queue_specify_priority(queue, priority);
fprintf(stdout, "%s: listening for workers on port %d...\n",progname,work_queue_port(queue));
if(progress_bitmap_file)
{
bmap = bitmap_create(xsize,ysize);
wavefront_bitmap_initialize(bmap);
}
task_prime();
struct work_queue_task *t;
while(1) {
if(time(0)!=last_display_time) display_progress(queue);
t = work_queue_wait(queue,WORK_QUEUE_WAITFORTASK);
if(!t) break;
if(t->return_status==0) {
int x,y;
if(sscanf(t->tag,"%d %d",&x,&y)==2) {
text_array_set(array,x,y,t->output);
task_complete(x,y);
fprintf(logfile,"%d %d %s\n",x,y,t->output);
fflush(logfile);
tasks_done++;
} else {
fprintf(stderr,"unexpected output: %s\nfrom command: %s\non host: %s",t->output,t->command_line,t->host);
}
} else {
fprintf(stderr,"function failed return value (%i) result (%i) on host %s. output:\n%s\n",t->return_status,t->result,t->host,t->output);
}
work_queue_task_delete(t);
if(work_queue_empty(queue))
break;
}
display_progress(queue);
return 0;
}
int main(int argc, char *argv[])
{
char c;
int did_explicit_auth = 0;
char *tickets = NULL;
struct fuse_args fa;
fa.argc = 0;
fa.argv = string_array_new();
fa.allocated = 1;
debug_config(argv[0]);
while((c = getopt(argc, argv, "a:b:d:Dfhi:m:o:t:v")) != -1) {
switch (c) {
case 'd':
debug_flags_set(optarg);
break;
case 'D':
enable_small_file_optimizations = 0;
break;
case 'b':
chirp_reli_blocksize_set(atoi(optarg));
break;
case 'i':
tickets = xxstrdup(optarg);
break;
case 'm':
fa.argc += 1;
fa.argv = string_array_append(fa.argv, optarg);
break;
case 'o':
debug_config_file(optarg);
break;
case 'a':
auth_register_byname(optarg);
did_explicit_auth = 1;
break;
case 't':
chirp_fuse_timeout = string_time_parse(optarg);
break;
case 'f':
run_in_foreground = 1;
break;
case 'v':
cctools_version_print(stdout, argv[0]);
return 0;
break;
case 'h':
default:
show_help(argv[0]);
return 1;
break;
}
}
cctools_version_debug(D_DEBUG, argv[0]);
if((argc - optind) != 1) {
show_help(argv[0]);
return 1;
}
fuse_mountpoint = argv[optind];
if(!did_explicit_auth)
auth_register_all();
if(tickets) {
auth_ticket_load(tickets);
free(tickets);
} else if(getenv(CHIRP_CLIENT_TICKETS)) {
auth_ticket_load(getenv(CHIRP_CLIENT_TICKETS));
} else {
auth_ticket_load(NULL);
}
file_table = itable_create(0);
signal(SIGHUP, exit_handler);
signal(SIGINT, exit_handler);
signal(SIGTERM, exit_handler);
fuse_chan = fuse_mount(fuse_mountpoint, &fa);
if(!fuse_chan) {
fprintf(stderr, "chirp_fuse: couldn't access %s\n", fuse_mountpoint);
return 1;
}
fuse_instance = fuse_new(fuse_chan, &fa, &chirp_fuse_operations, sizeof(chirp_fuse_operations), 0);
if(!fuse_instance) {
fuse_unmount(fuse_mountpoint, fuse_chan);
fprintf(stderr, "chirp_fuse: couldn't access %s\n", fuse_mountpoint);
return 1;
}
printf("chirp_fuse: mounted chirp on %s\n", fuse_mountpoint);
#ifdef CCTOOLS_OPSYS_DARWIN
printf("chirp_fuse: to unmount: umount %s\n", fuse_mountpoint);
#else
printf("chirp_fuse: to unmount: fusermount -u %s\n", fuse_mountpoint);
#endif
fflush(0);
if(!run_in_foreground)
daemon(0, 0);
fuse_loop(fuse_instance);
fuse_unmount(fuse_mountpoint, fuse_chan);
fuse_destroy(fuse_instance);
free(fa.argv);
return 0;
}
static void get_options(int argc, char **argv, const char *progname)
{
signed char c;
char tmp[512];
char *catalog_host = NULL;
int catalog_port = 0;
while((c = getopt(argc, argv, "p:P:n:d:F:N:C:s:r:R:k:w:c:o:uxvhaZ:")) > -1) {
switch (c) {
case 'p':
port = atoi(optarg);
break;
case 'r':
repeat_filename = optarg;
break;
case 'R':
retry_max = atoi(optarg);
break;
case 's':
rectangle_size = atoi(optarg);
break;
case 'k':
kmer_size = atoi(optarg);
break;
case 'w':
window_size = atoi(optarg);
break;
case 'c':
checkpoint_filename = optarg;
break;
case 'd':
debug_flags_set(optarg);
break;
case 'F':
wq_option_fast_abort_multiplier = atof(optarg);
break;
case 'a':
work_queue_master_mode = WORK_QUEUE_MASTER_MODE_CATALOG;
break;
case 'N':
free(project);
project = xxstrdup(optarg);
break;
case 'P':
priority = atoi(optarg);
break;
case 'C':
if(!parse_catalog_server_description(optarg, &catalog_host, &catalog_port)) {
fprintf(stderr, "sand_filter: catalog server should be given as HOSTNAME:PORT'.\n");
exit(1);
}
setenv("CATALOG_HOST", catalog_host, 1);
char *value = string_format("%d", catalog_port);
setenv("CATALOG_PORT", value, 1);
free(value);
work_queue_master_mode = WORK_QUEUE_MASTER_MODE_CATALOG;
break;
case 'u':
do_not_unlink = 1;
break;
case 'Z':
port_file = optarg;
port = 0;
break;
case 'o':
debug_config_file(optarg);
break;
case 'v':
cctools_version_print(stdout, progname);
exit(0);
default:
case 'h':
show_help(progname);
exit(0);
}
}
cctools_version_debug(D_DEBUG, argv[0]);
if(argc - optind != 2) {
show_help(progname);
exit(1);
}
sequence_filename = argv[optind++];
outfilename = argv[optind++];
outdirname = malloc(strlen(outfilename) + 15);
sprintf(outdirname, "%s.filter.tmp", outfilename);
if(mkdir(outdirname, S_IRWXU) != 0) {
if(errno == EEXIST) {
fprintf(stderr, "%s: directory %s already exists, you may want to delete or rename before running.\n", progname, outdirname);
} else {
fprintf(stderr, "%s: couldn't create %s: %s\n", progname, outdirname, strerror(errno));
exit(1);
}
}
sprintf(filter_program_args, "-k %d -w %d -s d", kmer_size, window_size);
if(repeat_filename) {
sprintf(tmp, " -r %s", string_basename(repeat_filename));
strcat(filter_program_args, tmp);
}
}
int main(int argc, char *argv[])
{
struct link *link, *list_port = 0;
char ch;
time_t current;
int is_daemon = 0;
char *pidfile = NULL;
outgoing_host_list = list_create();
debug_config(argv[0]);
while((ch = getopt(argc, argv, "bB:d:hH:l:L:m:M:n:o:O:p:ST:u:U:v")) != (char) -1) {
switch (ch) {
case 'b':
is_daemon = 1;
break;
case 'B':
free(pidfile);
pidfile = strdup(optarg);
break;
case 'd':
debug_flags_set(optarg);
break;
case 'h':
default:
show_help(argv[0]);
return 1;
case 'l':
lifetime = string_time_parse(optarg);
break;
case 'L':
logfilename = strdup(optarg);
break;
case 'H':
history_dir = strdup(optarg);
break;
case 'm':
child_procs_max = atoi(optarg);
break;
case 'M':
max_server_size = string_metric_parse(optarg);
break;
case 'n':
preferred_hostname = optarg;
break;
case 'o':
free(debug_filename);
debug_filename = strdup(optarg);
break;
case 'O':
debug_config_file_size(string_metric_parse(optarg));
break;
case 'p':
port = atoi(optarg);
break;
case 'S':
fork_mode = 0;
break;
case 'T':
child_procs_timeout = string_time_parse(optarg);
break;
case 'u':
list_push_head(outgoing_host_list, xxstrdup(optarg));
break;
case 'U':
outgoing_timeout = string_time_parse(optarg);
break;
case 'v':
cctools_version_print(stdout, argv[0]);
return 0;
}
}
if (is_daemon) daemonize(0, pidfile);
debug_config_file(debug_filename);
cctools_version_debug(D_DEBUG, argv[0]);
if(logfilename) {
logfile = fopen(logfilename,"a");
if(!logfile) fatal("couldn't open %s: %s\n",optarg,strerror(errno));
}
current = time(0);
debug(D_ALL, "*** %s starting at %s", argv[0], ctime(¤t));
if(!list_size(outgoing_host_list)) {
list_push_head(outgoing_host_list, CATALOG_HOST_DEFAULT);
}
install_handler(SIGPIPE, ignore_signal);
install_handler(SIGHUP, ignore_signal);
install_handler(SIGCHLD, ignore_signal);
install_handler(SIGINT, shutdown_clean);
install_handler(SIGTERM, shutdown_clean);
install_handler(SIGQUIT, shutdown_clean);
install_handler(SIGALRM, shutdown_clean);
if(!preferred_hostname) {
domain_name_cache_guess(hostname);
preferred_hostname = hostname;
}
username_get(owner);
starttime = time(0);
table = nvpair_database_create(history_dir);
if(!table)
fatal("couldn't create directory %s: %s\n",history_dir,strerror(errno));
update_dgram = datagram_create(port);
if(!update_dgram)
fatal("couldn't listen on udp port %d", port);
outgoing_dgram = datagram_create(0);
if(!outgoing_dgram)
fatal("couldn't create outgoing udp port");
list_port = link_serve(port);
if(!list_port)
fatal("couldn't listen on tcp port %d", port);
while(1) {
fd_set rfds;
int ufd = datagram_fd(update_dgram);
int lfd = link_fd(list_port);
int result, maxfd;
struct timeval timeout;
remove_expired_records();
if(time(0) > outgoing_alarm) {
update_all_catalogs(outgoing_dgram);
outgoing_alarm = time(0) + outgoing_timeout;
}
while(1) {
int status;
pid_t pid = waitpid(-1, &status, WNOHANG);
if(pid>0) {
child_procs_count--;
continue;
} else {
break;
}
}
FD_ZERO(&rfds);
FD_SET(ufd, &rfds);
if(child_procs_count < child_procs_max) {
FD_SET(lfd, &rfds);
}
maxfd = MAX(ufd, lfd) + 1;
timeout.tv_sec = 5;
timeout.tv_usec = 0;
result = select(maxfd, &rfds, 0, 0, &timeout);
if(result <= 0)
continue;
if(FD_ISSET(ufd, &rfds)) {
handle_updates(update_dgram);
}
if(FD_ISSET(lfd, &rfds)) {
link = link_accept(list_port, time(0) + 5);
if(link) {
if(fork_mode) {
pid_t pid = fork();
if(pid == 0) {
alarm(child_procs_timeout);
handle_query(link);
_exit(0);
} else if (pid>0) {
child_procs_count++;
}
} else {
handle_query(link);
}
link_close(link);
}
}
}
return 1;
}
int main(int argc, char *argv[])
{
signed char c;
random_init();
install_handler(SIGINT, handle_signal);
install_handler(SIGTERM, handle_signal);
install_handler(SIGQUIT, handle_signal);
install_handler(SIGCHLD, handle_signal);
install_handler(SIGHUP, handle_signal);
debug_config(argv[0]);
while(((c = getopt(argc, argv, "+d:o:O:m:M:s:S:vh")) > -1)) {
switch (c) {
case 'c':
check_interval = string_time_parse(optarg);
break;
case 'd':
debug_flags_set(optarg);
break;
case 'o':
debug_config_file(optarg);
break;
case 'O':
debug_config_file_size(string_metric_parse(optarg));
break;
case 'm':
min_wait_time = string_time_parse(optarg);
break;
case 'M':
max_wait_time = string_time_parse(optarg);
break;
case 's':
start_interval = string_time_parse(optarg);
break;
case 'S':
stop_interval = string_time_parse(optarg);
break;
case 'v':
cctools_version_print(stdout, argv[0]);
exit(0);
break;
case 'h':
default:
show_help(argv[0]);
break;
}
}
cctools_version_debug(D_DEBUG, argv[0]);
if(optind >= argc) {
show_help(argv[0]);
return 1;
}
program_argv = &argv[optind];
state_start = time(0);
state = STATE_READY;
if(program_argv[0][0] != '/') {
fprintf(stderr, "watchdog: please give me the full path to %s\n", program_argv[0]);
return 1;
}
while(1) {
time_t time_in_state = time(0) - state_start;
switch (state) {
case STATE_READY:
if(start_program()) {
change_state(STATE_STARTED);
} else {
change_state(STATE_STOPPED);
start_failures++;
}
break;
case STATE_STARTED:
if(program_exited()) {
change_state(STATE_STOPPED);
start_failures++;
} else if(time_in_state > start_interval) {
change_state(STATE_RUNNING);
} else if(want_to_exit) {
change_state(STATE_STOP_WAIT);
}
break;
case STATE_RUNNING:
start_failures = 0;
if(program_exited()) {
change_state(STATE_STOPPED);
} else if(program_changed()) {
change_state(STATE_STOP_WAIT);
} else if(want_to_exit) {
change_state(STATE_STOP_WAIT);
}
break;
case STATE_STOP_WAIT:
send_stop_signal();
if(program_exited()) {
change_state(STATE_STOPPED);
} else if(time_in_state > stop_interval) {
change_state(STATE_KILL_WAIT);
}
break;
case STATE_KILL_WAIT:
send_kill_signal();
if(program_exited()) {
change_state(STATE_STOPPED);
}
break;
case STATE_STOPPED:
if(want_to_exit) {
debug(D_DEBUG, "all done");
exit(0);
} else {
unsigned int wait_time;
int i;
wait_time = min_wait_time;
for(i = 0; i < start_failures; i++) {
wait_time *= 2;
}
if(wait_time > max_wait_time || wait_time < min_wait_time) {
wait_time = max_wait_time;
}
if(time_in_state >= (int) wait_time) {
change_state(STATE_READY);
}
}
break;
default:
fatal("invalid state %d!\n", state);
break;
}
if(state_changes == 0) {
sleep(5);
} else {
state_changes = 0;
}
}
return 1;
}
static void work_queue_status_parse_command_line_arguments(int argc, char *argv[])
{
signed int c;
static struct option long_options[] = {
{"statistics", no_argument, 0, 'Q'},
{"workers", no_argument, 0, 'W'},
{"tasks", no_argument, 0, 'T'},
{"verbose", no_argument, 0, 'l'},
{"resources", no_argument, 0, 'R'},
{"catalog", required_argument, 0, 'C'},
{"debug", required_argument, 0, 'd'},
{"timeout", required_argument, 0, 't'},
{"help", no_argument, 0, 'h'},
{0,0,0,0}};
while((c = getopt_long(argc, argv, "QTWC:d:lo:O:Rt:vh", long_options, NULL)) > -1) {
switch (c) {
case 'C':
if(!parse_catalog_server_description(optarg, &catalog_host, &catalog_port)) {
fprintf(stderr, "Cannot parse catalog description: %s. \n", optarg);
exit(EXIT_FAILURE);
}
break;
case 'd':
debug_flags_set(optarg);
break;
case 'Q':
query_mode = QUERY_QUEUE;
break;
case 'T':
query_mode = QUERY_TASKS;
break;
case 'W':
query_mode = QUERY_WORKERS;
break;
case 'l':
format_mode = FORMAT_LONG;
break;
case 'o':
debug_config_file(optarg);
break;
case 'O':
debug_config_file_size(string_metric_parse(optarg));
break;
case 't':
work_queue_status_timeout = strtol(optarg, NULL, 10);
break;
case 'h':
show_help(argv[0]);
exit(EXIT_SUCCESS);
break;
case 'R':
query_mode = QUERY_MASTER_RESOURCES;
resource_mode = 1;
break;
case 'v':
cctools_version_print(stdout, argv[0]);
exit(EXIT_SUCCESS);
default:
show_help(argv[0]);
exit(EXIT_FAILURE);
break;
}
}
}
int main( int argc, char *argv[] )
{
char c;
const char *progname = "wavefront";
debug_config(progname);
while((c=getopt(argc,argv,"p:Pd:o:vh"))!=(char)-1) {
switch(c) {
case 'p':
port = atoi(optarg);
break;
case 'd':
debug_flags_set(optarg);
break;
case 'o':
debug_config_file(optarg);
break;
case 'v':
show_version(progname);
exit(0);
break;
case 'h':
show_help(progname);
exit(0);
break;
}
}
if( (argc-optind)!=5 ) {
show_help(progname);
exit(1);
}
function = argv[optind];
xsize=atoi(argv[optind+1]);
ysize=atoi(argv[optind+2]);
infile=argv[optind+3];
outfile=argv[optind+4];
start_time = time(0);
last_display_time = 0;
cells_total = xsize*ysize;
xsize++;
ysize++;
array = text_array_create(xsize,ysize);
if(!text_array_load(array,infile)) {
fprintf(stderr,"couldn't load %s: %s",infile,strerror(errno));
return 1;
}
int count = text_array_load(array,outfile);
if(count>0) printf("recovered %d results from %s\n",count,outfile);
logfile = fopen(outfile,"a");
if(!logfile) {
fprintf(stderr,"couldn't open %s for append: %s\n",outfile,strerror(errno));
return 1;
}
queue = work_queue_create(port);
task_prime();
struct work_queue_task *t;
while(1) {
if(time(0)!=last_display_time) display_progress(queue);
t = work_queue_wait(queue,WORK_QUEUE_WAITFORTASK);
if(!t) break;
if(t->return_status==0) {
int x,y;
if(sscanf(t->tag,"%d %d",&x,&y)==2) {
text_array_set(array,x,y,t->output);
task_complete(x,y);
fprintf(logfile,"%d %d %s\n",x,y,t->output);
fflush(logfile);
tasks_done++;
} else {
fprintf(stderr,"unexpected output: %s\nfrom command: %s\non host: %s",t->output,t->command_line,t->host);
}
} else {
fprintf(stderr,"function failed return value (%i) result (%i) on host %s. output:\n%s\n",t->return_status,t->result,t->host,t->output);
}
work_queue_task_delete(t);
if(work_queue_empty(queue))
break;
}
display_progress(queue);
return 0;
}
int main(int argc, char *argv[])
{
batch_queue_type_t batch_queue_type = BATCH_QUEUE_TYPE_UNKNOWN;
catalog_host = CATALOG_HOST;
catalog_port = CATALOG_PORT;
debug_config(argv[0]);
int c;
while((c = getopt_long(argc, argv, "F:N:M:T:t:w:W:E:P:S:cd:o:O:vh", long_options, NULL)) > -1) {
switch (c) {
case 'F':
foremen_regex = optarg;
break;
case 'N':
case 'M':
project_regex = optarg;
break;
case 'T':
batch_queue_type = batch_queue_type_from_string(optarg);
if(batch_queue_type == BATCH_QUEUE_TYPE_UNKNOWN) {
fprintf(stderr, "unknown batch queue type: %s\n", optarg);
return EXIT_FAILURE;
}
break;
case 't':
worker_timeout = atoi(optarg);
break;
case 'w':
workers_min = atoi(optarg);
break;
case 'W':
workers_max = atoi(optarg);
break;
case LONG_OPT_TASKS_PER_WORKER:
tasks_per_worker = atof(optarg);
break;
case 'E':
extra_worker_args = optarg;
break;
case LONG_OPT_CORES:
num_cores_option = xxstrdup(optarg);
break;
case LONG_OPT_MEMORY:
num_memory_option = xxstrdup(optarg);
break;
case LONG_OPT_DISK:
num_disk_option = xxstrdup(optarg);
break;
case LONG_OPT_GPUS:
num_gpus_option = xxstrdup(optarg);
break;
case 'P':
password_file = optarg;
break;
case 'S':
scratch_dir = optarg;
break;
case 'c':
consider_capacity = 1;
break;
case 'd':
debug_flags_set(optarg);
break;
case 'o':
debug_config_file(optarg);
break;
case 'O':
debug_config_file_size(string_metric_parse(optarg));
break;
case 'v':
cctools_version_print(stdout, argv[0]);
exit(EXIT_SUCCESS);
case 'h':
show_help(argv[0]);
exit(EXIT_SUCCESS);
default:
show_help(argv[0]);
return EXIT_FAILURE;
}
}
cctools_version_debug(D_DEBUG, argv[0]);
if(batch_queue_type == BATCH_QUEUE_TYPE_UNKNOWN) {
fprintf(stderr,"work_queue_pool: You must specify a batch type with the -T option.\n");
fprintf(stderr, "valid options:\n");
fprintf(stderr, "%s\n", batch_queue_type_string());
return 1;
}
if(!project_regex) {
fprintf(stderr,"work_queue_pool: You must give a project name with the -M option.\n");
return 1;
}
if(workers_min>workers_max) {
fprintf(stderr,"work_queue_pool: --min-workers (%d) is greater than --max-workers (%d)\n",workers_min,workers_max);
return 1;
}
if(tasks_per_worker < 1)
{
tasks_per_worker = num_cores_option ? atof(num_cores_option) : 1;
}
if(!scratch_dir) {
scratch_dir = string_format("/tmp/wq-pool-%d",getuid());
}
if(!create_dir(scratch_dir,0777)) {
fprintf(stderr,"work_queue_pool: couldn't create %s: %s",scratch_dir,strerror(errno));
return 1;
}
char cmd[1024];
sprintf(cmd,"cp \"$(which work_queue_worker)\" '%s'",scratch_dir);
if (system(cmd)) {
fprintf(stderr, "work_queue_pool: please add work_queue_worker to your PATH.\n");
exit(EXIT_FAILURE);
}
if(password_file) {
sprintf(cmd,"cp %s %s/pwfile",password_file,scratch_dir);
system(cmd);
}
if(chdir(scratch_dir)!=0) {
fprintf(stderr,"work_queue_pool: couldn't chdir to %s: %s",scratch_dir,strerror(errno));
return 1;
}
signal(SIGINT, handle_abort);
signal(SIGQUIT, handle_abort);
signal(SIGTERM, handle_abort);
signal(SIGHUP, ignore_signal);
struct batch_queue * queue = batch_queue_create(batch_queue_type);
if(!queue) {
fprintf(stderr,"work_queue_pool: couldn't establish queue type %s",batch_queue_type_to_string(batch_queue_type));
return 1;
}
set_worker_resources( queue );
mainloop( queue, project_regex, foremen_regex );
batch_queue_delete(queue);
return 0;
}
int main(int argc, char *argv[])
{
int port = WORK_QUEUE_DEFAULT_PORT;
const char *port_file=0;
const char *project_name=0;
int monitor_flag = 0;
int c;
while((c = getopt(argc, argv, "d:o:mN:p:Z:vh"))!=-1) {
switch (c) {
case 'd':
debug_flags_set(optarg);
break;
case 'o':
debug_config_file(optarg);
break;
case 'p':
port = atoi(optarg);
break;
case 'm':
monitor_flag = 1;
break;
case 'N':
project_name = optarg;
break;
case 'Z':
port_file = strdup(optarg);
port = 0;
break;
case 'v':
cctools_version_print(stdout, argv[0]);
return 0;
break;
case 'h':
show_help(path_basename(argv[0]));
return 0;
default:
show_help(path_basename(argv[0]));
return 1;
}
}
struct work_queue *q = work_queue_create(port);
if(!q) fatal("couldn't listen on any port!");
printf("listening on port %d...\n", work_queue_port(q));
if(port_file) {
FILE *file = fopen(port_file,"w");
if(!file) fatal("couldn't open %s: %s",port_file,strerror(errno));
fprintf(file,"%d\n",work_queue_port(q));
fclose(file);
}
if(project_name) {
work_queue_specify_name(q,project_name);
}
if(monitor_flag) {
unlink_recursive("work-queue-test-monitor");
work_queue_enable_monitoring(q, "work-queue-test-monitor");
work_queue_specify_category_mode(q, NULL, WORK_QUEUE_ALLOCATION_MODE_MAX_THROUGHPUT);
work_queue_specify_transactions_log(q, "work-queue-test-monitor/transactions.log");
}
int result = work_queue_mainloop(q);
work_queue_delete(q);
return result;
}
int main(int argc, char **argv)
{
signed char c;
struct work_queue *q;
int port = WORK_QUEUE_DEFAULT_PORT;
static const char *port_file = NULL;
int work_queue_master_mode = WORK_QUEUE_MASTER_MODE_STANDALONE;
char *project = NULL;
int priority = 0;
debug_config("allpairs_master");
extra_files_list = list_create();
struct option long_options[] = {
{"debug", required_argument, 0, 'd'},
{"help", no_argument, 0, 'h'},
{"version", no_argument, 0, 'v'},
{"port", required_argument, 0, 'p'},
{"random-port", required_argument, 0, 'Z'},
{"extra-args", required_argument, 0, 'e'},
{"width", required_argument, 0, 'x'},
{"height", required_argument, 0, 'y'},
{"advertise", no_argument, 0, 'a'}, //deprecated, left here for backwards compatibility
{"project-name", required_argument, 0, 'N'},
{"debug-file", required_argument, 0, 'o'},
{"output-file", required_argument, 0, 'O'},
{"wqstats-file", required_argument, 0, 's'},
{"input-file", required_argument, 0, 'f'},
{"estimated-time", required_argument, 0, 't'},
{"priority", required_argument, 0, 'P'},
{0,0,0,0}
};
while((c = getopt_long(argc, argv, "ad:e:f:hN:p:P:t:vx:y:Z:O:o:s:", long_options, NULL)) >= 0) {
switch (c) {
case 'a':
work_queue_master_mode = WORK_QUEUE_MASTER_MODE_CATALOG;
break;
case 'd':
debug_flags_set(optarg);
break;
case 'e':
extra_arguments = optarg;
break;
case 'f':
list_push_head(extra_files_list,optarg);
break;
case 'o':
debug_config_file(optarg);
break;
case 'O':
free(output_filename);
output_filename=xxstrdup(optarg);
break;
case 's':
free(wqstats_filename);
wqstats_filename=xxstrdup(optarg);
break;
case 'h':
show_help(progname);
exit(0);
break;
case 'N':
work_queue_master_mode = WORK_QUEUE_MASTER_MODE_CATALOG;
free(project);
project = xxstrdup(optarg);
break;
case 'p':
port = atoi(optarg);
break;
case 'P':
priority = atoi(optarg);
break;
case 't':
compare_program_time = atof(optarg);
break;
case 'v':
cctools_version_print(stdout, progname);
exit(0);
break;
case 'x':
xblock = atoi(optarg);
break;
case 'y':
yblock = atoi(optarg);
break;
case 'Z':
port_file = optarg;
port = 0;
break;
default:
show_help(progname);
return 1;
}
}
cctools_version_debug(D_DEBUG, argv[0]);
if((argc - optind) < 3) {
show_help(progname);
exit(1);
}
struct text_list *seta = text_list_load(argv[optind]);
if(!seta) {
fprintf(stderr,"%s: couldn't open %s: %s\n",progname,argv[optind+1],strerror(errno));
return 1;
}
fprintf(stdout, "%s: %s has %d elements\n",progname,argv[optind],text_list_size(seta));
struct text_list *setb = text_list_load(argv[optind+1]);
if(!setb) {
fprintf(stderr,"%s: couldn't open %s: %s\n",progname,argv[optind+1],strerror(errno));
return 1;
}
fprintf(stdout, "%s: %s has %d elements\n",progname,argv[optind+1],text_list_size(setb));
if (!find_executable("allpairs_multicore","PATH",allpairs_multicore_program,sizeof(allpairs_multicore_program))) {
fprintf(stderr,"%s: couldn't find allpairs_multicore in path\n",progname);
return 1;
}
debug(D_DEBUG,"using multicore executable %s",allpairs_multicore_program);
xstop = text_list_size(seta);
ystop = text_list_size(setb);
if(allpairs_compare_function_get(argv[optind+2])) {
strcpy(allpairs_compare_program,argv[optind+2]);
debug(D_DEBUG,"using internal function %s",allpairs_compare_program);
use_external_program = 0;
} else {
if(!find_executable(argv[optind+2],"PATH",allpairs_compare_program,sizeof(allpairs_compare_program))) {
fprintf(stderr,"%s: %s is neither an executable nor an internal comparison function.\n",progname,allpairs_compare_program);
return 1;
}
debug(D_DEBUG,"using comparison executable %s",allpairs_compare_program);
use_external_program = 1;
}
if(!xblock || !yblock) {
estimate_block_size(seta,setb,&xblock,&yblock);
}
fprintf(stdout, "%s: using block size of %dx%d\n",progname,xblock,yblock);
if(work_queue_master_mode == WORK_QUEUE_MASTER_MODE_CATALOG && !project) {
fprintf(stderr, "allpairs: allpairs master running in catalog mode. Please use '-N' option to specify the name of this project.\n");
fprintf(stderr, "allpairs: Run \"%s -h\" for help with options.\n", argv[0]);
return 1;
}
q = work_queue_create(port);
//Read the port the queue is actually running, in case we just called
//work_queue_create(LINK_PORT_ANY)
port = work_queue_port(q);
if(!q) {
fprintf(stderr,"%s: could not create work queue on port %d: %s\n",progname,port,strerror(errno));
return 1;
}
if(port_file)
opts_write_port_file(port_file, port);
if(wqstats_filename)
work_queue_specify_log(q, wqstats_filename);
// advanced work queue options
work_queue_specify_master_mode(q, work_queue_master_mode);
work_queue_specify_name(q, project);
work_queue_specify_priority(q, priority);
fprintf(stdout, "%s: listening for workers on port %d...\n",progname,work_queue_port(q));
while(1) {
struct work_queue_task *task = NULL;
while(work_queue_hungry(q)) {
task = ap_task_create(seta,setb);
if(task) {
work_queue_submit(q, task);
} else {
break;
}
}
if(!task && work_queue_empty(q)) break;
task = work_queue_wait(q,5);
if(task) task_complete(task);
}
work_queue_delete(q);
return 0;
}
int main(int argc, char *argv[])
{
const char *host = NULL;
int port = MPI_QUEUE_DEFAULT_PORT;
char addr[LINK_ADDRESS_MAX];
char c;
int w, rank;
signal(SIGTERM, handle_abort);
signal(SIGQUIT, handle_abort);
signal(SIGINT, handle_abort);
MPI_Init(&argc, &argv);
debug_config(argv[0]);
while((c = getopt(argc, argv, "d:ho:t:w:v")) != (char) -1) {
switch (c) {
case 'd':
debug_flags_set(optarg);
break;
case 't':
idle_timeout = string_time_parse(optarg);
break;
case 'o':
debug_config_file(optarg);
break;
case 'v':
show_version(argv[0]);
return 0;
case 'w':
w = string_metric_parse(optarg);
link_window_set(w, w);
break;
case 'h':
default:
show_help(argv[0]);
return 1;
}
}
if ((argc - optind) != 2) {
show_help(argv[0]);
return 1;
}
host = argv[optind];
port = atoi(argv[optind + 1]);
if(!domain_name_cache_lookup(host, addr)) {
fprintf(stderr, "couldn't lookup address of host %s\n", host);
exit(1);
}
MPI_Comm_rank(MPI_COMM_WORLD, &rank);
if(rank) {
return worker_main();
} else {
return master_main(host, port, addr);
}
}
int main(int argc, char *argv[])
{
char *temp;
int did_explicit_auth = 0;
char *tickets = NULL;
char prompt[CHIRP_LINE_MAX];
char line[CHIRP_LINE_MAX];
char **user_argv = 0;
int user_argc;
signed char c;
int result = 0;
debug_config(argv[0]);
static struct option long_options[] = {
{"auth", required_argument, 0, 'a'},
{"debug", required_argument, 0, 'd'},
{"debug-file", required_argument, 0, 'o'},
{"tickets", required_argument, 0, 'i'},
{"verbose", no_argument, 0, 'l'},
{"timeout", required_argument, 0, 't'},
{"version", no_argument, 0, 'v'},
{"help", no_argument, 0, 'h'},
{0, 0, 0, 0}
};
while((c = getopt_long(argc, argv, "+a:d:hi:lt:v", long_options, NULL)) > -1) {
switch (c) {
case 'a':
auth_register_byname(optarg);
did_explicit_auth = 1;
break;
case 'd':
debug_flags_set(optarg);
break;
case 'o':
debug_config_file(optarg);
break;
case 'h':
show_help(argv[0]);
exit(0);
break;
case 'i':
tickets = strdup(optarg);
break;
case 'l':
long_information = 1;
break;
case 't':
timeout = string_time_parse(optarg);
break;
case 'v':
cctools_version_print(stdout, argv[0]);
exit(0);
break;
}
}
cctools_version_debug(D_DEBUG, argv[0]);
if(!did_explicit_auth)
auth_register_all();
if(tickets) {
auth_ticket_load(tickets);
free(tickets);
} else if(getenv(CHIRP_CLIENT_TICKETS)) {
auth_ticket_load(getenv(CHIRP_CLIENT_TICKETS));
} else {
auth_ticket_load(NULL);
}
getcwd(current_local_dir, CHIRP_PATH_MAX);
/* interactive mode if input is a TTY but we are not simply executing a
* command from argv */
interactive_mode = isatty(0) && !((argc - optind) > 1);
if(optind < argc) {
stoptime = time(0) + timeout;
if(do_open(1, &argv[optind - 1])) {
fprintf(stderr, "couldn't open %s: %s\n", argv[optind], strerror(errno));
return 1;
}
}
if((argc - optind) > 1) {
return !process_command(argc - optind - 1, &argv[optind + 1]);
}
while(1) {
if(interactive_mode) {
sprintf(prompt, " chirp:%s:%s> ", current_host, current_remote_dir);
} else {
prompt[0] = 0;
}
#ifdef HAS_LIBREADLINE
temp = readline(prompt);
if(!temp)
break;
strcpy(line, temp);
free(temp);
#else
printf("%s", prompt);
fflush(stdout);
if(!fgets(line, CHIRP_LINE_MAX, stdin))
break;
#endif
if(!line[0])
continue;
if(!interactive_mode && (temp = strchr(line, '#'))) { /* comment? */
for(temp--; temp > line && isspace((int) *temp); temp--); /* preceding space? */
if(temp <= line)
continue; /* else not comment */
}
#ifdef HAS_LIBREADLINE
add_history(line);
#endif
{
char *start = line, *last = strlen(line) + line;
while(*start != '\0') { /* process compound commands */
char *end = strchr(start, ';');
while(end != NULL && end != start && *(end - 1) == '\\')
end = strchr(end + 1, ';');
if(end == NULL)
end = start + strlen(start);
*end = '\0';
if(user_argv)
free(user_argv);
string_split(start, &user_argc, &user_argv);
if(user_argc == 0) {
start++;
continue;
}
result = process_command(user_argc, user_argv);
start = end == last ? last : end + 1;
}
}
if(!interactive_mode && !result)
break;
}
if(result) {
return 0;
} else {
return 1;
}
}
int main(int argc, char *argv[])
{
int port = WORK_QUEUE_DEFAULT_PORT;
const char *port_file=0;
const char *project_name=0;
char c;
while((c = getopt(argc, argv, "d:o:N:p:Z:vh"))!=-1) {
switch (c) {
case 'd':
debug_flags_set(optarg);
break;
case 'o':
debug_config_file(optarg);
break;
case 'p':
port = atoi(optarg);
break;
case 'N':
project_name = optarg;
break;
case 'Z':
port_file = strdup(optarg);
port = 0;
break;
case 'v':
cctools_version_print(stdout, argv[0]);
return 0;
break;
case 'h':
show_help(path_basename(argv[0]));
return 0;
default:
show_help(path_basename(argv[0]));
return 1;
}
}
struct work_queue *q = work_queue_create(port);
if(!q) fatal("couldn't listen on any port!");
printf("listening on port %d...\n", work_queue_port(q));
if(port_file) {
FILE *file = fopen(port_file,"w");
if(!file) fatal("couldn't open %s: %s",port_file,strerror(errno));
fprintf(file,"%d\n",work_queue_port(q));
fclose(file);
}
if(project_name) {
work_queue_specify_name(q,project_name);
}
int result = work_queue_mainloop(q);
work_queue_delete(q);
return result;
}
int main( int argc, char *argv[] )
{
signed char c;
const char *progname = "wavefront";
debug_config(progname);
progress_log_file = stdout;
struct option long_options[] = {
{"help", no_argument, 0, 'h'},
{"version", no_argument, 0, 'v'},
{"debug", required_argument, 0, 'd'},
{"jobs", required_argument, 0, 'n'},
{"block-size", required_argument, 0, 'b'},
{"debug-file", required_argument, 0, 'o'},
{"log-file", required_argument, 0, 'l'},
{"bitmap", required_argument, 0, 'B'},
{"bitmap-interval", required_argument, 0, 'i'},
{"auto", no_argument, 0, 'A'},
{"local", no_argument, 0, 'L'},
{"batch-type", required_argument, 0, 'T'},
{"verify", no_argument, 0, 'V'},
{0,0,0,0}
};
while((c=getopt_long(argc,argv,"n:b:d:o:l:B:i:qALDT:VX:Y:vh", long_options, NULL)) > -1) {
switch(c) {
case 'n':
manual_max_jobs_running = atoi(optarg);
break;
case 'b':
manual_block_size = atoi(optarg);
break;
case 'd':
debug_flags_set(optarg);
break;
case 'o':
debug_config_file(optarg);
break;
case 'B':
progress_bitmap_file = optarg;
break;
case 'i':
progress_bitmap_interval = atoi(optarg);
break;
case 'l':
progress_log_file = fopen(optarg,"w");
if(!progress_log_file) {
fprintf(stderr,"couldn't open %s: %s\n",optarg,strerror(errno));
return 1;
}
break;
case 'A':
wavefront_mode = WAVEFRONT_MODE_AUTO;
break;
case 'L':
wavefront_mode = WAVEFRONT_MODE_MULTICORE;
break;
case 'T':
wavefront_mode = WAVEFRONT_MODE_DISTRIBUTED;
batch_system_type = batch_queue_type_from_string(optarg);
if(batch_system_type==BATCH_QUEUE_TYPE_UNKNOWN) {
fprintf(stderr,"unknown batch system type: %s\n",optarg);
exit(1);
}
break;
case 'V':
verify_mode = 1;
break;
case 'X':
xstart = atoi(optarg);
break;
case 'Y':
ystart = atoi(optarg);
break;
case 'v':
cctools_version_print(stdout, progname);
exit(0);
break;
case 'h':
show_help(progname);
exit(0);
break;
}
}
cctools_version_debug(D_DEBUG, argv[0]);
if( (argc-optind<3) ) {
show_help(progname);
exit(1);
}
function = argv[optind];
xsize=atoi(argv[optind+1]);
ysize=atoi(argv[optind+2]);
total_cells = xsize*ysize;
if(!verify_mode && !check_configuration(function,xsize,ysize)) exit(1);
int ncpus = load_average_get_cpus();
if(wavefront_mode!=WAVEFRONT_MODE_MULTICORE) {
double task_time = measure_task_time();
printf("Each function takes %.02lfs to run.\n",task_time);
block_size = find_best_block_size(xsize,1000,2,task_time,average_dispatch_time);
double distributed_time = wavefront_distributed_model(xsize,1000,2,task_time,block_size,average_dispatch_time);
double multicore_time = wavefront_multicore_model(xsize,ncpus,task_time);
double ideal_multicore_time = wavefront_multicore_model(xsize,xsize,task_time);
double sequential_time = wavefront_multicore_model(xsize,1,task_time);
printf("---------------------------------\n");
printf("This workload would take:\n");
printf("%.02lfs sequentially\n",sequential_time);
printf("%.02lfs on this %d-core machine\n",multicore_time,ncpus);
printf("%.02lfs on a %d-core machine\n",ideal_multicore_time,xsize);
printf("%.02lfs on a 1000-node distributed system with block size %d\n",distributed_time,block_size);
printf("---------------------------------\n");
if(wavefront_mode==WAVEFRONT_MODE_AUTO) {
if(multicore_time < distributed_time*2) {
wavefront_mode = WAVEFRONT_MODE_MULTICORE;
} else {
wavefront_mode = WAVEFRONT_MODE_DISTRIBUTED;
}
}
}
if(wavefront_mode==WAVEFRONT_MODE_MULTICORE) {
batch_system_type = BATCH_QUEUE_TYPE_LOCAL;
max_jobs_running = ncpus;
} else {
max_jobs_running = 1000;
}
if(manual_block_size!=0) {
block_size = manual_block_size;
}
if(manual_max_jobs_running!=0) {
max_jobs_running = manual_max_jobs_running;
}
if(wavefront_mode==WAVEFRONT_MODE_MULTICORE) {
printf("Running in multicore mode with %d CPUs.\n",max_jobs_running);
} else {
printf("Running in distributed mode with block size %d on up to %d CPUs\n",block_size,max_jobs_running);
}
batch_q = batch_queue_create(batch_system_type);
if(verify_mode) exit(0);
struct bitmap * b = bitmap_create(xsize+1,ysize+1);
struct list *ready_list = list_create();
struct itable *running_table = itable_create(0);
struct batch_job_info info;
UINT64_T jobid;
struct wavefront_task *task;
wavefront_task_initialize(b,ready_list);
printf("Starting workload...\n");
fprintf(progress_log_file,"# elapsed time : waiting jobs / running jobs / cells complete (percent complete)\n");
while(1) {
if(abort_mode) {
while((task=list_pop_tail(ready_list))) {
wavefront_task_delete(task);
}
itable_firstkey(running_table);
while(itable_nextkey(running_table,&jobid,(void**)&task)) {
batch_job_remove(batch_q,jobid);
}
}
if(list_size(ready_list)==0 && itable_size(running_table)==0) break;
while(1) {
if(itable_size(running_table)>=max_jobs_running) break;
task = list_pop_tail(ready_list);
if(!task) break;
jobid = wavefront_task_submit(task);
if(jobid>0) {
itable_insert(running_table,jobid,task);
wavefront_task_mark_range(task,b,WAVEFRONT_TASK_STATE_RUNNING);
} else {
abort();
sleep(1);
list_push_head(ready_list,task);
}
}
save_status(b,ready_list,running_table);
jobid = batch_job_wait(batch_q,&info);
if(jobid>0) {
task = itable_remove(running_table,jobid);
if(task) {
if(info.exited_normally && info.exit_code==0) {
total_dispatch_time += info.started-info.submitted;
total_execute_time += MAX(info.finished-info.started,1);
total_cells_complete+=task->width*task->height;
total_jobs_complete++;
average_dispatch_time = 1.0*total_dispatch_time / total_jobs_complete;
average_task_time = 1.0*total_execute_time / total_cells_complete;
wavefront_task_complete(b,ready_list,task);
} else {
printf("job %" PRIu64 " failed, aborting this workload\n",jobid);
abort_mode = 1;
}
}
}
}
save_status(b,ready_list,running_table);
if(abort_mode) {
printf("Workload was aborted.\n");
} else {
printf("Workload complete.\n");
}
return 0;
}
int main(int argc, char *argv[])
{
struct link *link, *list_port = 0;
signed char ch;
time_t current;
int is_daemon = 0;
char *pidfile = NULL;
char *interface = NULL;
outgoing_host_list = list_create();
debug_config(argv[0]);
static const struct option long_options[] = {
{"background", no_argument, 0, 'b'},
{"pid-file", required_argument, 0, 'B'},
{"debug", required_argument, 0, 'd'},
{"help", no_argument, 0, 'h'},
{"history", required_argument, 0, 'H'},
{"lifetime", required_argument, 0, 'l'},
{"update-log", required_argument, 0, 'L'},
{"max-jobs", required_argument, 0, 'm'},
{"server-size", required_argument, 0, 'M'},
{"name", required_argument, 0, 'n'},
{"interface", required_argument, 0, 'I'},
{"debug-file", required_argument, 0, 'o'},
{"debug-rotate-max", required_argument, 0, 'O'},
{"port", required_argument, 0, 'p'},
{"single", no_argument, 0, 'S'},
{"timeout", required_argument, 0, 'T'},
{"update-host", required_argument, 0, 'u'},
{"update-interval", required_argument, 0, 'U'},
{"version", no_argument, 0, 'v'},
{"port-file", required_argument, 0, 'Z'},
{0,0,0,0}};
while((ch = getopt_long(argc, argv, "bB:d:hH:I:l:L:m:M:n:o:O:p:ST:u:U:vZ:", long_options, NULL)) > -1) {
switch (ch) {
case 'b':
is_daemon = 1;
break;
case 'B':
free(pidfile);
pidfile = strdup(optarg);
break;
case 'd':
debug_flags_set(optarg);
break;
case 'h':
default:
show_help(argv[0]);
return 1;
case 'l':
lifetime = string_time_parse(optarg);
break;
case 'L':
logfilename = strdup(optarg);
break;
case 'H':
history_dir = strdup(optarg);
break;
case 'I':
free(interface);
interface = strdup(optarg);
break;
case 'm':
child_procs_max = atoi(optarg);
break;
case 'M':
max_server_size = string_metric_parse(optarg);
break;
case 'n':
preferred_hostname = optarg;
break;
case 'o':
debug_config_file(optarg);
break;
case 'O':
debug_config_file_size(string_metric_parse(optarg));
break;
case 'p':
port = atoi(optarg);
break;
case 'S':
fork_mode = 0;
break;
case 'T':
child_procs_timeout = string_time_parse(optarg);
break;
case 'u':
list_push_head(outgoing_host_list, xxstrdup(optarg));
break;
case 'U':
outgoing_timeout = string_time_parse(optarg);
break;
case 'v':
cctools_version_print(stdout, argv[0]);
return 0;
case 'Z':
port_file = optarg;
port = 0;
break;
}
}
if (is_daemon) daemonize(0, pidfile);
cctools_version_debug(D_DEBUG, argv[0]);
if(logfilename) {
logfile = fopen(logfilename,"a");
if(!logfile) fatal("couldn't open %s: %s\n",optarg,strerror(errno));
}
current = time(0);
debug(D_NOTICE, "*** %s starting at %s", argv[0], ctime(¤t));
if(!list_size(outgoing_host_list)) {
list_push_head(outgoing_host_list, CATALOG_HOST_DEFAULT);
}
install_handler(SIGPIPE, ignore_signal);
install_handler(SIGHUP, ignore_signal);
install_handler(SIGCHLD, ignore_signal);
install_handler(SIGINT, shutdown_clean);
install_handler(SIGTERM, shutdown_clean);
install_handler(SIGQUIT, shutdown_clean);
install_handler(SIGALRM, shutdown_clean);
if(!preferred_hostname) {
domain_name_cache_guess(hostname);
preferred_hostname = hostname;
}
username_get(owner);
starttime = time(0);
table = nvpair_database_create(history_dir);
if(!table)
fatal("couldn't create directory %s: %s\n",history_dir,strerror(errno));
list_port = link_serve_address(interface, port);
if(list_port) {
/*
If a port was chosen automatically, read it back
so that the same one can be used for the update port.
There is the possibility that the UDP listen will
fail because that port is in use.
*/
if(port==0) {
char addr[LINK_ADDRESS_MAX];
link_address_local(list_port,addr,&port);
}
} else {
if(interface)
fatal("couldn't listen on TCP address %s port %d", interface, port);
else
fatal("couldn't listen on TCP port %d", port);
}
outgoing_dgram = datagram_create(0);
if(!outgoing_dgram)
fatal("couldn't create outgoing udp port");
update_dgram = datagram_create_address(interface, port);
if(!update_dgram) {
if(interface)
fatal("couldn't listen on UDP address %s port %d", interface, port);
else
fatal("couldn't listen on UDP port %d", port);
}
opts_write_port_file(port_file,port);
while(1) {
fd_set rfds;
int ufd = datagram_fd(update_dgram);
int lfd = link_fd(list_port);
int result, maxfd;
struct timeval timeout;
remove_expired_records();
if(time(0) > outgoing_alarm) {
update_all_catalogs(outgoing_dgram);
outgoing_alarm = time(0) + outgoing_timeout;
}
while(1) {
int status;
pid_t pid = waitpid(-1, &status, WNOHANG);
if(pid>0) {
child_procs_count--;
continue;
} else {
break;
}
}
FD_ZERO(&rfds);
FD_SET(ufd, &rfds);
if(child_procs_count < child_procs_max) {
FD_SET(lfd, &rfds);
}
maxfd = MAX(ufd, lfd) + 1;
timeout.tv_sec = 5;
timeout.tv_usec = 0;
result = select(maxfd, &rfds, 0, 0, &timeout);
if(result <= 0)
continue;
if(FD_ISSET(ufd, &rfds)) {
handle_updates(update_dgram);
}
if(FD_ISSET(lfd, &rfds)) {
link = link_accept(list_port, time(0) + 5);
if(link) {
if(fork_mode) {
pid_t pid = fork();
if(pid == 0) {
alarm(child_procs_timeout);
handle_query(link);
_exit(0);
} else if (pid>0) {
child_procs_count++;
}
} else {
handle_query(link);
}
link_close(link);
}
}
}
return 1;
}
int main(int argc, char *argv[])
{
struct catalog_query *q;
struct nvpair *n;
time_t timeout = 60, stoptime;
const char *catalog_host = 0;
signed char c;
int i;
int count = 0;
int mode = MODE_TABLE;
INT64_T total = 0, avail = 0;
const char *filter_name = 0;
const char *filter_value = 0;
debug_config(argv[0]);
static struct option long_options[] = {
{"catalog", required_argument, 0, 'c'},
{"debug", required_argument, 0, 'd'},
{"debug-file", required_argument, 0, 'o'},
{"debug-rotate-max", required_argument, 0, 'O'},
{"server-space", required_argument, 0, 'A'},
{"all", no_argument, 0, 'a'},
{"timeout", required_argument, 0, 't'},
{"brief", no_argument, 0, 's'},
{"verbose", no_argument, 0, 'l'},
{"totals", no_argument, 0, 'T'},
{"version", no_argument, 0, 'v'},
{"help", no_argument, 0, 'h'},
{0, 0, 0, 0}
};
while((c = getopt_long(argc, argv, "aA:c:d:t:o:O:sTlvh", long_options, NULL)) > -1) {
switch (c) {
case 'a':
show_all_types = 1;
break;
case 'c':
catalog_host = optarg;
break;
case 'd':
debug_flags_set(optarg);
break;
case 't':
timeout = string_time_parse(optarg);
break;
case 'A':
minavail = string_metric_parse(optarg);
break;
case 'o':
debug_config_file(optarg);
break;
case 'O':
debug_config_file_size(string_metric_parse(optarg));
break;
case 'v':
cctools_version_print(stdout, argv[0]);
return 1;
case 's':
mode = MODE_SHORT;
break;
case 'l':
mode = MODE_LONG;
break;
case 'T':
mode = MODE_TOTAL;
break;
case 'h':
default:
show_help(argv[0]);
return 1;
}
}
cctools_version_debug(D_DEBUG, argv[0]);
if(argc - optind == 0) {
// fine, keep going
} else if((argc - optind) == 1) {
filter_name = "name";
filter_value = argv[optind];
} else if((argc - optind) == 2) {
filter_name = argv[optind];
filter_value = argv[optind + 1];
} else {
show_help(argv[0]);
return 1;
}
stoptime = time(0) + timeout;
q = catalog_query_create(catalog_host, 0, stoptime);
if(!q) {
fprintf(stderr, "couldn't query catalog: %s\n", strerror(errno));
return 1;
}
if(mode == MODE_TABLE) {
nvpair_print_table_header(stdout, headers);
}
while((n = catalog_query_read(q, stoptime))) {
table[count++] = n;
}
qsort(table, count, sizeof(*table), (int (*)(const void *, const void *)) compare_entries);
for(i = 0; i < count; i++) {
const char *etype = nvpair_lookup_string(table[i], "type");
if(!show_all_types) {
if(etype) {
if(!strcmp(etype, "chirp") || !strcmp(etype, "catalog")) {
/* ok, keep going */
} else {
continue;
}
} else {
continue;
}
}
if(minavail != 0) {
if(minavail > nvpair_lookup_integer(table[i], "avail")) {
continue;
}
}
if(filter_name) {
const char *v = nvpair_lookup_string(table[i], filter_name);
if(!v || strcmp(filter_value, v))
continue;
}
if(mode == MODE_SHORT) {
const char *t = nvpair_lookup_string(table[i], "type");
if(t && !strcmp(t, "chirp")) {
printf("%s:%d\n", nvpair_lookup_string(table[i], "name"), (int) nvpair_lookup_integer(table[i], "port"));
}
} else if(mode == MODE_LONG) {
nvpair_print_text(table[i], stdout);
} else if(mode == MODE_TABLE) {
nvpair_print_table(table[i], stdout, headers);
} else if(mode == MODE_TOTAL) {
avail += nvpair_lookup_integer(table[i], "avail");
total += nvpair_lookup_integer(table[i], "total");
}
}
if(mode == MODE_TOTAL) {
printf("NODES: %4d\n", count);
printf("TOTAL: %6sB\n", string_metric(total, -1, 0));
printf("AVAIL: %6sB\n", string_metric(avail, -1, 0));
printf("INUSE: %6sB\n", string_metric(total - avail, -1, 0));
}
if(mode == MODE_TABLE) {
nvpair_print_table_footer(stdout, headers);
}
return 0;
}
int main(int argc, char *argv[])
{
enum {
LONGOPT_SERVER_LASTHEARDFROM = INT_MAX-0,
LONGOPT_SERVER_PROJECT = INT_MAX-1,
LONGOPT_WHERE = INT_MAX-2,
};
static const struct option long_options[] = {
{"all", no_argument, 0, 'a'},
{"brief", no_argument, 0, 's'},
{"catalog", required_argument, 0, 'c'},
{"debug", required_argument, 0, 'd'},
{"debug-file", required_argument, 0, 'o'},
{"debug-rotate-max", required_argument, 0, 'O'},
{"help", no_argument, 0, 'h'},
{"server-lastheardfrom", required_argument, 0, LONGOPT_SERVER_LASTHEARDFROM},
{"server-project", required_argument, 0, LONGOPT_SERVER_PROJECT},
{"server-space", required_argument, 0, 'A'},
{"timeout", required_argument, 0, 't'},
{"totals", no_argument, 0, 'T'},
{"verbose", no_argument, 0, 'l'},
{"version", no_argument, 0, 'v'},
{"where", required_argument, 0, LONGOPT_WHERE },
{0, 0, 0, 0}
};
struct catalog_query *q;
struct jx *j;
time_t timeout = 60, stoptime;
const char *catalog_host = 0;
int i;
int c;
int count = 0;
int mode = MODE_TABLE;
INT64_T sum_total = 0, sum_avail = 0;
const char *filter_name = 0;
const char *filter_value = 0;
const char *where_expr = "true";
int show_all_types = 0;
const char *server_project = NULL;
time_t server_lastheardfrom = 0;
uint64_t server_avail = 0;
debug_config(argv[0]);
while((c = getopt_long(argc, argv, "aA:c:d:t:o:O:sTlvh", long_options, NULL)) > -1) {
switch (c) {
case 'a':
show_all_types = 1;
break;
case 'c':
catalog_host = optarg;
break;
case 'd':
debug_flags_set(optarg);
break;
case 't':
timeout = string_time_parse(optarg);
break;
case 'A':
server_avail = string_metric_parse(optarg);
break;
case 'o':
debug_config_file(optarg);
break;
case 'O':
debug_config_file_size(string_metric_parse(optarg));
break;
case 'v':
cctools_version_print(stdout, argv[0]);
return 1;
case 's':
mode = MODE_SHORT;
break;
case 'l':
mode = MODE_LONG;
break;
case 'T':
mode = MODE_TOTAL;
break;
case LONGOPT_SERVER_LASTHEARDFROM:
server_lastheardfrom = time(0)-string_time_parse(optarg);
break;
case LONGOPT_SERVER_PROJECT:
server_project = xxstrdup(optarg);
break;
case LONGOPT_WHERE:
where_expr = optarg;
break;
case 'h':
default:
show_help(argv[0]);
return 1;
}
}
cctools_version_debug(D_DEBUG, argv[0]);
if(argc - optind == 0) {
// fine, keep going
} else if((argc - optind) == 1) {
filter_name = "name";
filter_value = argv[optind];
} else if((argc - optind) == 2) {
filter_name = argv[optind];
filter_value = argv[optind + 1];
} else {
show_help(argv[0]);
return 1;
}
stoptime = time(0) + timeout;
const char *query_expr;
if(show_all_types) {
query_expr = where_expr;
} else {
query_expr = string_format("%s && (type==\"chirp\" || type==\"catalog\")",where_expr);
}
struct jx *jexpr = jx_parse_string(query_expr);
if(!jexpr) {
fprintf(stderr,"invalid expression: %s\n",query_expr);
return 1;
}
q = catalog_query_create(catalog_host, 0, jexpr, stoptime);
if(!q) {
fprintf(stderr, "couldn't query catalog: %s\n", strerror(errno));
return 1;
}
if(mode == MODE_TABLE) {
jx_table_print_header(headers,stdout);
} else if(mode==MODE_LONG) {
printf("[\n");
}
while((j = catalog_query_read(q, stoptime))) {
table[count++] = j;
}
catalog_query_delete(q);
qsort(table, count, sizeof(*table), (int (*)(const void *, const void *)) compare_entries);
for(i = 0; i < count; i++) {
const char *lastheardfrom = jx_lookup_string(table[i], "lastheardfrom");
if (lastheardfrom && (time_t)strtoul(lastheardfrom, NULL, 10) < server_lastheardfrom)
continue;
const char *avail = jx_lookup_string(table[i], "avail");
if (avail && strtoul(avail, NULL, 10) < server_avail)
continue;
const char *project = jx_lookup_string(table[i], "project");
if (server_project && (project == NULL || !(strcmp(project, server_project) == 0)))
continue;
if(filter_name) {
const char *v = jx_lookup_string(table[i], filter_name);
if(!v || strcmp(filter_value, v))
continue;
}
if(mode == MODE_SHORT) {
const char *t = jx_lookup_string(table[i], "type");
if(t && !strcmp(t, "chirp")) {
printf("%s:%d\n", jx_lookup_string(table[i], "name"), (int) jx_lookup_integer(table[i], "port"));
}
} else if(mode == MODE_LONG) {
if(i!=0) printf(",\n");
jx_print_stream(table[i],stdout);
} else if(mode == MODE_TABLE) {
jx_table_print(headers, table[i], stdout);
} else if(mode == MODE_TOTAL) {
sum_avail += jx_lookup_integer(table[i], "avail");
sum_total += jx_lookup_integer(table[i], "total");
}
}
for(i=0;i<count;i++) {
jx_delete(table[i]);
}
if(mode == MODE_TOTAL) {
printf("NODES: %4d\n", count);
printf("TOTAL: %6sB\n", string_metric(sum_total, -1, 0));
printf("AVAIL: %6sB\n", string_metric(sum_avail, -1, 0));
printf("INUSE: %6sB\n", string_metric(sum_total - sum_avail, -1, 0));
} else if(mode == MODE_TABLE) {
jx_table_print_footer(headers,stdout);
} else if(mode==MODE_LONG) {
printf("\n]\n");
}
return 0;
}
int main(int argc, char *argv[])
{
batch_queue_type_t batch_queue_type = BATCH_QUEUE_TYPE_UNKNOWN;
catalog_host = CATALOG_HOST;
catalog_port = CATALOG_PORT;
batch_submit_options = getenv("BATCH_OPTIONS");
debug_config(argv[0]);
resources = rmsummary_create(-1);
int c;
while((c = getopt_long(argc, argv, "B:C:F:N:M:T:t:w:W:E:P:S:cd:o:O:vh", long_options, NULL)) > -1) {
switch (c) {
case 'B':
batch_submit_options = xxstrdup(optarg);
break;
case 'C':
config_file = xxstrdup(optarg);
break;
case 'F':
foremen_regex = xxstrdup(optarg);
break;
case 'N':
case 'M':
project_regex = xxstrdup(optarg);
break;
case 'T':
batch_queue_type = batch_queue_type_from_string(optarg);
if(batch_queue_type == BATCH_QUEUE_TYPE_UNKNOWN) {
fprintf(stderr, "unknown batch queue type: %s\n", optarg);
return EXIT_FAILURE;
}
break;
case 't':
worker_timeout = atoi(optarg);
break;
case 'w':
workers_min = atoi(optarg);
break;
case 'W':
workers_max = atoi(optarg);
break;
case LONG_OPT_WORKERS_PER_CYCLE:
workers_per_cycle = atoi(optarg);
break;
case LONG_OPT_TASKS_PER_WORKER:
tasks_per_worker = atof(optarg);
break;
case 'E':
extra_worker_args = xxstrdup(optarg);
break;
case LONG_OPT_CORES:
resources->cores = atoi(optarg);
break;
case LONG_OPT_AMAZON_CREDENTIALS:
amazon_credentials = xxstrdup(optarg);
break;
case LONG_OPT_AMAZON_AMI:
amazon_ami = xxstrdup(optarg);
break;
case LONG_OPT_MEMORY:
resources->memory = atoi(optarg);
break;
case LONG_OPT_DISK:
resources->disk = atoi(optarg);
break;
case LONG_OPT_GPUS:
resources->gpus = atoi(optarg);
break;
case LONG_OPT_AUTOSIZE:
autosize = 1;
break;
case LONG_OPT_FACTORY_TIMEOUT:
factory_timeout = MAX(0, atoi(optarg));
break;
case LONG_OPT_CONDOR_REQUIREMENTS:
if(condor_requirements) {
char *tmp = condor_requirements;
condor_requirements = string_format("(%s && (%s))", tmp, optarg);
free(tmp);
} else {
condor_requirements = string_format("(%s)", optarg);
}
break;
case LONG_OPT_WRAPPER:
wrapper_command = optarg;
break;
case LONG_OPT_WRAPPER_INPUT:
if(!wrapper_input) {
wrapper_input = strdup(optarg);
} else {
wrapper_input = string_format("%s,%s",wrapper_input,optarg);
}
break;
case 'P':
password_file = optarg;
break;
case 'S':
scratch_dir = optarg;
break;
case 'c':
consider_capacity = 1;
break;
case 'd':
debug_flags_set(optarg);
break;
case 'o':
debug_config_file(optarg);
break;
case 'O':
debug_config_file_size(string_metric_parse(optarg));
break;
case 'v':
cctools_version_print(stdout, argv[0]);
exit(EXIT_SUCCESS);
case 'h':
show_help(argv[0]);
exit(EXIT_SUCCESS);
default:
show_help(argv[0]);
return EXIT_FAILURE;
}
}
if(project_regex) {
using_catalog = 1;
}
else if((argc - optind) == 2) {
master_host = argv[optind];
master_port = atoi(argv[optind+1]);
}
else {
fprintf(stderr,"work_queue_factory: You must either give a project name with the -M option or master-name option with a configuration file, or give the master's host and port.\n");
show_help(argv[0]);
exit(1);
}
cctools_version_debug(D_DEBUG, argv[0]);
if(batch_queue_type == BATCH_QUEUE_TYPE_UNKNOWN) {
fprintf(stderr,"work_queue_factory: You must specify a batch type with the -T option.\n");
fprintf(stderr, "valid options:\n");
fprintf(stderr, "%s\n", batch_queue_type_string());
return 1;
}
if(config_file) {
char abs_path_name[PATH_MAX];
if(!realpath(config_file, abs_path_name)) {
fprintf(stderr, "work_queue_factory: could not resolve configuration file path: '%s'.\n", config_file);
exit(EXIT_FAILURE);
}
free(config_file);
/* From now on, read config_file from absolute path */
config_file = xxstrdup(abs_path_name);
if(!read_config_file(config_file)) {
fprintf(stderr,"work_queue_factory: There were errors in the configuration file: %s\n", config_file);
return 1;
}
}
if(workers_min>workers_max) {
fprintf(stderr,"work_queue_factory: min workers (%d) is greater than max workers (%d)\n",workers_min, workers_max);
return 1;
}
/*
Careful here: most of the supported batch systems expect
that jobs are submitting from a single shared filesystem.
Changing to /tmp only works in the case of Condor.
*/
if(!scratch_dir) {
if(batch_queue_type==BATCH_QUEUE_TYPE_CONDOR) {
scratch_dir = string_format("/tmp/wq-pool-%d",getuid());
} else {
scratch_dir = string_format("wq-pool-%d",getuid());
}
}
if(!create_dir(scratch_dir,0777)) {
fprintf(stderr,"work_queue_factory: couldn't create %s: %s",scratch_dir,strerror(errno));
return 1;
}
char cmd[1024];
sprintf(cmd,"cp \"$(which work_queue_worker)\" '%s'",scratch_dir);
if (system(cmd)) {
fprintf(stderr, "work_queue_factory: please add work_queue_worker to your PATH.\n");
exit(EXIT_FAILURE);
}
if(password_file) {
sprintf(cmd,"cp %s %s/pwfile",password_file,scratch_dir);
system(cmd);
}
if(chdir(scratch_dir)!=0) {
fprintf(stderr,"work_queue_factory: couldn't chdir to %s: %s",scratch_dir,strerror(errno));
return 1;
}
signal(SIGINT, handle_abort);
signal(SIGQUIT, handle_abort);
signal(SIGTERM, handle_abort);
signal(SIGHUP, ignore_signal);
queue = batch_queue_create(batch_queue_type);
if(!queue) {
fprintf(stderr,"work_queue_factory: couldn't establish queue type %s",batch_queue_type_to_string(batch_queue_type));
return 1;
}
batch_queue_set_option(queue, "batch-options", batch_submit_options);
batch_queue_set_option(queue, "autosize", autosize ? "yes" : NULL);
set_worker_resources_options( queue );
if (amazon_credentials != NULL) {
batch_queue_set_option(queue, "amazon-credentials", amazon_credentials);
}
if (amazon_ami != NULL) {
batch_queue_set_option(queue, "amazon-ami", amazon_ami);
}
if(condor_requirements != NULL && batch_queue_type != BATCH_QUEUE_TYPE_CONDOR) {
debug(D_NOTICE, "condor_requirements will be ignored as workers will not be running in condor.");
} else {
batch_queue_set_option(queue, "condor-requirements", condor_requirements);
}
mainloop( queue );
batch_queue_delete(queue);
return 0;
}
int main(int argc, char *argv[])
{
struct link *link, *master;
char *subject = 0, *type = 0;
time_t stoptime;
char line[1024];
signed char c;
int portnum = 30000;
char *hostname = 0;
int timeout = 30;
debug_config(argv[0]);
static struct option long_options[] = {
{"auth", required_argument, 0, 'a'},
{"port", required_argument, 0, 'p'},
{"host", required_argument, 0, 'r'},
{"help", required_argument, 0, 'h'},
{"debug-file", required_argument, 0, 'o'},
{"debug-rotate-max", required_argument, 0, 'O'},
{"debug", required_argument, 0, 'd'},
{0,0,0,0}
};
while((c = getopt_long(argc, argv, "a:p:r:d:o:O:", long_options, NULL)) > -1) {
switch (c) {
case 'p':
portnum = atoi(optarg);
break;
case 'h':
show_help(argv[0]);
exit(0);
break;
case 'r':
hostname = optarg;
break;
case 'd':
debug_flags_set(optarg);
break;
case 'o':
debug_config_file(optarg);
break;
case 'O':
debug_config_file_size(string_metric_parse(optarg));
break;
case 'a':
if(!auth_register_byname(optarg))
fatal("couldn't register %s authentication", optarg);
break;
default:
show_use(argv[0]);
exit(1);
break;
}
}
if(hostname) {
char addr[LINK_ADDRESS_MAX];
stoptime = time(0) + timeout;
if(!domain_name_cache_lookup(hostname, addr))
fatal("unknown host name: %s", hostname);
link = link_connect(addr, portnum, stoptime);
if(!link)
fatal("couldn't connect to %s:%d: %s", hostname, portnum, strerror(errno));
if(auth_assert(link, &type, &subject, stoptime)) {
printf("server thinks I am %s %s\n", type, subject);
if(link_readline(link, line, sizeof(line), stoptime)) {
printf("got message: %s\n", line);
} else {
printf("lost connection!\n");
}
} else {
printf("unable to authenticate.\n");
}
link_close(link);
} else {
stoptime = time(0) + timeout;
master = link_serve(portnum);
if(!master)
fatal("couldn't serve port %d: %s\n", portnum, strerror(errno));
while(time(0) < stoptime) {
link = link_accept(master, stoptime);
if(!link)
continue;
if(auth_accept(link, &type, &subject, stoptime)) {
time_t t = time(0);
link_putfstring(link, "Hello %s:%s, it is now %s", stoptime, type, subject, ctime(&t)); /* ctime ends with newline */
} else {
printf("couldn't auth accept\n");
}
link_close(link);
}
}
return 0;
}
